Watch this video from the stage at Adverting Week New York to learn more about identity, data, attribution and platforms from Dave Scrim, SVP, Product & Pricing, and Carl Madaffari, SVP, Database Solutions at Epsilon-Conversant.
You can also read the transcript of the 1 hour, 34 minute presentation below.
Dave Scrim: All right, welcome everyone.
Carl Madaffari: Hello.
Dave Scrim: All right. This is a back to basics session, if you're expecting Will Smith to walk out here in the wrong room. This is really broken up into four different sessions today, we're going to do kind of 30 minute modules, they get better and better so I'd encourage you to stay on. They do fit together into a story but they're also broken up into four components. I'm going to take half of them and Carl is going to take the other half, so my name is Dave Scrim. I've been working with Conversant Epsilon for almost 10 years now and we do kind of identity and measurements and all those kind of things, database management and prior to that, I used to work for a company called Experience, some of you guys might know, running databases for them.
Carl Madaffari: I'm Carl Madaffari. I've been with Epsilon Conversant for 19 years. Actually, this is my 19th year. I started off as a very technical person, managing databases, systems integrations, those types of things. In the last three or four years, I've migrated more to the marketing side, helping tell our story to our clients.
Dave Scrim: Awesome. There's four pillars we're going to talk about today as I mentioned. We're hearing a lot here at this session and they're all cool stuff, right, everything from voice activation, to AI, to social media, to addressable TV but really at the end of the day, you kind of need the pillars and the fundamentals to drive that and so this is the basics for some of those folks who are ... who haven't spend as much time in those basics areas and Carl and I have 20 years each experience in the space so we can go as deep as you want, we'll get to the Q and A but we've designed this at the base level for the New Years as mostly.
Dave Scrim: Also, it goes without saying, ask questions throughout. We have someone here who can take your questions through the app. What it's called, Slido? He can ... through the conversation, if you want ask a question or if you want to save them to the end, we'll leave five minutes at the end of each presentation. All right.
Carl Madaffari: Great.
Dave Scrim: Thanks, Carl.
Carl Madaffari: I'll step aside.
Dave Scrim: With that, I'll start with identity. Really one of the cornerstones of what you want to do and when we talk about identity, these are some players in the market, Signal, Neustar, Drawbridge, Epsilon and Conversant, LiveRamp. Just to get a sense of this audience, how many people are familiar with those names? Yeah, show of hand. Okay, right, so this is a pretty good audience. You guys knows what's going on. These are essential people for managing identity. Now, I'm going to define identity, I might not have the perfect answer to what identity is but this is the definition I'm going to be using throughout the presentation.
Identity resolution, what do we mean by that? The ability to accurately, super important word, and persistently, so over time, identify real people. Not just cookies, not just devices, not email address. Real people across channels and then the last word that if any of you are trying to do this today, you know it's a big problem doing that at scale, all right? As we dive into it, why is identity important? It really is the roots or the foundations to everything you're doing. If you're doing targeted marketing and you bought into personalization and you know that you need to get the right message, the right customer at the right time.
Well, how do you know that that's the right customer? How do you know that you are messaging the right person? What if you're getting the right message to the wrong person at the right time? This is why it's so commonly important and then, if you're going to measure anything and you're going to do close loop measurement and you're going to try to figure out what's working and what's not working, well, how do I know that the message I sent to this person is the person who converted maybe on a different device or a different channel? How do I know what's working if I don't have true identity?
We really think it's the cornerstone of what you're doing and if you did anything, first try to optimize your identity before you put a lot of special tools and additional things on top of it. Why is this so hard? Why is identity so hard, when we are running direct mail campaigns and sending mail and post cards to people's houses and catalogs, it wasn't that hard. We had their name and address. When we were sending emails, we might not have known exactly who the person was but we had to place to send them. The problem is there is no post office for the internet. There's not post office for digital.
There's no way to look up that Dave Scrim is this ID. We spend a lot of time trying to join the PII World, name and address and email with the non-PII world. What makes that super hard as well is we got to take people's privacy into account. We got to be super, super privacy because we haven't seen anything yet, as far as it goes on taking control and helping the customer be privacy centric and take the end consumer's space in mind. It's really hard to kind of manage this. One of the challenges is, cookies which was kind of the ecosystem of the digital world, they break all the time.
They break for all kinds of different reasons. People have antivirus, packages that automatically delete them. You change device, you change your browser, you go on a different device but cookies are always breaking. In fact, up to 40% of cookies are gone within the first 24 hours. How do you have accurate identity if you can't connect a person to a cookie on a consistent basis? Another problem is what we call clustering. In clustering, is when I try to overdo it. I can't keep track of these cookies so I'm going to use something else. I'm going to use a WIFI address at my house.
I don't know with you guys, but I have five people living in my home and any given month I have 25 people that have been on that WIFI. That's a real challenge because a lot of companies are combining those people, clustering and they're saying they're the same person so then, if I see an ad in my home, on my device, any one of those 25 people convert and that's called a conversion. You get into ... Again, you see the measurement trouble you get into. Another story on the targeting that that's a problem with, there was a media company and they were launching one of their big racy kind of shows, an adult show.
Not too racy but kind of in the HBO type framework. They pushed out an ad that was designed for people who were 18 plus and it wasn't ... there wasn't too much bad in it but all these kids all of a sudden got that ad because they'd all been at the person's house with a WIFI. There's a lot of complaints, it just shows you how audience targeting can be affected by the clustering as well. How do we get around that? How do we match to real people? There's a few ways to do that but I'm going to kind of try to go through a simple method and really foolproof and one that we stand behind. It's called deterministic matching.
One of the simple ways to do that is via transaction because when people buy things, they wanted to show up to their real house. They want to use their real name when they're buying something and so they don't use some fake email address. They're not just a cookie and they're not just device ID, they're giving you a name and address. What you do is, you take a customer order online and you associate that with a cookie. You have an order to a cookie, you have a match. Everyone does that day and night. At the same time what happens is that order goes through an offline transaction, all right?
It goes to the client called Signet Jewelers or call them Victoria's Secret or Home Depot. What happens is the name and address go to them and they strip that off and turn it into a number and you're onboarder strips that number into a common number across all clients. If I am the gap, Dave Scrim turns into the same number. If I am Home Depot, Dave Scrim turns into the same number. Now, that number is associated with the order ID. Over here, I've got a cookie and an order ID and over here I have an order ID and an anonymous number which maintains privacy. I join those two together and that's how you create a deterministic match because now, I have an anonymous name and ID which protects privacy, associated with the cookie.
I can do real targeting, real measurement and then the trick, we talked about how that deletes over time and gets erased and what you have you. If you're doing that across brands, over and over and over and over again, you hear about these ad consortiums that are trying to get together, that's what they're trying to do. That's kind of how you do deterministic matching. I think the big takeaway there is, when you're talking to your onboarder, talk to them about how many people they have behind those IDs they have, because we throw around a lot of numbers, a lot of match rates.
How many of them are backed by real people? Oldie but a goodie, on the internet no one knows you're dumb. This is literally from 15 years ago when I started in this space. Who are you matching? The main number you'll hear today if you work with most companies if you're like most people, you'll hear the match rate. My match rate is 40%, my match rate is 50%. I've got a match rate of 70%. Well, is that an accurate match rate? I talked a little bit about how companies can be better and accurate. See, you need to ask that question. A match is not a match. You need an accurate match.
The next question you say after you hear, "Oh my match rate is 80%," is can you tell me what the accuracy rate is of that match? How do you prove, how do I test your accuracy rate because what we've seen from a lot of providers is that accuracy rate is 60%. If I take my 80% match rate and I apply a 60% on top of that, now, I really only got a 50% match rate. All of a sudden the number is starting to fall down a little bit. The question to ask ... my big takeaway on this one is, a match is not a match. A match is match and accuracy and you should ask and they should be able to tell you how they measure accuracy and you should be able to prove that.
All right, that's a basic 101, you should be asking for. Scale. A couple of years ago, I think it's kind of gotten a little bit better now, I use to hear, "Oh my god, I've got 406 million consumers in the United States that you can target and message." It sounds a little off because there's not 406 million of us here and so you get these numbers inflated because everyone is trying to show that they have scale and the truth is you need to not only ask how many people base matches are because that number come down, right? You also have to ask what countries does it represent. Is this a match in Europe, is this a match in China, is this a match in Japan or is it a match in the United States?
Let's get our universe straight when you talk about how many IDs I have. The biggest point on this slide and the most important is ... so we talk about match accuracy, the reachable audience in a 30 day period, because if I match somebody to a Cookie that I haven't seen in six months ago, it doesn't mean anything. That person is not addressable. You need an addressable audience. You should be asking for a match rate, the accuracy of that match rate and how many people in that audience have been addressable in the last 30 days. That'll give you a good sense for who you can actually message to.
What I'm going to tell you is bad news, you then get another cut, you then get another hair cut of about 50% so now we're down from 80, we're down to 50, you're probably down at 25% now. Which leaves us to persistency. From a persistency standpoint, this is 10 second Bob, anybody watched 50 first dates? I love that movie. Ten Second Bob, says, "Hi, I'm Bob. Nice to meet you," and I'll say, hi, I'm Dave and what do you like Dave. I really like watching hockey and so Bob is like, 10 seconds later, Bob is like, "Hi. What's your name?" I was like, "I'm Dave." What do you like Dave? I really like watching hockey.
If think about it on the web, that happens all the time these days. You go to a website, you look at something, you leave and you come back three days later, either you have one of two things. Either they follow you around for two weeks and just chase you everywhere you go or they completely forget who you are because they've lost that identity. You log in on a different device, you're at your work versus your home. You're on a tablet, versus your mobile device. Your cookie got deleted. They don't know who you are and so they can't have that conversation with you.
The take away from this slide is you need to look for the persistency of that ID. In this day and age, when we're competing against big, big brands, mass retailers like Amazon, we need to build our brand. We need to build our relationship with our customers and the only way you can do that is by having a consistent conversation over time with your customers and to do that you need persistency and so all of this wraps into identity. A litmus test I use is you should be looking for something like, I can communicate over the course of the year with 80% of my audience.
I can talk to them on a regular basis. I'm not losing connection with 80% of those IDs. I'm going to talk quickly about a little case study, about a client we worked with at Conversant Epsilon and it's Road Scholar. They do adventure travel, education travel for adults. They get people in through all kinds of different marketing activities. They have a catalog, they bring them in but the real sweet spot is, repeat customers. They get most ... they don't get a lot of website traffic. They don't get a lot of conversions on their website. They got a lot of conversions from people who have been there before.
They came to us and said, "Look, we're not getting much traffic. We're having trouble reaching these people, we don't know where to find ... there's no directory to where to find our customers and they're not responding as much to direct mail anymore." What we said to them is, well, let's run a match test so give us a list of name and addresses and we'll see how many people we saw in the last 30 days. We saw 70 ... a more detailed version of this is available on our website, 71% of their offline customers that have bought before, we were able to reach online. Most of us people had never been to their website, ever.
We don't think about identity as the web based people or the offline people. They're really connected together. If you go a step further, of those people we messaged, 55% of the conversions were offline conversions. This could have been a call center conversion, they could have been a catalog or a write in or once he walked into the store but again, cross channel, we were messaging an audience who had never been on the website and many of them, even if they got their media online, converted offline and identity is critical to doing that. This is my favorite one of this, 28% of the people that did convert online, converted on a different device than they were messaged on.
We are messaging them on their mobile phone or we are messaging them on their PC or their tablet and they decided to convert on a different device. Now, I'll tell you that the biggest group of that is usually mobile devices. People get a lot of ads in their mobile devices and then convert somewhere else. We have a later case study that's going to talk a little bit about that. The quote from the client on this one was, "Conversant helps us reach an audience, we couldn't find before." If I go through the checklist, it's people, make sure you're talking to real people. Device IDs in themselves aren't real people. Email addresses aren't real people. I sometimes use ... I don't know about you I have a little fake email address. I have to sign up for a newsletter, I keep it over here, right?
Cookies aren't real people. They all have to be tied back and it's really, really hard to do. Make sure you can ... they have a way to show you accuracy and that you can test accuracy with identity. Make sure you can get it at scale and make sure that it's a persistent relationship over time. Those are the four questions you need to ask for identity. Nobody is perfect. It's a super, super hard thing to do and the best and smartest companies are trying to do it but there are better companies than others and so if you lay those four out and when you're looking at any of those vendors, you put them side by side, that's a great metric to use.
There is a more detailed blog post we have at conversant.com, blog which goes into, "Hey, look, if you're looking at a DNP or a CDP or you're evaluating a platform, here are some questions you want to ask." I gave you like four or five of them, has a lot more details behind it. With that, I'm going to pause for a second and see if there is any questions in the audience. Like I said, I try to keep it simple but if you guys want to ask more complex questions or less complex questions, that's fine too. Yes. We're going to ... You're jumping at third session here today, it's measurement.
We're going to talk about lookback windows for measurement but it's a great question. I'm talking about ... when I talk about 30 days, I'm talking about reachability. The ability to deliver that message but I think you're right, depending on the industry that you're talking about, that lookback window changes and we can have a whole conversation about lookback windows in the third session. All right. Yeah. Does it show up for me somewhere? What's the biggest misconception clients have with identity? That's a great question. That the match rate will let me talk to my audience.
The match rate from what I currently understand, you guys, tell me if I'm wrong is still the currency, is the word that's used, is the thing that's touted by clients. If I take that waterfall we talked about and I didn't talk about with persistency, if I really cascade that waterfall, the 80% match rate which most people don't start out at, probably more like a 60 turns into a 30% accuracy rate, turns into a 15% reach rate and all of a sudden the biggest problem is I can't talk to my customers and so that's one of the biggest challenges and misperception I think in the industry today. The other one is that anybody has got this completely figured out. Anyone that tells you they got it completely figured out, they're wrong.
New channels are popping up all the time, X-Box, Alexa, everything and so it's a moving target. Yes sir. That's a great question. Social, actually people based marketing is actually the best thing you can do for social because the walled gardens, one of the benefits they have is, they're one of the few people who actually deal in real people. Google and Facebook are for the most part, talking about real people. There's a measurement challenges there that we'll get into, the measurement section. If you could take your real people and port them over with name and address to this social networks, then it works, you're in a people based environment.
Again, there's a downside to that because there's a bit of a lack of transparency in measurement there that we'll talk about later but there's an upside. I would say they're the, one of the cleaner ways to do people based marketing.
Moderator: If you could give us a second and get the microphone?
Dave Scrim: Yeah, sure. I can hear you.
Moderator: We're live streaming.
Dave Scrim: Yeah.
Audience Member: First comes the money, "Hey, I've got a million dollars," and then great, let's do this. We do all this and then we have to go, whoever is this, yeah, we can only spend about 10 grand. That's it. Well, what do I do? I think how do we then work with customers? How do customers work with their partners then to understand that we have to do this legwork first to really understand how much budget can then go there because I just feel like what I'm saying, the money usually comes first then the project comes.
Dave Scrim: Got it. If I understand the question, just want to make sure I understand ...
Audience Member: How do we basically ... all this work, how do we then translate that into actual media budget? How do we then translate that better into how much we can actually spend against these uses?
Dave Scrim: Absolutely. What I would do is I'd have a little calculator and I'd say ... I'd backed into it and I'd say, so look if I get a million dollars and that's targeting this many customers, I'm only going to ... you're going to target with that million dollars, you can say, "Hey, I want to reach 10,000 people or I want to reach 100,000 cookies or whatever it is." You're going to need to understand, you're only going to be able to get 15 to 20% of that and we hear that all the time. I've got my budget, I want to spend more money, how come you can't execute?
You're going to take your little list, those four things and you stack them up into a waterfall and you're going to compare them across vendors and you're going to say, okay, if I really want to spend that million dollars and I really want to hit this many people, I'm going to have to start higher than I thought I did. Does that answer your question? Not quite. Okay. We do, I would recommend ... I'm really trying super hard not to do a sales pitch but I would tell you that if you go with ... let me put it this way.
If you go with the people base deterministic, those things that I talked about your fall off is going to be so much less than in some of those alternatives out there today. Companies like Google, Facebook, Conversant, Epsilon, if you do that, you're only going to get a 20% haircut and you're probably going to spend the majority of your money. Time for one, maybe two more if we have any questions. A couple on here, great. Thank you. I'm not used to this. Beyond asking the questions proposed, how do we hold vendors accountable for ensuring accuracy scale and persistency? This is down to the numbers.
I think you can't let up anymore that you can the performance that you're hearing, that you're getting back, "Oh, we had this many conversions, we had this many clicks. We had this many whatever," right? What is our match rate this month? What was it last month. Is it growing, is it shrinking? What is the accuracy rate, how are you showing me? This has to be a key metric for you because it drives everything else in your business. You can put the prettiest software on top of poor identity and you're not going to have an effective marketing program. To me, what I'd love to see and I'm closer to it than most people, this becomes a core KPI for your business identity. Good question.
How does identity work in the world of GPR and other privacy laws? You have to be privacy first. You can't get this wrong. It's not going away. What's happening in Europe, what's happening in California, it's the beginning, not the end. You have to treat people's privacy, their rights to what you collect, what you don't collect. You have to treat that very seriously and that has to be a part of the conversation with any of the clients you're working with. We have a privacy first, first. We tried to be ahead of the market on privacy and consumer privacy. We were the first company in Europe to launch a GDPR tool, to allow consumers to manage their consent and we're going to plan to stay ahead of that market.
That should be what you're hearing from whoever you're working with. If you're not using these identifiers available, cookies, email, et cetera, how can we identify our leads to target them with relevant ads? I like using this there. Whoever asked the question, what kind of identifiers are you using? Maybe that came in from offline. I apologize, I'm not sure I understand the question. Other identifiers you can use, traditional ones, direct mail, TV is becoming addressable. The world would become addressable, there will be ways to target and personalize across any channel at some point, even if it's not there yet today, 100%. Again, those four questions are what you need to focus on.
Carl Madaffari: On the different identifiers ... here we go. I know on the different identifiers. I know that there are traditional identifiers in keys that were built on top of traditional databases, those types of things. As Dave mentioned, we're working and the industry is working to bridge those together such that offline and online are tied together. It may not be in one key but with crossover between them, it allows you to link them over time. That is an evolving ...
Dave Scrim: Yeah, and with that, I'm going to, kind of switch gears. We're going to switch over to Carl. He's going to talk about data. When we originally set this up, they told us to be four sessions, people might be coming in and out, looks like we're good, we'll probably just go through one long thing and if you need a two minute break, you let us know.
Carl Madaffari: Yeah, it works.
Dave Scrim: Go ahead.
Carl Madaffari: Fantastic. Yes, as Dave mentioned, we got these four main topics. Identity, we're going to go into data next and data for us ... you don't need to see our ugly mugs there again. Data for us, this is a ... again, the idea here is that this is a session about introducing you to the concepts in this industry or bringing you back in. We've taken this down a level. We talk about data today, this is going to be the 101 version. This is the way I try and explain it to my parents as they ask what I do and they still don't understand after all these years. Data is one of those very ubiquitous terms. We all talk about data like we understand it. Data lives everywhere. It's in the nooks and crannies, it's in big giant chunks, it's in little pieces and it's become more and more important.
As we talk about that, it's important to understand who we're talking about in this space when we talk about data. On this slide, we obviously have Epsilon Conversant. We've got companies like Nielsen. Everybody is familiar with them. They're there to help track some of the traditional television ratings those types of things. IRI works very much in the CPG and retail spaces. Dun and Bradstreet for those of you who aren't familiar, has a compilation of data around businesses. When you're trying to solve a business to business problem, understanding who you're marketing to, who you're advertising to, data like that is very critical. ComScore is obviously there for other types of digital measurements and those types of things.
When we talk about data, why is more data critical? More is always better but there's risks in working with more data. The issue that we run into now is that the worlds of Adtech and Martech are coming together and both of those worlds use data a little bit differently and as evidence by Dave and I being up here on stage together, Epsilon and Conversant came from two different worlds. Epsilon was a traditional and we struggle with these delineators because the lines blur but we are traditional offline marketers, one to one marketers with email and direct mail and driving personalization that way.
Conversant was much more in the advertising space, the media space and they looked at data a different way as well but there's overlap and that overlap is starting to extend itself into this industry and it's something we found when we came together three years ago was that there was certain terminology, there were certain concepts around this data that sounded like we were talking about the same thing but we were just enough out of phase that it caused conflict, it caused confusion. What we're going to do today is we're going to talk a little bit about some of these at a very basic level and we'll let you through your questions take us deep as you need to going forward.
When we talk about data, it's important to understand what moves the needle. There is a lot of data and we're going to talk about these three types. We're going to spend a few minute on this slide and talk about the different types of data. First and foremost and attitudinal versus behavioral data. When you look at this, attitudinal data is the very squishy but very relevant stuff that lets you know how a consumer feels. This is data that you might collect through surveys, through preference centers where they tell you what subscriptions they want to hear about or what channels they opt into.
You get it through your call center data. You might get it through social networking where you start to see that linkage of who their friends are, who their influencers, those types of things. That's the attitudinal data that let's you know how somebody feels. It's very good data, it's very rich, relevant data but you actually ... if you're managing it yourself, you have to have an engaging brand to gather that type of data. Very few people ... and not to pick on anybody that works in the CPG space but very few people give a lot of feedback on paper goods. That's not going to get as engaging a brand as say, Axe Body Spray.
A lot of people have their opinions about that, one way or the other. Your engagement level as a brand dictates the level with which you can collect that attitudinal data. The other thing about that data is it's very hard to quantify, so when you take that data, it's great for measuring, it's great for when you look at something like say a brand awareness campaign, did it work? Did it drive awareness? Did it drive a certain feeling about a brand? Those are the types of things that you can sort of register a yes or no. Any depth or any quantifiable detail level gets much harder to do.
It's great data, attitudinal data but if you kind of take into an example of who maybe have this type of data, and it leads us to a transitional into behavioral is Netflix. Netflix is a place that when they started out, they asked me a lot of questions about what I like to watch. I might have said that I was very interested in period dramas or thought provoking documentaries. That's not really true, which takes us to the behavioral data. Behavioral data is not what I feel but what I actually do. Netflix as you all have noted, stopped giving me the recommendations based on the fact that I said, I like period dramas and thought provoking documentaries.
Gave me stuff that was based on what I actually watched which is just Adam Sandler movies and Avenger movies. Once they have that information, that is very concrete solid information. Behavioral data is the stuff that you gather from point of sale systems or the systems that collect and manage the dispositions of emails or ad servers or those types of things. It's concrete, it's transactional data so you're actually getting the data that says, hey, this is something that has happened. This is something that somebody chose to do. It's actually the behavior that we are trying to draw or something that we can tangential ... very concretely measure. Very powerful data.
The downside with that behavioral data is if you look at it in its own silo, is that it can be limiting and it can grow stale and it can be somewhat myopic in its focus. As we talk about the Netflix example, that is where that holy grail of putting them on an axis of how I feel and what I do is the type of data that works well in conjunction with each other. Attitudinal versus behavioral data, understanding that data as you collect and gather and prioritize it in your organization, so it's important to understand how it fits into your organization. Do you have an engaging brand, do you have the ability to collect this.
Those are the types of things to think about. Now, we move on to talking about first, second and third party data. This is one of those terms that I hear a lot of people talk about like they know what they mean and then they ... when you kind of question them, they don't really get it and it's really a very simple concept. First party data is the data that you as a brand, own and manage. This is your data. This is data you've collected about your consumers, be it attitudinal, behavioral, PII, non-PII. This is stuff you've collected about the experiences and the engagement you have with your consumers.
This is very, very powerful data. This is the space that Epsilon has worked for years in, it's where I personally grew up, in managing that consumer data. You heard about building 360 degree view databases. This was the start of that. The limiting factor in this type of data is really twofold. It's really hard to collect, there's a lot of it. It's builds, it evolves, it grows. The second part is that it's actually limiting in the sense that it is only data about you and your consumer. The experience you have, the channels you interact with, the transactions that they have with you as a brand, which is very powerful as you craft and deliver your messaging but it can be somewhat limiting.
It is limiting to their ... where are they in their overall lifecycle, where are they in their buying decisions and those types of things. First party data, very powerful but can be limiting. Second party data isn't really used very much but it's the idea of an exchange of data between partners. This is done in a lot of co-marketing efforts. A lot of efforts where you're maybe partnering with a brand to say, "Hey, if we work together, we'll share this data or we'll take data from my brand, what I know about them and give this to you in exchange for this or some sort of monetary exchange."
The problem with second party data, the reason it's not used very heavily is twofold. One, not a lot of people know how to monetize and put a value on the data that they have about their consumers. It's very hard to put a concrete number on it so it's very hard to make a transaction around that. The second reason is, it gets into that somewhat creepy area. That fine line between being a solid boyfriend who listens and that creepy stalker who knows way more about you than he should. That's second party data exchange, those fine lines are the reason second party data is not used as often but it is still a market that we think will evolve over time.
That takes us to third party data. Third party data is where you're going out and actually transacting. Buying data from typically a broker or compiled data sources. Data that is aligned to solve specific problems. In many cases, it's already been proven to solve that case so a lot of pros in third party data in bringing that together. Just like with the attitudinal versus behavioral, they all have their strengths, their weaknesses and triangulating across those is critical. The third area we're going to talk about is PII versus non-PII and I use this phrase like everybody knows what it means and inevitably somebody walks up to me afterwards and says, what does that mean?
I'll go ahead and explain it. PII is personally identifiable information. This is all the information around the name and the address. The contact information that Dave referenced earlier that made building IDs in the early days very easy. Non-PII information is everything that we know about an individual but that there are boundaries in place that you've got to keep that distance, that you can turn it into some sort of anonymize profile, that you can do your analysis on, that you can do certain decisionings behind walls but that you can't actively use it in that decision making process.
PII and non-PII, very critical parts, critical elements in bringing data together, about your consumer, helping you know who they are but different challenges and bringing them together throughout your organization. Where do you find this data. This data obviously lives all over the place. There's volumes of it everywhere. The three main sources we'll talk about today, your own backyard so you've got this data and the interactions you have with your consumers today and so this is where through either your own IT department or through maybe systems integrators, you gain access and collect this data about your consumers.
That is fundamentally where that first party PII centric data comes in, attitudinal behavioral data comes in. That is the first place to get that. Rich data, obviously it's great. The challenges as I mentioned earlier, it can be difficult to collect and difficult to manage over time. Getting it into a database is easy. Taking care of it, making sure it's evergreen is a challenge and has cost associated with it. Second source is compiled sources, data brokers, agencies ... I'm sorry, data brokers, list providers, those types that have built compiled data sources and deliver to you for a cost.
Again, as I mentioned earlier, this is data that is usually proven to have a good track record of the success and solving a very specific business problem and can just ... it just got to bounce out against the budget that you're working against. The third place is agencies. Agencies typically are sort of the best of both worlds. They take the data that you have in house. They take data from brokers, they apply a lot of analytic focus to it and prove out the value of that data set and did it work, does it apply to your business need and bringing that together. Between your internal, your own backyard, agencies and data providers is where you'll find that data throughout the ecosystem.
Let's talk about some of the challenges in making the most of your data. There's really three fundamental challenges that we talk about in this. First and foremost is that data lives in silos. It's inherently ... It's an idea that as a database administrator growing up, I envisioned this world where all data lived in this Utopian giant database that I can access through SQL or some sort of neural implant and that's not a reality and as much as I'd love it to be, it's not going to be a reality anytime soon. Mainly because this data is voluminous. It lives everywhere. It grows very quickly, trying to transfer it, bring it all together for no purpose is a big challenge.
That'll take us to our second challenge in a second. The other reason is, it's very utilitarian, it's very effective where it lives. Trying to move it and migrate it adds overhead to the management of that data, that makes it very difficult. What we recommend to get across that is something that Dave talked about earlier, is that identity layer. How can you link the data across those silos and have that decoder, if you will, that lets you know that the data about this person and this silo is tied to this person in this silo through identities, through keys, those types of things, is a much more effective way to do it than to try and do what is our second challenge, which is boil the ocean.
Trying to bring all of your data together in one ecosystem, I have been part of these projects. I have led and somewhat successfully gotten them off the ground but this idea of I'm going to build an enterprise data lake or an enterprise data warehouse that will be the end all be all monument to technology, that we'll never have to replace because we've done it so well this first time and those are really, really hard to get off the ground because you're trying to fight a lot of fires. You're trying to build for a lot of different challenges, some of which you can see. Most of which you can't and at the same time the technology is evolving and leapfrogging past what you can do today.
What I've seen a lot of cases in this data management mindset is, I'm going to stop down our business for nine to 18 months to build something that by the time I get there may not even solve my business challenges or worst yet, is I've seen companies get locked down in that sort of analysis paralysis of trying to decide what am I going to solve first. Boiling the ocean in this case, deadly, dangerous and then the third challenge ... and this mainly applies to the do it yourselfer is the regulatory concerns. This is an evolving space. This is something that as people become aware of the fact that when systems like Facebook are free, you're not really the consumer as much as you are the product.
Rules are evolving. Privacy rules are evolving and they're evolving in different regions and locations. What we have found is trying to do this yourself, micromanaging this is an ongoing battle, recommendation is obviously to work with somebody who has resources and expertise in this space. We'll wrap this section up with a case study about discovery communications that came to us several years ago. This is a case study that brings together all of the data that we've talked about. Discovery had a challenge. They were trying to evolve their audiences. They're trying to reach the right audiences. As tune in for the TV shows was dropping.
They were looking for ways to spend media dollars effectively and reach the right consumers, where it would be resonant and effective. They came to us with ... and one example of panel of say 10,000 households of people who actually watched their show. We took that along with our compiled data source and extended that out. We found lookalikes, people who had the same demographic and behavioral and attitudinal data, that expanded that audience out to say a million. We then took that million and took it to the digital ecosystem with Conversant to say, what are their online behaviors?
How do those look like the same consumers and expanded that audience even more and found audiences that we could reach, that the ads for these shows would be effective based on that information. What we found was ... in our first year that we launched this, they had ratings of 30% over prior season premieres and the other thing we were able to do is deliver this in a much shorter timeframe than they used to with a much smaller set of data. This used to be about a 12 week measurement turnaround. We turn this around for them in about four weeks. Again, part of that was all of that data lived in different silos but we had the ability to link it together, tie it together for purpose driven solutions to solve problems specifically for them.
When you are tasked with managing the data within your organization, few things you should think about. First and foremost, prioritize what you're focusing on. You can't get stuck boiling the ocean. Decide what problems you're trying to solve for and understand what data works in your industry and doesn't. This is one of those cases where I'd love to be able to tell you that, "Oh, yeah focus on behavioral, focus on attitudinal, focus on second party, third party." Each industry and each business has their own specific needs and knowing that will help you prioritize where to focus.
Second, centralize only what you need to, to drive that effectiveness. Don't boil the ocean, don't get stuck trying to build a monument to technology. Third, it's always important to measure and this is a nice transition into Dave's conversation here in a second. Measure not just your programs or your offers or the content but measure the data because every piece of data you manage, costs money to take care of. When we first started building databases for our clients 20 years ago, marketing database maybe had 50 attributes. We built one for a client recently, it had 3,000 attributes on any given consumer.
I know that even today, no more than 50 to 100 of those attributes are relevant at any given time but everyone of those attributes has a cost associated with it from managing the metadata around it, managing the interface agreements, the exchanges, the understanding of what that data is, all has a very high cost so measure what works, not just in terms of programs but in terms of the data and the effectiveness and finally, be prepared to evolve. This is not a one stop shop. There's not one universal project that will solve this problem for you, the data problem.
It will constantly grow, it will constantly evolve and our recommendations are to treat it that way and be prepared for this to be a lifelong pursuit to understand your consumers through better data. I'll stop there to see if there are any questions on data? All right, we have one here. I have heard the term data leakage used a lot recently, what is it and how do you prevent it? I have actually not heard the term data leakage in this regards. Does anybody here in the room that can ... Yes, come on up.
Dave Scrim: There's nothing more powerful to your business than data about your customers. You got to be protective about that data. You got to make sure ... I mean, there's a big ... anybody want to name one of the biggest companies out there that's responsible for data leakage right now? You all got it, right? Everything you do, every campaign you run on that company is sold. In fact, there was an article, I couldn't believe it, I read in two days ago or on Thursday, two business days, about the fact that your backup phone number that you're using in Facebook, you know, contact me if anything goes wrong. They're selling that and monetizing that.
Data leakage is about understanding when you're partnering with somebody, when you're allowing your data to be accessed, your data ... valuable data is brands and is customers, what happens to that data and where does it go because there's a lot of people who want to take that data and want to leverage it for you. That's a little bit about data leakage.
Carl Madaffari: The other thing I'd say about that is kind of what we talked about earlier, just because you can collect the data and manage it doesn't always mean that that's valuable data. To your example there, having that data sometimes that data gets misused so it's another point to managing only the data, centralizing only what's important especially from a marketing and ad perspective. Great question. Any others in the room. All I can see is bright shiny lights. All right.
Dave Scrim: Awesome.
Carl Madaffari: Fantastic. We'll jump ahead to ...
Dave Scrim: Measurement.
Carl Madaffari: Measurement.
Dave Scrim: This is a big one. Actually, it's called attribution and so I'm definitely going to spend some time on attribution but I'm also ... I'll open it up to measurement. We can take testing control questions. We can take how do you measure search. I want it to be ... this is a really tough space. Anybody feel like they're doing perfect measurement today in the market, anybody doing that? All right. We're not. I was surprised if ... most people I ran into this one is their biggest problems. How do I measure what's working, what's not working? That's going to be the gist of our presentation.
A couple of other level setting questions. How many people measure on clicks today. Are there still some click based folks out there? This is an advanced group. This shouldn't be called one on one. We should have up our game. They're still a lot of folks by the way that are click based measurement. A lot of folks out there. It was ... when it was the only currency out there, it was fine. There's a lot of better ways to do it today. I'll throw up a few companies. Attribution vendors, because the gist is attribution. Google is now a big attribution vendor. Facebook has its own attribution solution.
Bizible, Brightfunnel, Conversant, Epsilon, Visual IQ. Who uses a third party attribution solution here, can you raise your hands, anybody use a third party? All right. Anybody use one of these companies, anybody? Okay. I'm going to pick on somebody. How are you doing measurement today? You, yeah. You don't do measurement. Anybody, anyone who can tell me how they're doing measurement? Yeah, go ahead. Yeah, just taking the data and do it yourself. Yeah, higher media mix company, something like that, maybe a third party analytics company do it. Great, okay. There's a good reason why we don't have a lot of folks in here using these measurement.
Nobody is happy with the measurement providers out there. I'm going to get to that in a little bit. Before I start on that, what do I mean when I say attribution? What am I talking about and we can debate this but I want a level set. It's as simple as taking a marketing event. Some marketing event, an email that was opened, an impression that was sent to somebody, an affiliate message that somebody clicked on. An impression that was set on a mobile device. Something, some marketing event, a direct mail piece, right? Some marketing event that was sent to somebody and a conversion event.
That conversion event doesn't have to be a sale. It could be an online sale. It could be an offline sale. It could be a website visit, it could be sign up for an email or sign up for my loyalty program. Attribution is about connecting a marketing event to some kind of conversion event that I want to take place. Easy as that, pretty simple. What's the problem? Well, one of the problems is that most of the institutions still works on what's called last touch attribution. I say well, I did an email campaign, I sent a display ad to a PC and I sent it to a mobile provider but I was running an affiliate campaign and they clicked on the affiliate campaign and they converted and therefore the purchase goes to the affiliate provider.
That's called last touch attribution. The next one is, multi-touch attribution. More complicated and you could see why people have trouble especially now you're starting to see why identity might matter in the solution, why persistency might matter in keeping things together because multi-touch attribution tries to say how does different activities combine to drive a sale? You put a weighted score on the different channels. You look at the consumer's path to conversion so that you know where to invest so you know what's working and know what's not working. Somebody opened an email and then they saw an impression on their mobile device with media partner A.
Then, they looked on their computer and got an impression and then a click and then they visited a site and either did an online or offline purchase. That is called multi-touch attribution. All right but in order to do that, you need to see the full cycle of the person. Then, there was a question early in the audience about lookback windows and in fact, I'm going to stop after I do this next slide and we'll go back and see if this answers your question or if you had a different question on it. Lookback windows. A lookback window is how far do I look back because I might have been messaging that person for five years.
I'm not going to take into account every message in every channel and every contact I had with them for five years as part of the purchase cycle. I'm going to say, what's the reasonable timeframe I should look back on all my marketing activities. If I'm Walgreens or CVS, I might only look back seven days because the purchase cycle of convenience pharmacy is really quick. If I'm a retail clothing manufacturer, I might look back 30 days because that's kind of the decision cycle that it go through. If I'm selling cars, I want to go back months in time, three, four, five months.
It really depends on what the purchase cycle is of your industry, in order to decide what your lookback window could be and that's a really flexible thing and something that people should play with and figure out. Now, did I answer your question about attribution or did you have a different ... about lookback windows? They want more credit for the sale. One of the key things and I'm going to say this to the end because it's going to take away us in the flow but I want to come back to it. I want to encourage you because this is mostly about attribution. I would encourage you all where you can to use ... there's a classic thing that's been around since direct mail, we kind of forgot all about in the digital world.
It's called test and control. It's called creating a group of people that I'm going to send an email to and a group of people that I'm not going to send an email to and looking at the difference in sales between the two. It's about taking a group of people I'm sending a display ad and not sending a display ad and looking at the difference in the sales and creating a hold out group because you're exactly right, because for any of you in this room that have added up all of the conversions that your email company tells you they got, that your display company got out. If you got multiple escape companies, you end up with five times the sales, you actually have sitting in your conversion in your business.
A lot of people are claiming sales and that's why you want to centralize, you want to use testing control and you want to bring that all together to an individual level. So we can unpack that a little bit, that's a little ... not too much, it's a little more sophisticated so we'll save that one if we want to go deeper into that in the end. It can vary by type. It can vary by channel, how far you want to look back for attribution? Why do I care about attribution? The number one thing is I want to understand what media is working to drive budget allocation. Do I spend ... is this digital media provider working better for me than that one.
Is this website working better for me than that one? How about my channel, should I invest more in email, should I invest more in mobile, should I invest more in direct mail, should I cut back on direct mail? How do I invest my marketing budget in channels, how do we get the most leverage possible out of my marketing dollar? The good news is, it's more measurable than it ever has been, especially in digital today. The bad news is it's super complicated because of some things we'll talked about. We're going to talk about that in a minute. Understand the customer value ... sorry, customer journey path, what path do people take to purchase, what steps do they take along the way.
Again, so you know where to invest. Understand the customer value of the customer by channel or by conversion path. Maybe the people who are on mobile are more or less valuable than the people that are only answering emails. Maybe it's the opposite, right? How do I understand the value of the customer, with those channels? The thing is, attribution is really just math. It's not that hard, you come up with a lookback window. It's seven days, it's 30 days, you come up with a fractional attribution. I'm going to give 10% to email. I'm going to give 30% to display ads. I'm going to weight that a little bit, depending on how long each of those were but it's just a math formula at the end of the day.
Why is it so hard, why is the net promoter score for attribution vendors negative 29%? The net promoter score is don't work with this company. I don't care which of those companies it was out there. Nobody is happy with attribution vendors. Nobody is happy with their attribution. It's messy. It's messy because there's a lack of trust. It's messy because there's a lack of transparency. It's messy because data management is really, really hard. I'm going to talk about each of these in a little more detail and as we learned earlier today, it's messy because identity is super, super hard.
We'll just going to go into that in little bit more detail. Trust. I put up a bunch of names up there and there's only one of them that's still branded as the same company that was Visual IQ. There was Visual IQ, this is only two years ago. Convertro. Adometry, Datasong. These were independent attribution companies that said, "Hey look, I know you're working with Google. I know you're looking with Facebook. I know you're working with Conversant. I know you're working with whatever, media company. I'm going to look across all those, I'm going to help you with this. I've got some great math and I'm going to help you understand what's working for you and what isn't."
As a standalone company, none of those companies exist today. The only branded company that exist today is Visual IQ but it's not independent. These companies are owned by ... Visual IQ is owned by Nielsen, who's in the media business. Convertro is owned by Oath, who's in the media business. Adometry is owned by Google, who's in the media business and Datasong is now owned by Neustar who's in the media business. A little bit of fox guarding the hen house. How do you trust the providers, how do you find an independent provider? Well, the truth is, these marketing platforms are really complicated and they take a lot to run and that's why some of these companies have been absorbed.
You can have trust if some of these companies are doing it, if you have transparency. This would be the number one ... my number one takeaway from this whole session is not how you do the math, because you can argue about that. It's not who to trust because trust is earned. It's that trust comes through transparency. I'll use an example of ... As I was taking the cab from the airport today or last night, I was thinking about the old days when I used to take the cab to New York from the airport and I was sure that person was driving me in circles all the way around. I didn't know where it's going.
The meter was going up. You might be right, you might be wrong. I didn't have trust and I didn't have transparency. What came along today on the ride this time and I don't ... the guy was a great guy. I said, I do not have any reason to double check him. Let's say I was nervous. I could have pulled out my Google Maps. I could have seen the right path to go and I could have directionally had transparency that I was going in the right direction. I actually think attribution is hard enough that you do want to work with somebody else, whether it would be one of these big partners.
Whether it would be a small boutique shop who takes in data and analyzes it for you but if you are a larger organization, you can certainly have the analytics people to do it yourself but the key, the key to it is ask for transparency and that means, you're saying, give me the data. Give me all of the conversions you saw, give me all of the impressions you saw across all of the channels and let me validate that myself. I might get you to do this on a regular basis, but occasionally I'm going to test you. I'm going to spot check this. I'm going to have another party take a look at this data.
Without the transparency, then you're just taking their word for it that the results are right there. There is a real lack of transparency industry and I think that's the number one thing, we all need to drive for. I'm not saying anyone is out to cheat anybody but they're also out to make their solutions look good and you validating that is an important part of your business. These are two of the least transparent organizations and they are the two of the biggest organizations out there. We all should be doing social media, it's a great channel.
It's where our customers are and yet even with them, you need to drive for transparency because what you're going to find at some of these companies and you probably know if you're using them, you're going to get aggregated results. You'll get nice reports, you'll get some performance goals and stuff like that but you're not going to get the raw data. My big question would be why, what's anybody afraid of. It's working, it's working. Just show me the individual level results of those programs and that would be the question and it's going to take a lot of time.
It's an industry problem, it's going to take the big players, pushing, and pushing and pushing but we got to keep doing that. Data management, this sounds about as dry ... this is about as dry as it's going to get. The data management is how do I collect that data in the first place? Hi Stacy. Stacy is the data expert at Epsilon Conversant and runs our whole data business. She is the one to go to for questions after the show for you, if you have that. Data management. How do I collect data, it seems easy, right? Just send me over a file and I got ... I know how many people I direct mailed. I know how many people I emailed.
What about mobile apps? How many of you are doing a really good job at connecting which individuals visited your mobile apps? What about tagging all the different media companies that you're working with and having those tags on the internet flow into one individual place across all the different vendors. What about your Facebook and your Google, right? How easy is it for you to do data management? According to Forrester, it was the number one criteria of importance when people are evaluating attribution vendors. Can they even collect the data in the first place?
That's a super, super important question to ask. How do you collect data, can you collect it on all the channels? Can you collect the search data, the paid search data that you're running against? There are ways to do that. I don't know if you guys know but through AdWords, you can actually identify who the individual was that hits your site from your paid search thing. Ask your vendor, can they do that? Can they capture that? We can talk about that later if you want. Identity, again, most of you here were in the first sessions so I'm not going to bore you with that again.
Number one thing so I grab the data, I pull it all in. I have my math but none of it matters if I don't have the accuracy persistency and scale to identify this at an individual level because all it says is, I put 25 cookies out and I got a conversion. Well, did those 25 cookies go to 25 different people or those 25 cookies all of the same person? You've got to come up with a common denominator around identity. Client success story. This is a retail brand in the clothing space, 30 days as we talked about attribution window. What they found was, they were running attribution and their mobile conversions were less than 1% and the division head of the mobile unit said, I got a real problem.
They're divesting in my business, they're not funding us. They don't think mobile does anything. I'm telling you, I know, I've been out there, I talked to the ... it was an 18 to kind of 25 year old brand. The kids are all on mobile. I don't get what's going on and the truth was he couldn't track it. What happened was, as he migrated to a system that could track identity, could track persistency, could track all across these channels, he found out that prior to a conversion, he might only be getting 1% of the sales on their mobile device but 42% at the time, people were seeing mobile ads and consuming the media and getting the brand consideration on their mobile device.
Huge change within the organization. Huge shift in understanding about how mobile affects the sales process. That's just one little story. You'll all have your own stories but it's super, super important and shows you the difference between, where you might invest your money or not. You must have checklist. It's just math at the end of the day. If it's just math, then demand trust and transparency. Make sure you're getting the best data management and you're able to capture the data from all those channels. Another thing I didn't hit was you get some people in the space who were good at online data capture and some people that are good at offline data capture.
It's a very rare combination to find a company that was designed to do both. I would keep looking for those companies. Epsilon Conversant, that's one of the unique and powerful things of Epsilon and Conversant so a machine was plug there. We've been doing offline forever and we've been doing online forever, now, we've kind of put them together. Then, make sure identity is correct. Again, we'd point you to that blog, more questions. The questions are aligned around measurement, attribution, identity data, the questions to ask at that blog that we talked about earlier. With that, if there's any questions, I will take them now. Yes sir. Yeah.
Audience Member: How do you get to the right combination mostly when you have non-transparent channels like Facebook for example and Google but mostly in Facebook, that you'll not even get the order ID to be able to max the data.
Dave Scrim: Yeah, if I could solve the Facebook attribution problem, I probably wouldn't be standing here right now. It's the biggest problem. Social is so important to us but it's such a black hole and I wish I had a magic answer for you. It's the one channel we can't. The only thing I can say is look, you're going to ... remember the days when AOL was the only thing in the world or Yahoo was the only thing in the world. Well, more pressure is going to come, more players are going to come. Amazon is getting ... What we saw upstairs, Target has got a big ad initiative.
The more pressure, the more competition that comes to the market, the more those companies are going to be transparent with their data, the more we're going to see and force everybody to be transparent with the data. I would say, you got to put your money where your mouth is. Yeah, you put some dollars in but maybe don't put as many dollars in and I will tell you, the biggest clients we have are making some dense, are getting more than the average person, is starting to lead the way in putting pressure on this companies. Yes.
Audience Member: The campaign managers of the ad server, they've said they're not going to release event level or ... yeah, event level data that's required for attribution.
Dave Scrim: This would be ... is that Facebook campaign manager?
Audience Member: Google.
Dave Scrim: Google Campaign manager. You are ... you're just at the mercy of their attribution. You're going to ... if you're going to use them, you're going to have to not have individual level insight to who is being measured or not. You're going to have to wonder if you were here in the identity session, if there's clustering going on or not and you're going to have to take the word for it. That's the only answer I could have on Google and Facebook. Yes.
Audience Member: When you talk about transparency, with some of the big players besides Facebook and Google and you say allow for spot checks. I mean, just the hurdle in getting everything where you talked about data management, getting everything to that, getting everything flowing. How do you do the spot checks?
Dave Scrim: That's a great question. What I would do, if I was a mid-sized company and I had mid-sized resources, I would say, look, I really trust you, I went through my checklist, you passed most of the things or at least you're better than most people out there, I want you to kind of do my attribution for you. I don't have a ton of resources, I'm going to do but I what I want you to do is I want you to give me a data feed. I don't want you to send me that data feed every week. It's going to have the impressions in it, served across all media, whether that would be my media or somebody else's media. I'm going to send you my direct mail and email files and you're going to send those back through the impression feed.
It's going to have all the conversions across any channel in it and you're going to send me that feed each week. Now, you might not do anything with that feed for week one, week two, week three, week four, week 20 and then one day, you hire a third party, a nice little analytics firm, you say, "Hey, here's this data feed. Could you validate what we're hearing from this company and just double check that everything is okay?" One of two things is going to happen. You're going to have a big discrepancy and you're going to need to get everyone in the room and start talking about that or things are going to look directionally correct.
It's okay, if they aren't perfect. They're going to look directionally correct and they're going to go, "Okay, I think this is working. I think this is fair. I double checked that person with somebody else and I'm going to move on now for another five months or six months. I think that's the way I would do it." Yes, in the back.
Audience Member: Advertiser who's run the marketing mix, so I'll give you actually, a real example. Trying to break into Nielsen's Black Box and I'm getting ... asking some of these questions but understanding what are those questions I'm not asking. Their panel doesn't even match the data and the targeting we're selling so we don't even know what that is, where ... our uses are mobile first so we already calculated a 70% drop off of the measured media.
Dave Scrim: Yeah.
Audience Member: Do I ask for data feed as well? How do I recheck the actual sales that I know are modeled on modeled data?
Dave Scrim: Right. It sounds like you're asking some of the right questions but you're not necessarily breaking through and one of the suggestions I would have is, find a third party to help be your advocate. Find a firm who's doing this on a regular basis and if you ... it doesn't have to be us, right, if you want to ask after we can give you a few names of places or it can be us. I mean, we do this all the time and we help companies with the right questions and like I said, there's a list on the blog but there's a list of bigger questions and we can put a little pressure on those. What I can't promise is, I can't promise, we're always going to get the magic answer.
I can promise we'll get to the right questions or we'll get some more data probably than you do but again, they're called walled gardens for a reason, right? Nielsen, maybe not so bad. Nielsen, I think we ... we know a lot about Nielsen on ROI where we can help you a little bit more of that. Any other questions? All right. Thanks so much and, at the end I'll be around if you want to talk anymore.
Carl Madaffari: Great. Well, we are in the home stretch, moving along quickly here. Our last section is platforms. I'll jump ahead here and give a few people, if you want to switch sessions, you're good. We're going to talk platforms now. Platforms is sort of the final of these pillars and it's an important one. What we found is over the last ... in my 20 years in this industry, there's a pendulum that swings back and forth, between the, we need to own this ourselves internally and manage this data, this orchestration, this content and switching in the other way which is, well, we need to just outsource this.
This is too expensive, too costly, we can't keep the resources in house so we need to outsource this. Then, the problem becomes, well, I've outsourced this to some agency that I now can't see what's going on. Now, I need to pull it back in house. That pendulum in my 20 years has swung back and forth. A couple of things have happened over the last five years that have really led to the advent of the platform. Obviously, the cost of storage and housing and horsepower in cloud computing has been universal in helping all of these. There's a couple of big changes, am I cutting it out there?
First and foremost the big software providers years ago started acquisitions. They started buying in pieces and parts and stitching them together into a larger ecosystem. The second thing that happened is that these big social media networks went from Farmville to Madison avenue. They sort of realized that they had this audience, they had this group of engaged individuals and they found a way to sell advertising on that. Then, the third thing is that an advent of venture capital has started to apply platform technology. The problems that were ... have been around forever but it's just always been solved in a very manual focus.
Those three initiatives have led to this advent of age of the platform. The problem with the platform is that there are a lot of players. In this case, we're just listing a handful, the Salesforces, the Adobes, the Oracles. There's obviously the IBMs, the Facebook, all the social media providers. Epsilon and Conversant is sort of a different breed and that we kind of come from the ... we were a services organization and agency, if you will and we are pushing to be more platform oriented. To take our intellectual property and our experience and package it in such a way that we sell it as platforms. You have everybody converging towards this software as a service model. It becomes a problem, it becomes a struggle because they all have an overlap.
We're all sort of playing in the same space. In fact, if you take all of our story and you turn it to the elevator pitch, it's, we help you understand to know your consumers and reach them across every channel. That's the story of everybody in the platform page, whether it's in detail or not, it's the same story at the highest level. How do you get through that? How do you solve for this? How do you pick the right platform? I'm going to just roll really fast ahead. There's not one platform that solves it all. I mean, I would love to tell you it's us, it's not. There's not a single platform that does it all and it's the single easy choice to make that will get you promoted to be a CMO.
There is a lot of things you have to go into the decisions around your platform. First and foremost, can you articulate what you're trying to do? Each of these platforms that we discussed has a different strength, has a different focus, came from a different set of DNA that means it solves problems better than others or maybe it started at a point where you got to leapfrog certain problems. Understanding, what you're trying to do, what you're trying to achieve is important because the other problem you get is you will get bombarded with the big stack, it can do everything and anything and if you buy one tool, you'll get four more thrown in and that can lead to a lot more confusion down the road.
Can you articulate what your problems are, what you're trying to solve for. The second is, can you prioritize that, because the reality is, as we talked about earlier, these projects take a while to get stood up. Even the platforms that are multichannel marketing hubs that have it all put together, are still built out of acquisition and some of these are built out of 10 to 12 acquisitions that have been stitched together or are being stitched together after you close the deal and stitched together behind the scenes, by companies like ours to make them work together.
Being able to prioritize what's important, three months, six months, 12 months, 18 months down the road is important as you make these decisions because as we talked about earlier, technology is constantly leapfrogging and evolving. A player in a certain space may acquire somebody that makes that solution something they could do themselves and yet, you've already purchased something that now has to be stitched together. Prioritizing is key. Then, the third, and this is really the most important as you look internally, is are you building around an organization or are you going to shape your organization around a new vision.
A lot of tools and a lot of people where I've seen go wrong in the platform decision is they buy a platform and they try and make that platform fit a broken model and in many cases, I even see this broken model shape how the decision for platforms are made and cause of the problems. You have the VP of email marketing, making the decision on the platform they're going to use for email and the VP of measurement and analytics is going to choose their partner and somebody else is going to choose one for web analytics and media and now you're dealing with different pieces and parts from the different platforms.
Are you trying to work around that? As you get into the level deeper, the orchestration of those audiences is now going to become more complicated who approves this audience being used in this channel. Those are things that technology by itself and these platforms will not solve for, you as an organization have to be ready to understand that there are going to be changes made around that to fix them. The other big challenge in these platform decisions are what's in place that works for you today? Either because it works really well or because it would be difficult to replace and that shouldn't be underrated or third because you've already got some sort of long term engagement.
Some sort of licensing deal already in place but the provider that can provide that functionality and you or somebody in your organization is incented to get the most and leverage that platform. Those deals are signed and sometimes they sit idle for two, three years on the back end of those long term deals and those platform decisions ... your best bet is try to make the most out of those while you've got that licensing in place. A lot of decisions you got to consider, that then takes you to the ... which of the platforms, what does the landscape look like? We talked a lot in this about the marketing clouds, the big software giants we talked about.
What I would say in this space is that, this is why it goes back to understanding, what it is you're trying to do and what you're prioritizing. Each of the clouds has a strength, a position they may have started from or an acquisition that was foundational. Adobe is really solid in web analytics space and in managing content. IBM is obviously fantastic at their analytic capabilities and some of their orchestration. You've got Salesforce which is great taking from the call center and understanding those consumers. Each of them has a strength and each of them has tried to bolt on certain components to give them the complete stack.
Sometimes, it was a game of musical chairs and some were left without an E-commerce platform or others are finding strategic partnerships. One thing that I would note is ... an announcement came out last week and I think you're going to see more of this. Adobe announced the strategic partnership across, I think it was SAP and Microsoft. An open API set that allows for the exchange of data and information between these three platforms. I think you're going to see more of this as each of these software giant sort of realizes that no one is buying their big black box of technology and landing it in a data center and dealing with that as their sole tool set.
Everybody is working around the different tool sets. The other thing that's important to note when you're buying, when you're looking at marketing clouds the one thing that they don't tell you about is that for every dollar in licensing, they look to spend between four and eight dollars to either integrate or manage on an ongoing basis those platforms. A lot of times, the salesperson for those platforms is in there telling the story about hey, look, this tool is going to solve this problem and it's very compelling but the problem is there is another four to eight times that spend in managing that. It's something that you got to understand and prepare for.
Moving on walled gardens. We've talked quite a bit about them. I think we've kind of kicked them around a little bit and the reality is there are strengths and we see a lot of clients, depending on the problems they're looking to solve, maybe just fine working in this space and saying, look, I'm going to take my consumer base and I'm going to fish where the fish are. I know where they're at. I'm going market to them and I'm going to trust that there ... that that's effective and for several clients, that's all they need and that's where they spend and that's where they spend a bulk of their spend and that works for them.
For everybody else, and as we've heard today and I heard in this room, very few people are looking to put all their eggs in that one basket. Looking to exchange, looking to work across that is where most people are looking to take their platform investments. It's something to keep in mind as you consider them. The final frontier ... and this is sort of that new area that I talked about in the introduction. This is that idea of customer data management. This is an evolving space. This is really my sweet spot, this is where I grew up. For Epsilon, we built these as one off custom databases.
We use to start working with a client in a blank whiteboard and we'd say, what do you want to do? Great. What data do you have and we'd evaluate that and we'd put together. More and more companies are looking for ... the other thing about that, that engagement was, everything we go into a client even in a particular vertical, we would say, look is there somebody else you want to model your business after, somebody else you want to model your marketing? We were always told, "Nope, we're unique, we're different. We do things completely different than anybody else. This has to be a custom solution."
Nowadays, over the last two years, CMOs and CEOs are saying, if you can give me 80% of the functionality, out of the box and prevent me having to build and manage my own monument to technology, I would rather do that. This is where about three or four years ago, this idea of managing it yourself with customer data platforms, customer data management platforms and there's a huge list of different types that you see up there, that are all up there competing in the same space and they're all saying the same thing in a much smaller sphere around understanding your consumers.
This is a report from Gartner that came out earlier this year to help people understand what CDPs are. A year and a half ago, no one knew what a CDP was. Now, everybody believes, they have to have a CDP. The reality is when you look at what platforms do ... and this is really important. I'm not going to go through everything on the left side there but I think it's important to talk about the categories on the top. Across the top, you've got the idea of data collection. This is that idea of bringing data together, putting the identity together to that, keying it, bringing it together.
Fleshing that out with profile unification, which is a deeper level of that ... that identity resolution. Taking your online and offline, matching it together. That takes you into that segmentation layer so you now know who your consumers are. How do you bucket them into audiences and groups that you want to drive behaviors in. How do you then activate on those. This is a really important one. Activation versus native marketing execution. What you find a lot of is these platforms will talk about the ability to activate and the reality is, you can activate an audience by pulling a list and handing it off to any number of providers.
That's activation, that's the definition. How can you orchestrate that. There's a level of complexity in that, that gets very, very deep and very sophisticated if this is the type of client or the type of audience you're looking for. How do you manage multi-wave touchpoints that if they saw this banner ad and then they click through the website, do I then send them a follow up email and if I sent them a follow up email and it wasn't read, what do I do there? That's that level of activation and then the native marketing execution is, do I have the ability to actually push that message in my channel?
This is where you've seen some of these big cloud providers buying up email platforms so that they can drive that and drive the push messaging and those types of things. All of them getting into that space. The next one there, and this is a really important one and this is an interesting final frontier for a lot of people. The marketer managed. We work at Epsilon and Conversant in the fortune 500 enterprise client set. The idea of marketer managed works really, really well in a single brand organization or maybe an 800 pound gorilla brand within an enterprise type organization.
Marketer managed means that the marketer is in control of ingesting the data, bringing it together and then driving through all of the activation and orchestration. The reality is, stitching that data together has a level of complexity, it's a very unsexy part of this but it's about managing that metadata and the data about the data. How that data is stitch together. How do you bucket large sets of data. Do you bucket it in groups of five or do you give the actual identity or the actual number, integer number for each of those? That's a very unsexy part of it but if a brand does it one way and another brand does it another, then trying to do that across an enterprise is very, very complicated.
We see that marketer managed is being something that everybody is looking for and it's definitely something we push for in the segmentation and activation but the idea of ingesting and bringing that data together in enterprise level is something that is best left to IT organizations or systems integrators or others to drive the least sexy part of this which is data governance and how do you manage that across that so that data is consistent across brands and relative to each other at the enterprise level. Marketer managed is a very big buzz word that you're seeing driven through the CDP platforms today but as you see, it's not really done in many other places.
Then, the final one is real time decisioning. This is obviously critical. This involves both the ability to ingest data in real time, get it back out decisioned with some sort of real time decisioning, machine learning that builds and learns on its own modeling and drives those results out. Again, this is where we think the industry will leapfrog a lot of the, sort of the orchestration levels but that's going to take time. There is a lot of technology debt, there's a lot of investment and time and organizations that manage those workflows. Letting the machine do that is something that's technically capable today but is something that will take challenges, hurdles to overcome organizationally over the next several years.
All of that, those are the top decisions around the platforms if you will and then the final question is who is going to do the work? This is one ... we talked about earlier, that on the walled gardens in particular and true on the customer data management as well, there is a licensing cost and then there is a people cost. That people cost is not just in terms of the bodies it takes to do the work, to push the buttons, to drive the integration but it's also the expertise and the know how to say, "Hey, look, if I do this, is it going to drive efficiency downstream?"
"Is it going to be manageable in the next six months? How do I take care of that?" There's really several options here. A lot of agencies do this type of work. A lot of the big consultancy firms do this work. A lot of what you're seeing is, very much like in the late 90s where all of the accounting software was sort of centralized into one platform. The big consultancies has built practices around that. You're seeing that now with platforms that those consultancies are building big expertise, basis behind that, to drive off of these platforms. The good news on this front is there are people out there to do this work.
It's definitely worth considering as you budget for this, that doing this, without side help is critical because one of the biggest challenges that I have found organizations have is managing the career path for people that integrate or manage these systems over time. A lot of companies have made a big investment in a big cloud provider technology, they've hired somebody they thought that could take care of the job, that person comes in builds their resume, halfway through their project and leapfrogs their career and they're left holding half built, half integrated technology and try to chase that down.
The decision, not just around the technology platform but the availability of resources is a critical decision as you look at decisions around what you're going to decide. We're going to wrap with a case study around a retailer, that was a multi-brand retailer that had a big challenge and I would love to tell you that they picked one platform, it was ours Mesobase and it solved all their problems and that would be a fantastic case story for us, case study for us. The reality is, it's deeper than that. It takes ... this was a client of ours that already had big chess pieces on the board with Oracle and Adobe.
It was up to us to help them make sure they were leveraging those, not just for the length of the contract they had left but for the technology debt, they had already paid into it that could go even longer. We work with them not just to stand up a platform we call Mesobase but we work with them to build out a strategy and a technology roadmap for them that took and help them focus and articulate very clearly the problems they were trying to solve with their platform decisions. We help them build the roadmap and prioritize the problems they were going to solve at which point so that everybody had transparency and visibility into what to expect.
Then, we help them even with some of the organizational decisions around what does their organization need to look like today and tomorrow to support the evolution of this platform decision as they go forward and then putting in the governance and the processing around that to help them get the most out of not just the platform we delivered for them but the systems they already had in place. As you get to making decisions around which platforms to put in place, which ones to implement into your ecosystem, it's important to know what problems you're trying to solve.
It's important to internalize and that means the resources, the skillsets, those types of things, only where critical. Only where it moves your business. Focus on what your business is, where those people have career paths within your organization. Drive to measurement, excuse me. Drive to measurement. As Dave talked about earlier, measurement is the only way that you're able to see the results and articulate them very clearly. That doesn't just happen with math, with good data. It takes a drive and a push and a prioritization of that within your organization. Everything should be measured and everything should be measured from front to back.
Finally, budget. Budget not just for the licensing cost, not just for the infrastructure, not just for the operational cost, but budget for the know how and the expertise and the skillsets to stand up and help you drive the most value for the investments that you've made and those that you're about to make. With that, I will wrap for questions. There is one question on the screen. We got one on the screen and then ... where would you rank attribution in terms of ... that's from earlier. Okay. You got another for me? Do we want to come back to ... is this one that Dave answered before or ... okay.
It's funny you say the must haves. Attribution and identity are book ends in our opinion. Your attribution with bad identity leads to fragmented or bad attribution, whether Dave is one person or two, if you see him as two, then you're going to get double counts on what works or half counts where it did and didn't work. For us, identity and attribution go hand in hand. When you get the attribution right and you have the transparency into all the data, the attribution is just math. It just adds up and gives you what you need. For us, it goes hand in hand.
All right, any other questions online, in house? Fantastic. Well, thank you all very much. We appreciate your time and enjoy the rest of the show. I do. They're on the inside. Thank you. Yes.
Audience Member: Nice presentation.
Carl Madaffari: Thanks. Thank you.