Why a Strong Data Foundation is Business-Critical

Kailey Raymond: People might think data and emotion are opposites. The unsexy stuff like definitions and processes, these are the nitty gritty details that allow marketers to make campaign decisions and build audiences. But in the publishing world, data also serves as the foundation for emotion. With the advent of AI technologies like natural language processing, we're able to capture topics, themes, and yes even emotions within an article.

This new data set allows advertisers to match their campaigns with an article's tone in setting, emotional responses, and in turn delivering higher conversions for the advertiser. Knowing how to implement these tactics has made Kendell Timmers an industry leader and her role as SVP and head of Data and Insights at the New York Times. In this episode, Kendell and I discuss NLP, personalizing paywalls and creating a unified view of the customer in a privacy forward way. Kendell, welcome to the show. Thanks so much for being here.

Kendell Timmers: I'm excited to be here. Thanks for having me.

Kailey Raymond: Awesome. Well, I just wanted to start off to get to know you a little bit better and your career journey a little bit more deeply. As I understand it before the New York Times, you're at American Express, ZS Associates and US Airways. But before we get there, I want to go back a little bit further. Kendell, what was your first job?

Kendell Timmers: Well, I grew up in Hershey, Pennsylvania, so naturally it's chocolate related. I worked at a place called Chocolate World and basically what was an all chocolate grocery store. So fun fact, I actually put out my back lifting chocolate improperly, which was a really fun visit to the ER to explain exactly what I had done to myself.

Kailey Raymond:That is a very unique occupational hazard. I hadn't considered that for a chocolate store employee. I also have a bad back, so I really empathize with starting off your job like that. So I'm so sorry. I feel like that might go one of two ways. Do you love chocolate or do you hate chocolate?

Kendell Timmers: Oh no, I very much love chocolate and I'm very biased towards Hershey. In fact, after working there, it was probably several decades before I ate M&M'S again, just out of company loyalty.

Kailey Raymond: A brand loyalist, marketing works. I love it. That's amazing. I'm a dark chocolate lover myself. I don't have a brand bias necessarily. I also have a weird thing where I sneeze when I eat anything that's 78% or over. I don't know what is with that.

Kendell Timmers: That's very specific.

Kailey Raymond: Yeah, if there're any doctors in the house that want to tell me why I sneeze when I have a certain percentage of cocoa, keep me posted. So after your Hershey ER days, you probably took a few different steps in a little bit of a different direction. What happened from there in your career journey? How did you get to where you are today at the New York Times?

Kendell Timmers: So I always tell everybody I did my education backwards. So I do think for a lot of analytically based careers, you need some kind of background in numbers and you need some kind of understanding of how businesses work. So a lot of people, they'll get a technical undergrad and then polish it off with an MBA on the top. I did it the other way around, because I did not realize that I liked math until I hit calculus and 18 year old Kendell thought, "Oh, it's too late to get into math now at 18." So I majored in business and then I only realized at my first job out of school like, "Oh wait, no, I want something much more technical than this." I really wanted to be deeper in the math of it. So I ended up going back to school to get a master's degree in operations research. And the reason I love OR so much, is it basically provides you with this whole tool set for how to solve a whole bunch of analytic and business problems, almost agnostic of what the problem is, it's just the tool set.

Kailey Raymond: That's amazing. I love Calc too. It's also so funny to be at 18, you're so young that you're just kind of dismissing an entire career path because you didn't like it for elementary school basically. But I love that you challenged yourself and went back. You position that in a really cool way too, the fact that you're aiming to provide a tool set for problem solving with data without regard agnostic to what the problem actually is. I love that kind of definition of what OR does. So tell me a little bit about how you've done that at American Express and The Times.

Kendell Timmers: Well, first at Amex, I was there for about 12 years and I did a whole bunch of things from loyalty campaigns to prospecting. My last role there, I owned all the prospect data, whether it was in the DMP or whether it was mail, mailing addresses and emails, and really we were focusing all our time on what is the right data and the right way to reach out to people to make them aware of our product and to get them to convert. And then when I went to the New York Times, I was actually switching my perspective. I kind of had to convince them that because I knew what I wanted when I was more of a marketer, I could tell them as an advertiser what the marketers who they were talking to would want from a data perspective. And so, I started off on advertising side at The Times, but what I run now, is the whole data group, which kind of runs the gamut from helping to maximize our subscriptions, working on making sure we get the right reach for our journalism, increasing engagement and marketing and advertising.

Kailey Raymond: That's amazing. So it sounds like a lot of what you do reaches out across tons of different departments and business partners, and you've become really critical to the way that a lot of these folks actually make decisions. You mentioned loyalty, advertising, user engagement, those are all kind of music to my ears as a marketer, and I can definitely tell you that I wouldn't be able to do my job without people like you. So thank you. And I'm not alone actually, we actually did a recent survey.

Our growth report came out recently and we found out that 85% of marketers believe that their roles are becoming more data driven, and in the past year they feel like they become more data driven. And so, data has really just become this competitive asset that powers decision making across every single team in the organization. The ability for your team to kind of be that centralized resource of excellence is really incredible.

So something I'm curious about, you've been in analytical and tech roles for a long time in different industries and especially industries that I feel like have had a major shift towards digital. What are some of those major industry trends that you're currently seeing in following?

Kendell Timmers: I think the biggest one is that scramble to truly understand your customer, really get that unified customer view and to do it in a privacy forward way. And those two things, I think often in the industry, they're seen as being an opposition. But one of the things I'm really happy about working at The Times is we've really been able to do things that accomplish both, that allow us to understand our customers, but to do so in ways that respect not just the letter of the law, but also the spirit of the law with privacy.

That's a big part of it. An example of that would be first party data. So we are large enough to have a good base of subscribers and registered users for whom we can understand how they use our product, and to be able to take that information and turn it into first party audiences. And we tried this experiment, I guess I about two years ago, and our first party audiences were so successful that we were actually able to get rid of third party. So on our sites, we only use first party data for audiences and just completely got rid of third.

Kailey Raymond: That's amazing. And I mean, I'm sure you might imagine, we've talked about privacy quite a few times on this podcast, but I think what you just said, a lot of people think it's an opposition, and I've said this before, that there's this paradox that people think exists between privacy and personalization. You just prove that it doesn't exist if you are collecting the right data in the right ways and you can build audiences and maintain the conversions, and you're just calling out something that's really interesting, which is a win for all of the different groups that you service for your readers, for your advertisers, for yourself. So I think that's really interesting that you can create those relevant personalized performative ads and probably reduce costs from third party vendors too.

Kendell Timmers: Yeah, we no longer buy third party data at all.

Kailey Raymond: That's amazing. So I mean that ability to reduce costs I think is, especially in this macro environment right now, really, really interesting. Again, from that growth report that I was talking about, we were talking to, I think it was like 1500, maybe 2000 marketers and 93% of them say they're taking steps to adapt to this current macro environment, but 63% of them are looking to decrease spending on marketing tech. So the ability to take out some third party resource and only use your first party data, certainly interesting. And it sounds like you guys are a bit ahead of the curve as it comes to some of the privacy conversations, which often comes up as a challenge that people are going through. But what are some of those biggest challenges that you are facing at The Times?

Kendell Timmers: Well, so I'll speak more from a data standpoint, but I think many organizations, we're trying to rethink how our data is organized. I'm kind of anti buzzword, and so I hear the words data mesh and I cringe, but I think that's exactly what we actually need. So when I joined The Times, I was really pleasantly surprised. Hey, all our data is together in one place, that's leaps and bounds forward from what I'd experienced before. And it's really helpful, but there's always a desire for more data and to get the data in the right places, we actually need to go past just getting all the data together and we need to get it together in a way in which people can actually use it. So it's about keeping data from getting siloed into different domains where everybody's kind of creating their own definition and having it all in one place where it's discoverable and usable, so that the dashboards that everybody needs to use to make decisions all reflect consistent definitions for instance.

There was an example we had, we had two dashboards that were talking about our audience numbers, and one of them included cooking and games and the other one just included core news, but they both used the same word to talk about audience, the same definition. And then when that was pointed out to both of them, they both changed it. So now the one that had cooking and games took it off and the one that didn't put it on, so they still disagreed. So having a place where you can say, especially if you're a new person who's trying to build your first dashboard, "Oh, what's the definition of audience?" There are like 50 definitions in here, which one should I use and why, and where and how do we name them differently so that people understand what this audience is versus that one and make sure that we're telling consistent stories so that people can make consistent business decisions.

Kailey Raymond: I love that you're calling that out. I've encountered that so many times in my career and I feel like the first couple of months, mainly what you're trying to do is figure out how things are defined, so that you can really understand your scope and your ability to influence things in an organization. Because you're right, definitions and naming are so critical. And to your point on data silos, we hear that all the time. An incomplete view of that customer, everything becomes more expensive, more challenging, it's a mess.

And bringing that all together really helps make sure that that customer experience is improved, that you're actually speaking to the customer in the right way at the right time. Something that I think is a little bit squishier maybe as well, but really important when it really relates to data management, is that it can make or break trust in an organization. If everybody thinks that something is named a different thing, there can be some interesting conversations that you have to reconcile. And so, having that upfront is super critical and important. So knowing that those definitions are really important, how would you define good data?

Kendell Timmers: I think I'd say good data is accurate, actionable, and findable. So obviously the most important part is that it's right. And so, if you're building a data organization, you have to build in QC processes and alerts, and I think everybody's got some version of that, but everybody could probably stand to make that a little better. And on alerts in particular, there's always this tightrope you have to walk between, you don't have enough alerts to prevent the things that are really important, or you have so many alerts that everybody just ignores them because they are getting too many…

Kailey Raymond:Alert overload

Kendell Timmers: …a day, that's not helpful either. So figuring out that balance. On the actionable side, can you actually use it to make a business decision? And some of that has to do with timing. So knowing that a reader completed a particular article is really useful to know quickly, but could become less and less helpful over time. A year from now, I don't know how much particular information that gives me. And for findable, it goes without saying, if you can't find it, you can't use it. And that again goes back to the, if I'm new here and I want to know, how do I figure out what a page view is? And you either can't find it or you find 30 things called page view, that's not findable.

Kailey Raymond: Yeah, definitely not. For me, it's searching in Looker and finding a million different things that say the exact same thing but mean totally, completely different things. And I love the fact that you mentioned time. I mean, time decay, especially in my world, is so, so important. The ability to act on that real time information is super important to be able to personalize those interactions and really drive conversions. The most obvious thing that comes to mind is like cart abandoners.

If you get that email a week after you put shoes in your cart or whatever, you're probably not going to convert as readily as you would if it's today or within a couple hours. So that's really important. And one of the ones that we hear about now all the time too, is real time audience suppression. So being able to understand exactly who that person is that purchased, you don't want to serve them that ad. And that's going to save you tons of money as well, which is kind of a really cool way to make sure that you're doing the most for the business and being an owner and acting like an owner. What are some of the ways that you're using this good data across the New York Times? Any tactics or programs that you're willing to share?

Kendell Timmers: So one of the ones we've spoken about a little bit recently that I've been excited about for years is the dynamic meter. So this is maybe not the first thing that pops to people's mind when they think about personalization, but it's really important. When we started having a pay wall back in, I think it was 2011, it was a fixed pay wall. So at the beginning of the month you had X number of articles. When those articles were out, you couldn't read anymore unless you subscribed.

And over time we've realized that we could be actually a little more granular with this. So we want to be able to base how many articles you have on the customer's usage. So I can use how many articles that you read in the past to be able to predict how many articles I need to give you to get you to convert. So I want to strike that balance between letting the reader kind of sample our product, understand it's worth, and actually at some point having to make that sale and make that conversion. And with this, we really have an ability to fine tune that at a more personal level.

Kailey Raymond: That's so interesting. So you're telling me that when I'm on Twitter, scrolling around, clicking into New York Times articles, I'm a subscriber, by the way, I just haven't logged in necessarily on Twitter all the time, but that article limit that you're showing me is personal to me?

Kendell Timmers: That's correct.

Kailey Raymond: You're right, it's not the most obvious example of personalization, but I imagine it probably helps a whole lot with your conversion and your customer acquisition costs. I won't ask the specifics, but I'm sure both of those metrics moved in the exact right direction with that initiative. I'm also curious, I worked at an AI company, my last company, it was in the research and insight space, and they used NLP to help people basically find information more quickly and really dense documents like bankers used it to make decisions, that kind of thing. But I read about a product that your team was working on and helping develop using NLP that was helping your marketing team and your advertisers make more informed decisions. Can you share a little bit more about what that is?

Kendell Timmers: Yeah, I think you're talking about Kaleidoscope, which originally was called ReaderScope. And the idea was particularly when an advertiser is thinking about advertising with The Times, it can be kind of hard to figure out what should I be buying, both in terms of what audiences, what contexts and are these audiences the people that I think they are, how do they consume the time? So we've actually been able to create this product that the backbone of it is on one hand the audiences. So you're the gender and the age and the household income and those other kind of standard audiences that again, are first party at this point, against a topic model that was created with natural language processing that has more than a hundred different topics like education or football to be able to understand how audiences consume topics or what topics over index on what audiences and vice versa, and allows you to make a better, more informed decision on what that right audience or context is for you as an advertiser.

Kailey Raymond: That's cool. So it's more than NLP, it's attaching audiences to that theme. So it's way more powerful than simply NLP alone, in my opinion. You're attaching a couple of different data sets together. Very cool. So do you have an example maybe of when data might have helped you surface something that surprised you?

Kendell Timmers: Yeah, so it's interesting, NLP always surprises me with the ability to pick out things that a human wouldn't necessarily see. So for example, in an early version of that topic model, it was taking all the articles and creating groupings of topics. And as it went along, when it came to the sports articles, it was consistently grouping them into two categories. Now, as a human I would've said like, "Hey, there's soccer, there's football, there's hockey, there's baseball." 

Kailey Raymond: That's logical.

Kendell Timmers: They were grouping them, they the algorithm. 

Kailey Raymond: I love that, “they”, the computers.

Kendell Timmers: Yes, the computers were grouping it by articles that were more of a transactional nature about reporting on a sporting event versus articles that were about the sport as a whole or an athlete's career or something that was going on in the world of sports. And when you think about it, you do use very different words and themes, but as a human, that would not have been the first way that I thought to classify those articles. So I found that a really interesting insight.

Kailey Raymond: Definitely, that kind of rub between humans and machines right there. I was talking to the head of ad operations at this big anime brand a couple weeks ago, and he was saying a very similar thing where I was asking him about topic grouping and how they actually do it. And he said the best way to do it is that they use humans. There's still some things that machines, at least we don't intuit in the same way that they might, of course a baseball player and a baseball score are likely more relevant to each other than a baseball score and a hockey score that might not be the same kind of fan base, but machines, they see things that maybe we don't. That's really cool. What would you say is your favorite piece of data that you use in your role to drive customer engagement?

Kendell Timmers: It's really, really hard to pick favorites, but I'll talk about one that happened just before I joined The Times, that was maybe one of the reasons that I joined The Times because I found this so interesting and inspiring. It was called at First Project Feels and the name kind of stuck, but basically we use machine learning to classify all of our articles by the likely emotional response. And you hear about sentiment analysis everywhere. This isn't just like, is this a happy article or is it a sad article? It really gets to very niche emotions like this article makes people feel self confident or adventurous or nostalgic, and that really allows an advertiser to think about exactly what mood their ad creates and how can they be surrounded by content that matches that mood. And it's a different way of thinking about how you want to advertise, but in my mind, a really performative one from what we've seen, and we now have more than 30 of these different niche emotional responses that we can link to.

Kailey Raymond: That is so cool. Nostalgic adventures, those are such specific emotions and nostalgia I'm sure is one they latch onto, because aren't there tons of studies that showcase that it's one of the most powerful emotions to bring people into a different kind of space. I assume that helps a lot with contextual advertising, but how has Project Feels impacted the way that you're interacting with your advertisers?

Kendell Timmers: I mean, what's really cool is it makes it more of a conversation. So with all of these tools, whether it's Kaleidoscope or Feels or just even the first party audiences themselves, we sit down and have a conversation about what's relevant to you as an advertiser and what do you need? And again, because all of these are very privacy forward solutions, we have a really good positive conversation to have about that, which is always really nice.

Kailey Raymond: That's so important. I mean, it's something that has been on the top of the minds of advertisers for years now. There's so much legislation from governments with GDPR and CCPA to the platforms who frankly are I think driving most of the conversation at this point with Google and Apple, such fear, such fear from advertisers that they won't be able to do that personalization that you were kind of talking about at the very top. And it seems like you've tackled a lot of this at The Times, and you have this kind of blueprint for how you're actually creating this data maturity and how you're staying compliant and how you're able to perform over time. So it's awesome to hear this and very inspiring, I'm hoping for a lot of the people listening in, but that's what you've done so far. What about what's on the horizon? What do you see on the horizon for customer data for publishing in the next 6 to 12 months?

Kendell Timmers: I think there's really two things. One, on the consumer data side and one more just a publishing thing. So on the consumer data side, the obvious answer is the looming cookie apocalypse that we've sort of alluded to a couple times. Although the dates are changing, the fact remains it's coming, and we need to find ways to comply. And again, not just with the letter of the law, but with the spirit of it, customers aren't dumb. And if you do things with your customer's data that they wouldn't expect, you lose their trust and their loyalty and that's really important. What I'm seeing is kind of three responses that are going. So one, is companies trying to find ways around the restrictions, new ways to connect data across the surfaces, and they become more guessy as they go, as accurate as they try and go through that.

The other two, companies making a bigger effort to get the readers to self-identify so that they can use first party data. And then the third would be contextual. So other alternatives, then audience targeting to make sure that you're hitting the right moment for the ad. Obviously, of those three things, I'm really a fan of those second and third ones, and I think they're both really important. Contextual, is one that once people hit audiences and the age of the DMP, they sort of forgot about contextual, but machine learning has progressed so much since then, that there's just so many ways to dive into contextual in really smart and very granular ways that I think is going to be every bit as important as audience. And there's nothing in my mind more privacy compliant than something that's fully contextual, because anybody reading that article is going to see that ad, it's nothing about that person personally at that point.

Kailey Raymond: You're right, I think it's kind of growing once again. And I was reading about one of IBM's use cases there, and this makes perfect sense. They have weather triggered advertising, so of course, it's such a specific behavior, it's time bound. If a North Easter is blowing through, you're going to want salt, you're going to want shovels, and those advertisers are going to want to hit you exactly in that timeframe or they lost you. So such an easy example to showcase the power of what that might actually look like. You mentioned a second trend related maybe to publishing.

Kendell Timmers: Yeah, and it's a little bit related I think, that there's this more and more crowded subscription services space. Everybody wants first party data, and so the best way to get that is to get people to subscribe. So it used to be you'd have cable, maybe a magazine or two, but now every service is a subscription service. And it gets really confusing as a consumer, "Did I sign up for this via Apple? Where do I go to cancel this?" It gets really crowded and I think we're going to hit a limit and people are already hitting a limit on how many subscriptions they're really comfortable signing up for. And so, publishers who maybe before were just competing against other publishers, you're now competing against the whole big space for somebody's sort of subscription budget.

Kailey Raymond: It's unbelievable actually. You're talking to a cord cutter, so I've never had cable, but I have reached my limits. I feel like over the past three to five years, it's just been this absolute proliferation of services and I'm like, which ones do I actually subscribe to through what? To your point of how can I actually think about canceling this or decreasing what my subscription looks like? It's so hard to keep track, but I'm also, the content's really good. So what do I do? Rock and a hard place.

Who do you think is doing it right as it relates to data in the ways that they're activating it?

Kendell Timmers: So I do think the whole Disney Plus, Hulu, ESPN combo was interesting as a parent of young children. The Disney Plus thing was almost inevitable, but it's an interesting compelling example of putting together a range of different things. Of course, I'm excited about what The Times is doing along those exact same lines to bring together things like games and cooking and wire cutter and the athletic together to provide that whole well-rounded kind of bundle that's like journalism, recipes, smart, fun, kind of along those same lines as Disney.

Kailey Raymond: That's awesome. And I imagine you're trying as much as you can to unify all of those profiles across all of those different brands to make sure you're getting the most out of those customers?

Kendell Timmers: That's right.

Kailey Raymond: That's awesome. Well, Kendell, my last question for you of the day. If you had any piece of advice that you could give somebody that's trying to build their data foundation, what would it be?

Kendell Timmers: Well, I mentioned it a little bit before, but I think the thing that people most overlook is the organization and structure and rules behind their data. And you don't want it to be so restrictive that you can't explore, create new definitions, but you also don't want it to be so anything goes that you end up with 30 definitions of page view. So if you're really starting out, try and build that in as you go. But if you're already in a working system and you're trying to make it better, I mean, join the club. There are lots of other data people in other companies we should all be talking through it, you should be reading the articles and figure out with an open mind, what is the best way to organize data going forward? And we all need to really spend some time on that. Not easy, not fun kind of process.

Kailey Raymond: Yeah, it's sometimes the unsexy stuff that is really the thing that drives the business forward, it's so important. Process, structure, ownership, those clear rules that are building trust within an organization. And if you do have that right structure, you actually can adapt and you can be flexible as well. Kendell, thank you so much. This has been so awesome, I've learned so much from you today.

Kendell Timmers: Well, thank you so much for having me.