A History of Marketing / Episode 39
A recurring theme on A History of Marketing is the tension between marketing as an art and marketing as a science.
Lately, we’ve explored the former. David Gluckman shared how he invented Baileys Irish Cream in 1973 based on gut instinct and “the benefit of ignorance.” Scott Reames revealed how the team that birthed the Nike brand in 1971 had no formal training as marketers.
This week, the pendulum swings in the other direction in my excellent conversation with Jim Spaeth, Ph.D.
Jim’s career places him at the center of the industry’s shift toward rigorous measurement. From his early days at Young & Rubicam and General Foods, Jim pioneered Marketing Mix Modeling (MMM), a discipline designed to measure marketing’s ROI in financial terms and further optimize investments in marketing.
The Origins of MMM: How General Foods used early models to uncover granular insights to bridge the gap between marketing and finance
Connecting Ads to Sales: How the ambitious ScanAmerica venture attempted to measure actual SKU-level supermarket purchases to locally-aired TV ads
Standardizing Internet Advertising: His time leading the Advertising Research Foundation (ARF) during the dot-com boom, where he fought to standardize the chaotic new language of clicks and views
The Future of Measurement: How deep learning and AI are addressing the lingering challenges of causality and creative assessment
If the last few episodes demonstrated the power of creative intuition, this conversation explores the discipline of proving that intuition actually works.
Listen to the podcast: Spotify / Apple Podcasts
Thank you to Xiaoying Feng, a Marketing Ph.D. Candidate at Syracuse, who volunteers to review and edit transcripts for accuracy and clarity.
Defining Marketing Mix Modeling
Andrew Mitrak: Jim Spaeth, welcome to A History of Marketing.
Jim Spaeth: Thank you.
Andrew Mitrak: I’m really excited to speak with you. Prior guests, Shelly Zalis and Bill Moult, both highly recommended speaking with you, especially on the history of marketing mix modeling, which is something that has been brought up on this show before, but we haven’t fully delved into.
So, I thought I’d ask you about some of that. And first, I thought maybe we could define that for listeners. So if you were at a dinner party or having a glass of wine with somebody and you had to explain marketing mix modeling to them—to somebody who hadn’t heard of it before—how would you overall describe that to them?
Jim Spaeth: I would describe it in a few ways. The simplest is just from what its goal is. It attempts to decompose the contribution to sales of all of the marketing and non-marketing factors that drive sales. And at its simplest, in some ways, it’s to prove that marketing is accomplishing something. To quantify what it’s accomplishing and to measure the return on investment. So, it’s a way for marketing and finance people to talk to each other. That’s kind of one part of it.
But also, by understanding all of the factors that are driving sales, you can begin to think about how to optimally allocate budgets, which things are working, which things need help, how external factors that you don’t control impact your sales. So good examples would be things like the weather, the economy, competitive product pricing. You don’t control everything, but you need to react to certain things. So, in a way, conceptually, it’s the engine for an ideal dashboard for a marketer.
Andrew Mitrak: And do you think of it more as a backward-looking dashboard analyzing results or more of a forward-looking model of where to apply things? Or is it a little bit of both?
Jim Spaeth: It’s a little bit of both. It’s accused of being backward-looking because the problem is, typically marketing mix models need two or three years’ worth of data. So, what they learn or figure—learn tends to make it sound like AI, and these days there is AI in marketing mix, but not always, and certainly there hasn’t been—but what you can infer, I guess is a better way of saying it, from marketing mix is what the impact of those historical activities have been.
So if you ran the same television campaign for the last 10 years, that’s what you’re going to understand about, that’s what your model is going to tell you about. It’s not going to tell you about this great new campaign and what it might do. Now, that said, you can use marketing mix to, as I said, optimize your spend going forward based on what historical responses have been. You can use it to forecast, and forecasting can be at a broad level or it could be at a very granular level to help, you know, production, inventory, things of that sort.
But you always have to recognize that it’s based on history. So that’s why it has that backward-looking reputation. However, I would say any technique is backward-looking because any technique based on data is based on something from the past. I mean, it might be yesterday or last week or last month, but it’s something from the past. It’s just a question of how far back in the past you have to go.
The Origins of MMM: From Marketing as Art to Science
Andrew Mitrak: You mentioned earlier in your career, you were at General Foods in-house. And by the way, General Foods is probably better known today for the Post cereal brands, is that right?
Jim Spaeth: Oh, it was the home of Post, Maxwell House, Jell-O, Tang that went to the moon, or at least in space, I can’t remember anymore. So, it was merged with Kraft, then Kraft has been merged with Heinz...
Andrew Mitrak: And so really mass CPG products. Were you involved in marketing mix modeling there?
Jim Spaeth: So that was kind of the beginning. GF was the beginning of marketing mix modeling, quite frankly. I would say the prologue is, before that, I worked for nine years at Young & Rubicam when it was a full-service ad agency. And those were the days of the 15% commission, so the agency had a lot of money to bestow many, many benefits and services upon their clients, and they weren’t always squabbling over every nickel and dime.
So we had a big, big research department and we did a lot of different stuff. But these guys were thinking about how to use science in marketing. Marketing was an art. I mean, it was just all an art. It was judgment, it was creativity, it was... and it worked. You know, we just couldn’t really prove it, we couldn’t measure it, but you could see it work when well-marketed brands beat the crap out of the more commodity brands, and they could charge more money and get away with it because they seemed to be higher value. So we knew marketing worked, but it hadn’t really been measured and it hadn’t really been submitted to the science of optimization and all those kinds of things. So that work was beginning at Y&R.
So, those are kind of the prologue days, really trying to bring science methods, econometrics, and such into the marketing world.
An Early Insight: Connecting Stove Top Stuffing Sales with Potato Prices
Andrew Mitrak: You worked at Y&R before General Foods. Did having this in-house and agency perspective on marketing mix modeling shape your views in any way? Did you notice sort of gaps or things that didn’t work? Because not everybody has the luxury of working at both.
Jim Spaeth: Yeah, that’s a really good, really, really good point. Absolutely did. My knowledge from Y&R of how media is bought and sold made a big difference, because why do you need to learn something when you can’t act on it, right? So, you buy your television largely in the upfront market. While you have some flexibility, you’re pretty much, let’s say, 80% locked in. So, we’re not going to be able to change it anytime soon, whereas radio was pretty nimble, and if you learned something about radio, you could act on it over the weekend and make a change.
So things like that, knowing how media really worked was very, very helpful. And then the view from inside the business, inside the food business, was amazing because you really got to see how it worked—both organizationally, how it worked, what the decisions were, when they got made, what the decision-making process was like.
I came in initially into the market research department where we began to build an actual, honest-to-God marketing mix modeling practice, which meant, to boil it down simplistically, using econometrics. So using regression modeling and all the sales and marketing data you could lay your hands on. And in those early days, we built models for all of the GF brands.
And they were very simplistic. They would just say, “Here’s my sales. To what degree is...” Now remember, we’re talking in the ‘80s now, right? “To what degree does television drive sales versus radio versus outdoor versus magazines? What’s the impact of pricing? What about my in-store promotions? What about my coupons?” So what I just named, seven or eight factors, maybe there would be 10 or 12 factors in a model. My personal big breakthrough was when I discovered that Stove Top stuffing did well when the price of potatoes was high. So, you know, starch on your plate could be Stove Top, could be potatoes. When potatoes were expensive, you go to Stove Top. When potatoes are cheap, Stove Top didn’t do as well. So, that was pretty sophisticated. That was a cross-elasticity in the model. So they were really, really simplistic.
Who Actually Used Marketing Research?
Andrew Mitrak: When you built those early models at General Foods, who do you present that to, and who makes use of that information? Is it to product groups as far as, “Okay, we’re going to change our products”? Is it to advertising agencies who then use it to inform their campaigns? How does that research actually get put into practice?
Jim Spaeth: That’s a very, very good question because that development of that kind of organizational process has been equally, if not more, important than the development of data and statistical technique. In those days, we worked collaboratively. So every division had its own little research group and a head of research. And so we would collaborate with them because we wanted to be somewhat cohesive and consistent in our story. And then ultimately, we would report to the marketing director or the division president.
Andrew Mitrak: The marketing director or division president, they are looking at all this information and then they’re using it to just inform their overall marketing strategies.
Jim Spaeth: Right. Setting budgets, allocating budgets. That was the primary application, deciding what’s working, what’s not working. Back in those days, we weren’t able to break out creative, so we couldn’t really try to understand whether this particular campaign was working. That has only come very recently. That was a desire from the beginning, but not something we could accomplish.
Andrew Mitrak: Did it feel like General Foods was early to this kind of practice? That this was sort of the frontier of using research and analytics?
Jim Spaeth: My impression is back then, GF was clearly the first. And it disseminated from there. So one of the reasons mix modeling in its early phase was pretty much just a CPG practice is because people moved from GF to Pepsi, to Kraft, to Clorox, to wherever, and brought the practice with them.
The ScanAmerica Project: “Ahead of its Time”
Andrew Mitrak: After General Foods, at some point, you get involved with ScanAmerica and Arbitron. Can you tell me this story?
Jim Spaeth: That was an early attempt at measuring both product purchasing and television viewing among the same households so that you could look at not the audiences among women, but the audience among Maxwell House purchasers, or frequent Maxwell House purchasers. That was ScanAmerica. That was actually a joint venture between Burke Marketing Services and Arbitron. So Arbitron did the TV part and Burke did the sales part, and they also had expertise in test marketing.
So that was very, very exciting, and it didn’t make it. The first of a number of startups that I’ve been involved in that have not been successful but have been exciting and groundbreaking.
Andrew Mitrak: It seems like a really ambitious project because this would have been in the ‘80s?
Jim Spaeth: Second half of the ‘80s.
Andrew Mitrak: The idea really was the ads you’re running, really connecting those to actual purchases and having a set of SKUs that are at a grocery market in an area that saw those ads, and really connecting end-to-end, which seems like a complex, ambitious challenge to solve. It still seems challenging today, especially 40, 50 years ago or so, it sounds like a really tough one.
Jim Spaeth: Yeah, it was a little bit ahead of its time.
Andrew Mitrak: I’ve worked for a handful of startups and it seems like there’s an interesting thing as a marketer to join a product as part of the vision, right? Where the product is not quite reality yet, but if you get enough people on board and it seems like it could cross the chasm and become real, there’s a big opportunity. So it seems like it’s important to just swing for the fences and try things out, even if they don’t work.
Jim Spaeth: Absolutely. That’s been the story of my career. It’s like, when you see it, go for it. And we really, really believed in it. And we had clients who really, really believed in it. It was just too hard to do, too hard, too expensive to do at that time. And the other thing is you run into those organizational issues. That’s I think where I first learned, it may make complete sense, total sense, it might have demonstrable economic benefit, but before you really push too hard, make sure you understand what the industry or organizational constraints might be.
Standardizing a New Frontier: The Internet and the ARF
Andrew Mitrak: You mentioned the Advertising Research Foundation (ARF), and you became president of the ARF in the mid-’90s. What is the ARF for somebody who hasn’t heard of that before?
Jim Spaeth: Sure. The Advertising Research Foundation is, I want to say the oldest... Now, I don’t know how far back the AMA goes, but the oldest or one of the oldest marketing-centric trade organizations in the United States. It was founded by the ANA (the Association of National Advertisers) and the 4A’s (the American Association of Advertising Agencies) because they needed something, they needed to stimulate better research.
Andrew Mitrak: Yeah, and I think the AMA was the ‘30s or ‘40s. I’d have to look it up.
Jim Spaeth: So I think the ARF actually predates it. And it was focused on advertising because it was the child of the advertisers and the agencies. So obviously marketing is a broader topic, but that’s what their focal point was. And again, many, many years later, ARF was still respected around the world as a preeminent authority with respect to advertising and media in particular, maybe not marketing more broadly or market research more broadly, because you had other organizations in Europe and in Asia and elsewhere. But it’s an organization with global stature that’s been around for a long time and does a lot of leadership. We’re doing some great work right now on this topic of marketing mix modeling.
Andrew Mitrak: That’s awesome. By the way, it looks like ARF, based on Wikipedia at least, was founded in 1936, and AMA was founded in 1937. So it predates the AMA. [laughs]
Jim Spaeth: Slightly earlier, all right.
Andrew Mitrak: Why the jump to ARF after having a career at startups, at agencies, at large CPG companies?
Jim Spaeth: The ARF was very important to me personally. The reason is, back when I was at General Foods, the head of research at GF was involved, I think he might have been on the board of ARF. They needed someone with some media expertise in a volunteer role. This was just early on. And he kind of volunteered—he went to my boss and he said, “I want to volunteer Jim for this job.” Which was great because I was a little bit bored, frankly.
And I had very little exposure to the outside world. I’d been in my two companies I worked at and was fine and comfortable, but didn’t really get out a lot. So this kind of put me out in the outside world. And I will never forget being in a meeting with five or six legends of the day who were on this committee. They just didn’t want to chair it. And somehow, not because of me, but because of my company, I had the stature to chair it. So they gave it to me. And the kid walks in, not knowing what he was getting himself into, and suddenly I’m talking to these people who I’ve read their papers, I’ve read their articles, I know all about them. It was like... I was awestruck.
And I will say, just without making this too much about me, it enabled me to find my footing. And I think that’s what ARF does for a lot of people in the industry. It really brings them on board, broadens their perspective. And a lot of my ambitions where maybe I had something that was innovative I really wanted to push, but the company I worked at wasn’t really quite ready for it, I could go... I had another gig. I could do whatever ARF volunteer work I was doing and kind of try to push the industry in a certain direction. My whole career has been about innovation. What better place to try to drive innovation? So that’s what brought me there.
Andrew Mitrak: As far as driving innovation when you’re there, this is the late ‘90s through early 2000s that you’re president, and this is the dot-com bubble era. The rise of the internet and the rise of early digital advertising as well. As I was researching and prepping for this interview, I came across an article that was published in 2002, and I’m going to read you a quote. The article is all about how advertising measurement models are changing as the internet is developing. And I’m going to quote: “We’ve invented this jargon: clicks and ad views and page views,” says Jim Spaeth, president of the ARF. “We need to direct people to standard media terminology and get people to talk the same language.”
I’m still talking about ad views... yeah, ad views and page views and clicks. Those are still used today. So I was just thinking of this, putting myself in this era. It’s like, okay, the internet’s developing, it’s brand new for everybody. There’s a lot of uncertainty around it. It kind of sounds somewhat familiar with AI today as well. But there’s new standards that need to happen, people are measuring different things, there’s uncertainty. And then you, as the head of ARF, need to help standardize, need to kind of bring people together. Can you just paint a picture of that era and some of the challenges and opportunities there, and who you needed to persuade to adopt the same language around things?
Jim Spaeth: The internet was just happening. You had a lot of really smart people coming on the scene with absolutely no background in advertising or marketing, for that matter. And they were reinventing the world. Now, that’s great, they’re reinventing the world, but as you will appreciate as a historian, it pays to understand where this world came from and what the framework is you’re moving into. Particularly when you think about it from an ad spend perspective, digital was new. It was like novelty money, it was experimental money, it was some extra cash here and there. The bulk of the spend was still... we were spending more in outdoor advertising than in digital, right?
But suddenly digital is reinventing everything. So we don’t have impressions, we have clicks, and we don’t have reach, we have uniques, and we don’t... It’s like, can you just stop confusing people? Do you want to be part of the scene, or do you want to have somebody have to have a special training course just to understand your vocabulary? And then by the way, what happens when you put your clicks into a media plan? Do we add them to the impressions or do we have to create clicks for television? It was like... it was kind of stupid, frankly. And it continues.
Why Programmatic Advertising was Digital Marketing’s Breakthrough
Andrew Mitrak: It continues. It is shocking how little digital was for as far as a percentage of marketing spend for as long as it was, but it grew at a rapid rate. Were there any milestones at standardization that you saw as big wins in this period?
Jim Spaeth: I think the biggest breakthrough was programmatic advertising. Digital always had a problem because the audiences to any one thing you bought were so tiny that instead of going out and buying a 12-rated television program, now I got 12% of the country watching me right now. That’s it. Let’s put another one of these in. Let’s put 12 of these in. Now I’ve got a plan. You could do it on the back of a napkin. You could do it on an Excel spreadsheet for sure.
But digital was tiny, and you needed to buy thousands and thousands of units to add up to anything. So it was manually prohibitive. Labor was prohibitive. It was not efficient for agencies to do at all until you could do it in an automated fashion, which was supported by the fact that digital is data-driven, right? So there’s metadata and so forth associated with it.
So programmatic was possible for digital and absolutely necessary. And it took a while to get it to work right. And I’m not sure it still works right, but... and then we had some experiments trying to do television programmatically, and the early ones were kind of nuts because you didn’t... whatever, it’s a long subject, but they ignored some basic things that we understood from the beginning of time. Like, people pay a lot of attention to ads on primetime, so that’s important. They pay a lot less attention to ads in daytime because they’re busy, if they’re home, they’re busy around the house doing whatever they’re doing. So there’s different value propositions, and that was often not seen in programmatic. Now, attention has become an important variable, which I think is great, and that really differentiates the quality of impressions in a big way.
So, automation, right? Sorry, long-winded answer. Automation was necessary and possible for digital. It made it more affordable, and then the other thing that happens is digital has more or less an infinite supply, right? So, as an economist, I would say when your supply is infinite, your price gets really low. So that’s why digital impressions are so inexpensive. “Oh, you want some more? We’ll make you some more.” It’s a little oversimplified, I will say, but that dynamic is at work. So they became cheaper impressions. It was good for advertisers because they were cheap. It was good for agencies because it was labor... it was not labor-intensive, it was efficient and profitable for the agencies. And now, of course, we’ve got television and out-of-home and other media, radio, going through that same kind of automation process now.
Recent Innovations in Marketing Mix Modeling and the Impact of AI
Andrew Mitrak: Kind of reflecting on this conversation in your career and the development of marketing mix modeling, are there any sort of remaining challenges with it that haven’t fully been solved yet? Or are there still things you would have expected to be more buttoned up and dialed in and more of a solved problem by now that still haven’t been resolved? Are there lingering things where you think, “Wow, more innovation needs to happen in this space that will \pave the way for it in the future”?
Jim Spaeth: There’s always that. But you know, I have to say, marketing mix is, in some cases, conducted by bigger companies as a business, and like a lot of businesses, they’re really trying to drive costs down and profits up. Not to say they don’t do a good job, I’m not saying that at all, but it changes the orientation a little bit.
And then there are a lot of modeling companies that are driven by entrepreneurs who are really motivated to do a great job. It’s one of the things I’ve enjoyed about being involved in this field is a lot of very highly motivated—and not just about making money, I mean, that’s obviously a good thing—but highly motivated to just advance the state of the art and do a great job.
And just off the top of my head, I’ll call out Nancy Smith at Analytic Partners, who has created one of the biggest independent marketing mix companies in the world. I would call out Ross Link, who developed his company and then actually ran Nielsen’s company for a while and is back doing his own thing in his own way. And Steve Cohen at in4mation insights and his partner, Mark, who just keep pushing the envelope. I mean, these people just want to do great work, and it’s so refreshing to see. There’s real pride of their services and their products.
Andrew Mitrak: That’s cool. It’s great that it continues to develop and continues to get better, and it’s an area where entrepreneurs continue to make an improvement for folks. Do you have any other kind of final reflections or takeaways on this conversation?
Jim Spaeth: You know what is interesting? Kind of just go back to your previous question more directly. The practice, under the stewardship of those people I’ve mentioned and others, has continued to push in the directions I always, and my partner Alice, have always seen as where the benefits are—getting faster, getting more granular, more inclusive. And that just keeps happening, which is great.
A big breakthrough was a few years ago, we finally started seeing creative being assessed through these same tools, which was really, really good. So that’s finally happening on a bigger scale, I think, than ever before. And then, the frontier right now is AI, which offers to make certain levels of analysis more possible, affordable than would be if it was all manual.
But the other thing with AI and deep learning in particular is, mix modeling was always a bit of an art and a science. The science was the data and the statistics and the models, but the modeler had to look at the results and go, “That’s not possible. Everyone knows when prices go up, sales go down. So what’s wrong with the model?” Right? That was the art part. You had to have a common sense sort of understanding of how these things work because the model can come up with some cockamamie solutions, and you have to really make sure you’re looking at something that really makes sense from every angle.
And deep learning—I have not practiced it, but as I’ve read about it and understood it—has a better grasp of causality and can give you more assurance that what you’re looking at is actual causality and not coincidence. I think that’s another important breakthrough. And I hope we see that spread quickly through the industry.
Andrew Mitrak: Jim, thanks so much for your time. I really enjoyed the conversation.
Jim Spaeth: Great talking to you, Andrew. Thank you so much.
