
Loading summary
A
Hello, hello, hello. Welcome to another episode of the Always Be Testing podcast. I'm your host, Ty degrange, and I'm really excited to talk to Tom Rathbone today. Tom, how are you, man?
B
I'm excellent. How are you doing?
A
I love that. I love that. Got a little time to take, take care of some stuff yesterday we were talking about the weekend. You got some things accomplished yesterday in your free time.
B
Yeah, yeah. I was fortunate enough to be able to take the day and just bundle it all and just go down the to do list. And I got halfway bit, but that's pretty good. So I got that off my back for a couple of those things.
A
It's the best, man. Sometimes you gotta do it, just take a day. And it. It sounds like, you know, I talk about cowboy code and some of the ranch lifestyle stuff. You've had some experience around horses and that lifestyle yourself, which is pretty cool.
B
Yeah, we grew up with them. I grew up with a horse and working. Like my first jobs are all at down at the stables. And yeah, we live out in pretty much horse country here too, so it's. I love the animals. I can't stand them. They're idiots, but I love them. So hopefully we never get one.
A
Yeah, that's hilarious. It's like we were saying, you kind of know too much sometimes around them, but knowing that is good. But very, very cool, man. A lot in common.
B
Small world.
A
So for those of you who don't know, Tom is an amazing resource, affiliate leader, partner marketing leader, analytics product leader and guest on the pod. And we're excited. It's going to be a good one today. So right now you're focused on product at TV Scientific by Pinterest, and prior to that did some really interesting things in agency measurement and product development. So you've got some good experience to bring to what's been a good roster of folks on the pod. Thanks for coming on.
B
Yeah, I'm so excited to be on here.
A
Heck, yeah, man. So you've seen, you know, years and years of this, the space in affiliate. You've seen some amazing. The always be testing pod is perfect for you because you've seen so much measurement, testing, experimentation, seen scaling programs. What's a big learning that sort of changed how you think about affiliate marketing.
B
Yeah, I think there's a lot. I'll kind of back into this one, I think. So you see a lot of feedback in this space about affiliate marketing not being a set it and forget it channel. And we don't need to harp on that. But I think a lot of that comes from relying on really simple measurement. If you're going to use last click single touch point measurement as a metric for success and that's it. In my opinion, it's just not productive. Like you're going to end up with a program that ends up being a cost center and a time suck and all these things, but it's not actually producing growth for your business. And I think the biggest learning for me that kind of came into that was when I started throwing away last click just entirely and started looking at the nuance in marketing measurement and trying to scale that nuance. So let's scale the look into a linear attribution model or a first click or compare sessions to clicks and see what we learn from there and try and scale that. And I think when you start doing that, you really peel back the onion in a nice way and then you start getting into really productive conversations on a per partner basis. Like what are they doing, where are they falling in a customer journey and how can we get more of the good stuff and less of the bad stuff and ultimately just have more productive conversation that leads to growth and not just a lot of frustration at vanity metrics.
A
Yeah, I love that. And it's such a philosophical alignment to help people unpack where the value is happening, where the measurement is happening, where we should be focused and not just looking at it from a, you know, myopic, you know, once one lens perspective. And I think that's, that's what you're bringing to this, which is. And you have brought to it, which is really fantastic for people.
B
Yeah, I find that there's a lot. The best data analysis that I've seen usually has a really strong storytelling component to it. It's not something you can just put in front of someone's face and say like, here it is good or bad. There's a narrative, there's a story here that we had to kind of unpack because I compare marketing to economics a lot. It's like a dismal science. People do stupid, irrational things, myself included. And we're not trying to understand like perfect intent all the time. There's a story underneath the data that we need to uncover to get to the more productive decisions.
A
Yeah, I love that. Economics is a fun discipline and one with tons of corollary to marketing, right?
B
Yeah, totally.
A
I love that. You know, you've, you've gone through so much in measurement in your career and now you're finding yourself in TV measurement with TV Scientific and Pinterest how is it different?
B
It's, it's a lot different. Especially like historically, you know, TV was probabilistic to a pretty extreme degree with Nielsen boxes and things like that. And fortunately with CTV we've gotten so much closer, but we're still relying on trying to tie together an ad view to something down the road. And I think that TV measurement is really tricky because a, you are relying on more view through attribution. So you're not, there's no, no clicks in the tv, at least not to any significant degree right now like that consumer behaviorism hasn't caught up. And then it's also like these are, these are upper funnel behaviorisms. People discover brands and CTV and there's often another device in play as well. So I think we're trying to tie together more of, you know, are we introducing, are we reaching new customers, are they ultimately converting, are we hiring the funnel and then try and use all the rich data we have in TV to do a better job of that? One of the cool things about CTV is, is that there's a ton of data signal in as well. You can get a ton of signal from some of these providers for, on a content basis, on a viewership basis. That is really, really strong data that you can't really get elsewhere. You don't really get that from other, like on the publisher side, like the affiliate publisher side, into more traditional affiliate marketing. So it's nice to have tie into more upstream data sources.
A
Yeah, I love that. I, I think there's, you know, I was on a panel with your teammates years ago and like our team has just really jumped into this world and been a, been a big advocate and seen some of the value that you guys are providing and, and I think to be able to, I think there's also, and I don't know, not to get too far afield, but I, I feel like there's this trend of sort of this convergence of brand and performance that's happening a little bit. And I think that, and I think it also is the trend of recognizing that affiliate as we know it is actually an enablement of like so many other channels like connected tv. And I think there's this weird misnomer of the name of affiliate and it's, it's not necessarily, it's not like we can just easily slap another name on and assume it's going to be understood, but it's really cool to know that you can get access to that. It's one of the things we really enjoy bringing to brands. And it's so fascinating to hear how that performance and how that measurement's landing for your stakeholders and your clients and your brands that are on Connected tv.
B
Yeah, I mean we can, I'll join you in getting too far afield because we can get pretty philosophical about this. But I feel like for, yeah,
A
yeah.
B
For a decade plus brands, agencies alike have all been clamoring for upper funnel influence. You know, anything beyond what Last Click really measures in the, and the technology, the industry. Like nothing's really caught up to that I feel like because we've had a lot of entrants like TV scientific to the space. But even more recently the, the change in paradigm that AI is bringing in GEO to the mix now there's more pressure to measure things upper funnel and other signals that we're getting like a knock on effect in affiliate and we're seeing those things actually pick up pace now. I think there's more interest and more accessibility and more direction on a product side towards measuring those more, you know, influential touch points that are growing businesses and not just pushing conversions over the line or latching onto something that was going to happen anyways.
A
Yeah. And I think like it's not. Oh, and you might, I know you'll have some strong opinion on the measurement side of it but I feel like from my knowledge the IP based information is just so helpful and to be able to see halo effect, some performance, some brand impact within your ecosystem of Connected tv it almost again this is a little bit of a leap here but it almost feels like there's a flavor of what MMM is attempting to be and that goes into a whole nother measurement question. But I, I just think there's, there's some of that is why is. Do you think some of that is why it's really catching fire in a good way? The measurement, the, the measurability of something that used to be very unmeasurable, like less measurable.
B
Yeah, yeah, I think so. I think that there's a lot more technology around measuring upper funnel impact. I think there's more appetite and more acceptance of those not being perfectly deterministic all the time. I feel like in the affiliate channel everything has to be so auditable to like a pure deterministic effect historically which is great but it's also an impediment to growth. It's not always productive to be able to do that. And if we can get something that is very, very reliable but not perfect all the time, then I think we can grow more, we can test more and we can learn more and hopefully grow a business instead of trying to find something that has a margin of error of zero.
A
Yeah, exactly. Sometimes you got to think a little bit bigger. You mentioned deterministic and I think, you know, you, you've obviously lived and breathed measurement. For those in the audience that are maybe not as familiar with some of the terminology, can you maybe just share more, a little bit more about what you mean by deterministic for folks that are not as measurement centric?
B
Yeah, that's actually a fair question because sometimes I see it used a bit differently. So deterministic will be if we can trace an individual touch point that happened to an actual purchase with very strong data signals, with a very high confidence. You know, a perfect confidence, like a perfect deterministic will be like if we could tie it to an individual email address that they put in. Right. That is really hard to argue against. Probabilistic will be more of like, well, we saw like this fingerprint, this device ID or like earlier in the funnel or a household IP or a business IP and we saw a lift down the road. These are more correlation, maybe really strong correlation, but not perfect causation. So I think it bleeds over. You know, you, you get as you get more as a probabilistic gets better and better. At some point it became comes kind of deterministic. So there's definitely a gradient there.
A
Yeah. What a great description, Tom. Thank you. That's super helpful for people.
B
Yeah.
A
And like it's fascinating to think about how you. I think another piece that I find to be really compelling about what TV scientific and what you've. The team has been able to do is not only have a measurement lever that's pretty, you know, as you said, pretty viable, but also be willing to test obviously the theme of the pod. And so I'd love to hear a little bit more about just some of the things that you view as how does that measurement work for people that are not less familiar, kind of setting it up for true success. How are you kind of seeing folks counsel and do it the right way?
B
Yeah, it's tricky. We've got to make sure that because we're tying together a view on a device, this is cross device tracking. A view on one device with a sail on the other device. We've ultimately got to make sure that the tracking is universal and like rock solid. So we don't want scenarios where we're conditionally firing tags or they're a little buggy or they're slower on the slow on a load Order. We've got to be able to see all of the data coming in because sometimes brands will try and engineer attribution upstream because they see certain signals that they feel are indicative of it coming from a certain partner or a channel or something like that. But in this case they don't see that we see the stuff upstream, those views and we need to see what happens downstream to be able to connect the dots. That has to be very, very tight. And we get into trouble when it's not. When there are gaps. People think and brands think they have a tag deploy universally, but actually they don't. Or they have some conditional firing logic they didn't know they had because, you know, one of the web devs has a certain policy uncertain tag and loading orders and things like that. So we want to make sure that that's like in a really good place for the data in perspective. And then you kind of get into like, all right, data from the actual test model. You know, what else, what are we gonna be testing? And how you kind of structure that overall test media plan, I think has to be pretty strong. Well as well. So you're gonna have a good model of like, all right, what are we measuring against? What is our. What's our control, et cetera. And that's gotta be. That's the other half to it. That has to be. You have to be pretty specific about.
A
Yeah, and maybe like just taking that and riffing on it a little bit. It's sort of related to our next topic. But do you, do you think brands take advantage of like the holdout testing opportunity enough? Do you think they get enough signal that as they should, generally speaking, from your perspective? I find it kind of fascinating of, you know, obviously they have to prioritize and think about resource and time and cost. Those are not insignificant things. But curious to know what your perspective is on like, do brands really take advantage of the testing capabilities as much as they should?
B
I don't know if it's as much as much as they should. They certainly don't as much as they could. I think we often run into that challenge of, you know, is it going to be productive or are we getting too academic with it? And I think that it is often a trap to want to over test sometimes. Yes, because you want to know the answer to it and there's like some anxiety for future budget and you want to protect it and that's all very valid. But if you overdo it, then you're just, you're. You're stuck trying to answer Questions that aren't actually going to move the needle. So I think it's trying to strike a balance of like, what are we testing and what are we going to do with it when we do test, and if we can have a good perspective on that, then I think then you're in. It's worth it to test more.
A
Yeah, I couldn't agree more. And I've seen, I've seen brands get too much minutiae there for sure. And I think there is a happy balance. And I think, like, that's, that's the magic that I think it's hard to strike. I don't see it always struck just right. Often enough.
B
Yeah. Like, if you're going to say, you know, we have a budget of, hey, we're going to spend 50 grand in this campaign over the next year and we want to spend 30 grand testing or something. This is obscure example, but it's always not going to be worth it. But if you're saying, like, hey, we've got a $10 million campaign on the line, we want to take a beat and do a quick early test on like a 50 grand study just so we can feed that into the model, like, absolutely. Makes sense. You should do it. Do it quick.
A
Yeah, Fail fast, learn fast.
B
Yeah, all that fun stuff.
A
What are, what are some of the misnomers you see in testing, you know, in myths, like, obviously you've seen a lot. I'm curious to hear your perspective there. Kind of, kind of dovetails into what we just shared.
B
Yeah, I don't. I think often there's. The studies will come out with some sort of a lift, but is that lift one that's worth chasing? Is this lift going to be something that's going to grow the business and is it statistically significant enough to do so? And I think that often we'll kind of get stuck in that minutia, as you said, and start looking at that and starting to be like, okay, we have a moderate confidence in this moderate lift that was moderately going to grow the business if we scaled it. And I think we kind of. Or it can be that we're going to have a low impact in the business, but a high, you know, a high confidence and like a low lift. You know, there's like those things where we can just get stuck on the confidence rating of what we're measuring and we kind of lose sight for why we're doing the thing that we're doing. And I always fall back on that. Like, all right, what's the point of what we're doing here. Are we going to make an impact with it? And I think there's also just so much. I'm definitely an acolyte of Dr. Foe and all his great content out there. There's just so. Oh yeah, I mean, side note what an absolute like content machine he is. But there's just so much junk data out there and junk measurement and it doesn't make sense when you dig into it a little bit too. And I think that that's when it comes to testing there's such a data fidelity problem and you've got to have such control around your inputs and your outputs and feel really good about those so that the signal you're going to get is one you can bank on.
A
But yeah, yeah, I, I guess I also like the theme that I think you're, I'm picking up what you're saying around kind of sanity checking the direction and the human element and the qualitative element, which is what has come up a lot on the POD recently around the topic of testing and the old. I think Bezos made it famous and a few other people talk about it a lot of like, you know, you run an experiment, the qual and the quant are kind of, you know, tying up or not really decisive. You, you tend to like lean into the human element of it or sort of like what you set out to do as opposed to like reading too much into the data. You know, obviously there's a. Various ways to approach something like that, but I don't know, I just. It sounds like you, you're taking a pretty common sense approach to it, which I think is pretty refreshing.
B
Yeah, I think I definitely, like, even in my personal life I have a bit of a productivity complex. So at some point when we start getting a little academic, I start getting a little like, I get a spider sensei going on and I'm like, hang on, why are we doing what we're doing here, guys? Are we getting lost in the smarts of it all? And it's really hard. I work with so many brilliant people much, much, much smarter than me. But sometimes we get stuck on the smarts and we get. Start drinking our own Kool Aid. But it's we, we've lost the productivity signal on it. We need to back up for a second. And recenter. Yeah.
A
One of my dad's famous cowboy quotes, as I'll call them, is don't mistake motion with progress.
B
That's a great one.
A
You said it. You said it as well, which I think is awesome. This is a good one around the testing topic, which could be more appropriate. But like deciding what to run, what test is worth doing, what, what kind of guides you and what. What's the framework you use to think about experimentation?
B
I mean, so you asked what guides me. So I'll take like a personal perspective on it because it's certainly not like the right perspective. But I tend to like, I tend to like bigger swings. So I want to have a plan for what are we doing after this. Like, where are we looking to go? What's the thesis that we're looking to prove and how far are we looking on that? If we're looking to do something kind of optimization or incremental improvement level, like, okay, then sometimes it's really not worth doing. But I want to have a plan for like we, we want to do this, take this big swing, let's backtrack and build out a model for what that swing needs to be. And then here's our, like our core model that we're going to start feeding data into with that testing. So we, we want to have a means to that end. And is that end compelling enough to justify all the work that ladders up to it? And if I don't have a good, not like perfect, just directional plan on like what this is going to look like and what we're kind of going towards, then I'm probably not going to look for something else that's going to be more compelling. It's going to be easier to get resources for because it's more exciting and there's more upside and then is going to be, you know, worth the squeeze down the road.
A
Yeah, I love that. Spoiler alert. I think that is the Tom approach is the right approach on that one. I think you have to have enough juice there to justify all that effort and time and energy. And I just. Yeah, I think, I think you're on.
B
Yeah. But it's also got, it's. You got to have like a plan for what you're going to do with it too. You get not like a perfect. We don't have to design the whole product around it. But just the worst thing is when you have a vague idea of something and then you do a test and the test is great. And then, and then you're stuck at like now what you don't have a plan for like where you're going after that. And then you just, again, it's. The excitement wears off, it gets deprioritized and it gets relegated to the world of academia. Then you Just get stuck in like this cool thing that you did. But the business is off looking at other bright shiny things because you weren't ready for then what, then went and then what. I think you need like that phase approach. We're going to get the data, we're going to feed into the model that's going to justify this next level of investment where we do these new things and build these new things and kind of have everything ready.
A
I love that planning and prep and we've instilled and talked about that a ton around the hypotheses documentation. And then like if we, if this does prove out, then what do you do about it? Which I think you're totally highlighting.
B
And are we ready to do that too? Because you don't want to test too early. Otherwise if you, if you test, then six months go by and then you're ready to build something. But gosh, right now six months is a freaking eternity. Like everything has changed. So you gotta test again.
A
You know, that's, that's a really good point. Yeah, that's, that's amazing. Obviously, measurement's your strength. What do you, what do you think you're bringing, you know, to the Pinterest team within TV scientific team around, you know, kind of principles of measurement, if you will. And I think you've touched on some of those themes already, but just curious.
B
Yeah, I mean there's a lot of great, great people here that measure that are just brilliant minds in measurement here. It's just a pleasure to work with them. I think one thing I'm bringing, which I've always kind of brought around with me wherever I was, is my, and we talked about earlier too, I did talk about earlier that like getting away from these single touch point measurements, especially last click type measurements, and then trying to focus on broader attribution models, even first click, but then more linear, more fractional models. Just where are we contributing and where are we driving growth? And how can we kind of measure that and get away from these more like classic affiliate vanity metrics. And given that TV Scientific and falls into kind of a affiliate ish, right on the edge there, sometimes depending on the brand more than others, we can kind of get stuck on those single touch point measures of success. And that doesn't do a good job of measuring whether or not this partnership is productive. And I want to always focus on, even if it's not our measurement, even if we're integrating with an MMM or MTA partner, having that data there and understanding it and learning to storytell with it so that we can have like a better growth strategy with that brand longer term and as opposed to a one off campaign, was it good or bad? We can say well here are all the signals we saw from it across the funnel. Like halo. We had a great halo. We got beat out by you know, these other lower funnel touch points. That's fine. But we contributed in these ways. I think that kind of tells a story of like now where do we go next quarter? What do we change? How do we invest more, more budget and do something differently on top of that. Yeah,
A
you mentioned MMM a little and reminds me of some conversations I've had around affiliate and connected TV sort of separately ironically based on our previous mentions but in the context of like a CMO fractional CMO heavy measurement approach like very rigorous measurement and hold out testing and incrementality testing and a lot of very smart objective individuals that are looking across multiple channels but also have some expertise in connect TV and affiliate many of them found and it's not always going to be the case with every test or every brand but really positive correlation and positive view on the incrementality of affiliate done the right way and, and connected tv. So I think it's kind of, you know, it's not always the case, it's not always going to win every time. But I think there's a, just a really good body of evidence to support, you know, what essentially we're both doing in a way that's like hey, we want to put these things up to the test. We, we don't want them to claim they're the winner when they're not. And in fact they're frequently coming up as a, an incremental, incremental lever for brands. So I figured you have some thoughts on that point but I just think it's, it's so nice to see it being able to be put up to the test. And I like to see that. You know, I think you and I are both aiming for that for the industry.
B
Yeah, I mean CTV benefits from some of that. In a lot of those like more CMO level conversations and mmms they build upon impression level data and that's what CTV uses one of. I think boy, we can get philosophical real easily on all of these. I think one of the biggest disservices that the entire affiliate channel has done to itself over the last decade is not prioritizing impressions and just accepting it and not pushing harder for that. And I understand technically why they don't exist and they're unreliable but These are choices we made and that has resulted in lower spend and lower trust and confidence over time. And had we 20 years ago put the basis on like we need to make sure impressions are always there, I think the conversations over the last 20 years would have been a lot different.
A
Yeah, amen to that. That's spot on. We're, we're moving and counseling people there quickly and it's, it's, you know, there's a lot of things that make it tricky, as you know.
B
Yeah, I mean it definitely opens the door to like crappy impression data. Right. And that needs to, you need to have a solve for that. But that is, that is a very accessible solution. There's technology and platforms, a ton of them out there that can help you validate that data as being legitimate impression level data. Let's use that.
A
Yeah. Yeah, I love that. I love that. We talked a little bit about, you know, redesigning or thinking how could you think through affiliate. And it's sort of, you kind of touched on a great one around impressions and exposure. But there was something else that you think you could expand on redesigning affiliate ecosystem. What would you consider making changes to or would you suggest
B
besides getting rid of like the last click thing, which everyone talks about a lot and thankfully people are talking about it more because I feel like I've been screaming into the void for a long time on that one. But I also feel like one of the better things that when I've worked with brands Direct and the better things we could do, more productive things we can do is to stop using affiliate network data as your source of truth for performance. Use it as your source of truth for cost. Because it is. But look elsewhere. Look in GA4, you can do a great job with it. You can get a lot more out of that than you probably are. But then look at other things, great platforms like Triple Whale, like North Beam, like Rockerbox, whichever, and start looking at it on a per partner basis outside of that network. Do the do the things. Tag it appropriately. So you can do that. That's like table stakes. But you've got to pull out of using those affiliate platforms as like the measurement for success because invariably it's only you can do a lot with these networks. I know you can measure a lot of channels in them and you can do a lot more, but those are edge cases. Most brands aren't set up to do that. So it always has a narrow view and it's not productive. You're not going to get, you're not going to be able to scale the nuance that the data can provide for you in other platforms if you're just using an affiliate network.
A
Yeah. It's amazing to me, similar to what you're saying, how many people are overly reliant on last click, how many people over reliant on the network. And like you, I mean, we absolutely love diving in and looking at data in house data, other platform data, doing some real proper comparison around that because like you said, it's, it's not telling the whole story, which is really unfortunate.
B
Yeah. I think it does everybody a disservice. I think there's a lot of, on the brand side, a lot of affiliate managers who are managing the affiliate channel. Before they came on board, success was measured based on the data from the affiliate platform and then they came on board and they're just kind of continuing that because that's what the, that's what the, the company kind of the structure of it demands. But I don't think it does them a service because I think you're gonna, someone's going to come in that room and question the actual growth that's happening and you're not going to be able to defend it because you don't have that perspective outside of that. So you got to almost retrain the business around you and I think that's where you're going to lead out and get more productivity out of the channel.
A
Yeah. Nailed it. Tom, this has just been amazing. It's been awesome to have you drop knowledge. Have you come in and chat. We always have. Always have such good conversation. Just coming down the home stretch to get to know you a bit better. What's something that people might not know about you?
B
This is like a really hard question because I'm, I'm, I'm. I'm like privileged to have a lot of really great friends in the industry and those friends love to pick on me and embarrass me incessantly. So I feel like everybody knows everything about me because there's, because I just feel like everyone's little brother sometimes and they just kind of rat me out. Am I embarrassing things? So I don't, I don't know, I
A
need to step up my game and giving you more shit.
B
That's, that's, yeah. Oh, there's a, there's a whole bunch of bullies you can join in with. I think it's also tricky because I'm a bit of a hobby, like a hobby hoarder. So I've always got something going on, something new that I'm working on or building or. Or doing with my hands. I think that's just kind of like me. Outside of. Outside of work is still kind of work, but I just like to move things forward and test and learn. And whether that's building something or I'm doing a lot of painting lately or backpacking or something like that, that tends to be. Yeah. Hobbyist, collector, I think is the thing that people may not realize about. It's a problem.
A
I love that. And you know, you joined the ranks of dads. Congratulations.
B
Thank you. Thank you. Yeah, we just had our four month appointment this morning. She did great. She did great. She did not like the shots. They were tough. It was a new scream we discovered this morning. But it's been a great journey. I love being a girl dad so far.
A
That's amazing. Nothing better. That's so cool. Anything that surprised you or you thought about that was maybe better or worse or just. Any takeaways from the four months so far?
B
I talked to you about this one, but I'll repeat it because I'm still flabbergasted by the time paradox that is the first few months where I can't believe it's only been two months since we were back in the doctors because it seems like it's been a year, but it also has flown by at the same time. I don't understand how things can seem so long and so short, so fast and so slow at the same time. It is an absolute like. Yeah, it's. It's melting my brain sometimes. And that has been very, very interesting to realize. I understand that it'll go a lot faster. The. It'll fly by soon like that. That's gonna pick up speed, especially now that she's moving, she's rolling around and stuff. But for now, I'm just kind of enjoying the. The mind melts. That is the time paradox.
A
That's awesome. I couldn't agree more. It's like you don't know what's up or down or how quickly or slow things are going and you're. It's. It's like joyous and then it's hard and you're sleep deprived. And it's been a wild one for us too.
B
Yeah, it's. It's been fun. I mean, I don't know about you, but I'm my, my second job is definitely like a baby photographer now. But I spend like so much of my time just I, I have to look back on past pictures from months ago to realize how far we've come because when she was just like a little tiny peanut on my chest. That feels like yesterday. It also feels like last year, but I had to look at, like, physical evidence of that to realize how big she's gotten. Just four months.
A
Wild. Yeah. It's crazy. It's crazy. Congratulations and thank you. Yeah, awesome, man.
B
I'm happy to join the. The dad club that you've been in for a while now. It's. It's great to be a part of it.
A
Yeah. Welcome. It's so much fun. It's the best. Best thing in the world, as you know. So. Yeah, thanks for sharing. Thanks for coming on, Tom. I think this was a great episode on it Just totally nailed the. The theme of the pod and your measurement experience is a phenomenal fit. And just thanks for coming on, man.
B
Appreciate you, like, literally anytime. It was a blast. I really, really enjoyed it. Thanks for having me.
A
Heck yeah. Talk soon. Thanks, everybody.
B
Later, Tyler.
Host: Tye DeGrange
Guest: Tom Rathbone (Product Leader, TVScientific by Pinterest)
Date: May 19, 2026
This episode dives deep into the evolving landscape of measurement and experimentation in affiliate and partner marketing, with a special focus on Connected TV (CTV). Host Tye DeGrange is joined by Tom Rathbone, an analytics and partner marketing leader at TVScientific. Together, they unpack incrementality, attribution models, the shift away from last-click measurement, and the real-world complexities of testing and scaling in CTV and affiliate marketing. Tom shares actionable insights from his experience on the agency side, product development, and life inside a top CTV measurement company.
“If you're going to use last-click single touch point measurement as a metric for success and that's it...you're going to end up with a program that ends up being a cost center and a time suck...but it's not actually producing growth for your business.” (02:33 – Tom)
“You really peel back the onion ... and then you start getting into productive conversations on a per-partner basis: what are they doing, where are they falling in the customer journey?” (03:29 – Tom)
CTV Measurement Challenges:
Convergence of Brand and Performance:
Industry's growing appetite for upper-funnel measurement.
Acceptance of less-than-perfect determinism to unlock testing and growth:
“If we can get something that is very, very reliable but not perfect all the time, then I think we can grow more, we can test more...and hopefully grow a business instead of trying to find something that has a margin of error of zero.” (10:32 – Tom)
Deterministic vs. Probabilistic Attribution:
“Sometimes I see [‘deterministic’] used a bit differently...Deterministic will be if we can trace an individual touch point...Probabilistic will be more of like, well, we saw like this fingerprint, this device ID or...household IP and we saw a lift down the road.” (11:19 – Tom)
Holdout Testing is Underutilized:
Common Misnomers & Data Fidelity:
“There’s just so much junk data out there and junk measurement and it doesn’t make sense when you dig into it...you've got to have such control around your inputs and your outputs.” (18:37 – Tom)
The Human Element:
“Are we getting lost in the smarts of it all? ... Sometimes we get stuck on the smarts and ... we’ve lost the productivity signal.” (20:43 – Tom)
Go for bigger swings, not just minor optimizations.
Have a clear thesis or “then what” after the test—ensure results can be operationalized.
“You need that phase approach. We’re going to get the data, we’re going to feed into the model that’s going to justify this next level...” (23:23 – Tom)
Don’t test too early—move quickly to application or risk obsolescence.
Push beyond single-touch, last-click metrics—embrace advanced attribution models and cross-channel insights.
Impressions matter: CTV’s focus on impression-level data is a missing piece in affiliate; ignoring impressions hurt the industry’s credibility and budgets.
“One of the biggest disservices that the entire affiliate channel has done to itself...is not prioritizing impressions...” (28:53 – Tom)
Use network/platform data for cost, not performance measurement; integrate with external analytics tools (GA4, Triple Whale, North Beam, Rockerbox) for richer insights.
Retrain the business to look beyond the affiliate network dashboard for true performance.
On Last Click:
“The biggest learning for me ... was when I started throwing away last-click entirely and started looking at the nuance in marketing measurement.” (02:40 – Tom)
On Data as Narrative:
“The best data analysis that I’ve seen usually has a really strong storytelling component to it. ... There’s a story underneath the data that we need to uncover...” (04:16 – Tom)
On Being Practical:
“Don’t mistake motion with progress.” (21:22 – Tye quoting his dad)
On Test Planning:
“The worst thing is when you have a vague idea of something and then you do a test and the test is great. And then you're stuck at ‘now what.’” (23:23 – Tom)
On Affiliates & Impressions:
“One of the biggest disservices that the entire affiliate channel has done ... is not prioritizing impressions and just accepting it and not pushing harder for that.” (28:53 – Tom)
Authentic, candid, and pragmatic. Tom and Tye discuss sophisticated measurement topics with both technical rigor and plainspoken, story-driven clarity. The episode balances philosophical considerations with hands-on advice, integrating real examples and friendly banter throughout.
For those navigating modern affiliate, partnership, or CTV growth, this episode is a masterclass in measurement mindset, practical execution, and the importance of always challenging your assumptions and frameworks.