#127: Is Multitouch Marketing Attribution Dead? Should It Be? With Priscilla Cheung

Multi-touch attribution is like fat free cheese: it sounds like a great idea, it seems like technology would have made it amazing and delicious by now, and, yet, the reality is incredibly unsatisfying. Since we’ve recently covered how browsers are making the analyst’s lot in life more difficult, and since multi-touch attribution is affected by those changes, we figured it was high time to revisit the topic. It’s something we’ve covered before (twice, actually). But interest in the topic has not diminished, while a claim could be made that reality has gone from being merely a cold dishrag to the face to being a bucket of ice over the head. We sat down with Priscilla Cheung to hash out the topic. No fat free cheese was consumed during the making of the episode.

Items Referenced in the Show

Episode Transcript


00:04 Announcer: Welcome to The Digital Analytics Power Hour. Tim, Michael, Moe, and the occasional guest discussing digital analytics issues of the day. Find them on Facebook at facebook.com/analyticshour, and their website, analyticshour.io. And now, The Digital Analytics Power Hour.


00:28 Michael Helbling: Hi, everyone. Welcome to The Digital Analytics Power Hour. This is Episode 127. Well, according to our content calendar, we haven’t bitched and moaned about multi-touch attribution for quite some time. 91 episodes, in fact, when we last had Jim Novo on to talk about it. So, Tim, is attribution dead, then?

00:54 Tim Wilson: I wish it was dead. I’m doing everything I can to kill it, unsuccessfully.

00:57 MH: Clickbait. All right. Well, Moe, we do have the top five attribution hacks for marketers. Do you think CMOs are gonna hate number three?

01:06 Moe Kiss: Oh, I’m just hanging on for dear life. Like, you know, when you’re on the bonnet, like nails digging in, not ready to let go.

01:15 MH: Oh, I would totally pay to see you in an action movie, Moe.


01:22 MH: Analytics action star, Moe Kiss. Yes, I am Michael Helbling, and attribution has been mostly a cop out, in my opinion. But, that is until you meet someone who’s using it in really cool ways and is really smart about it, and then I always get energized about the possibilities all over again, and it’s amazing, and everybody’s gung-ho once again. Well, by now, our listeners have a pretty good sense for how the three of us probably think about this topic, but we wanted to bring a fresh perspective. Priscilla Cheung is an analytics and conversion specialist at Prospa. Prior to that, she’s held marketing analytics roles at THE ICONIC, Helloworld Limited and Datalicious. We are glad to have you on the show, Priscilla. Welcome.

02:08 Priscilla Cheung: Yay, thanks.

02:08 MH: Hey, thank you.

02:10 PC: I don’t actually know how I feel about attribution.

02:13 MH: Well, we’re gonna try to tease it out right now.

02:16 TW: Try to make her defend it. She’s like, “Wait a minute.”

02:19 PC: I feel like you’re all on a one side of a fence.

02:23 MH: Well, first off, let’s just start with this. Moe says that she’s basically learned everything she knows about attribution from you. What have you taught Moe, and why? No, no. [chuckle]

02:37 PC: I can probably say why I care about attribution. It’s because it’s a lot to do with understanding the customer, and that’s where it’s come from. And so, everything that I’ve tried to impart to Moe and to everybody else is like how important attribution is in understanding how customers behave with you and to be smarter with the money that you spend on marketing. One of my pet peeves is how much money is wasted on marketing. That’s essentially what I’ve been trying to help people understand, is that attribution is not necessarily a perfect solution, but a way for us to be able to understand a little bit more about that.

03:21 TW: So, is it dead?


03:24 PC: In a sense, yeah. Well, can it be like a zombie, where it’s reincarnated with a bit of something else?

03:33 TW: Mm-hmm.


03:34 TW: I do feel like we… When we say… We say attribution, and a lot of times, that’s used as shorthand for multi-touch attribution, and I think attribution can’t be dead. You can’t go to marketing and say, “Hey, now that you have all this digital data, just spend it wherever it feels good. Instead of half of your marketing dollars being wasted, why don’t you waste them all, but you won’t know about it?” That, to me… I guess, defining multi-touch attribution, the way that it seems like I have most seen it classically defined is, you are trying to track all of your marketing touches and interactions over the course of a customer’s journey so that you can then assign value to each of those marketing touches. Is that a fair common definition of the platonic ideal of what multi-touch attribution is?

04:29 PC: Yeah. In my understanding, yes, though in how it’s actually dealt with practically, I feel like a lot of companies are just using that to understand the online journey. A lot of the time, what we’re missing is that offline journey when companies use above-the-line media and other types of… Even traditional direct marketing in that journey. And they’re also missing that journey in terms of when the conversion isn’t necessarily online, you actually convert offline. And so, multi-touch attribution, I think, covers that whole aspect of the journey, but I don’t think a lot of companies are using that well. But why do you think it’s dead? What’s the… Yeah, behind that.

05:12 MK: Because all the bloody marketing platforms. I feel like they are just making it harder. And so, really ironic twist of fate, Pris has worked her entire career in marketing attribution and now is working in product analytics, and I’ve done the opposite, where I was working product analytics, and now I’ve gone back to marketing. And I kinda feel like I’m banging my head against a wall because all of the platforms are just making it harder and harder and harder for you to actually attribute yourself as a company across all channels by basically being like, “Here is what we tell you about how we’re performing. That’s all you got, so just make do with what we’re giving you,” which tends to often just be like last-click or they’re setting some weird attribution window that they think works or… Basically, you just have to accept it at face value ’cause they’re not giving you granular enough data, which means that you can’t do your own attribution, you just have to take their word for it.

06:11 TW: You’re talking about the platforms that are reporting attribution, or are you talking about the… Right, I think it’s struggling. Well, one, that’s only one solution for doing attribution. And it is kind of a “We’re just gonna take what we arbitrarily executed and then try to use that data to build a model,” I think is a huge fucking misconceived way to even go about it. It is trying to shortcut everything that social sciences and pure analytics has learned where you actually recognize their… It’s trying to have no cost of data, just saying, “Do whatever you’re doing. If you invested the exact same for five years in every channel and you have no variability, how would any model actually work?”

07:03 TW: But it’s also struggling, because even in the online, and I completely agree, Pris, that the… I have a diagram that I talk about, the, “Oh, you’re only stopping at the lead. You’re not actually going offline to the opportunity, the conversion, or even the customer lifetime value.” That’s actually a little more solvable ’cause you know who they are and you just have to wire your systems together, but as even just in the digital world, trying to link back to… With cookie deletion, your digital platform has to recognize that you came a month ago from paid search and now you’re the same person coming back through paid search or organic search, certainly trying to bridge cross-device and recognize that’s the same person. All of the regulatory-led GDPR, CCPA, but now every browser and app platform ratcheting down on the ability to track, to me, that’s what’s really dying, is this idea that we can actually see those touch points. But then, I would also say, maybe my last little rant and then I’ll just go drink, but…


08:21 TW: Even when looking at advanced attribution or multi-touch attribution, when it was a little more workable, before even those limitations, there wasn’t that much variation. Last click is easy to beat up on, you can talk about why that’s flawed, but if you say, “Well, let me compare last click to linear to… ” Which again, flawed to say, “I’m just gonna pick a model,” but if I compare a bunch of them, a lot of times, my paid channels, they didn’t move that much in the first place, and now, I’m introducing a lot of weird challenges when it comes to actually acting on the data.

09:00 PC: I don’t know. I think as Moe said, I was very in attribution back in the day, [chuckle] when we did have display impression logs available and we actually stitched that to the customer journey. And yeah, it didn’t change the overall share of the pie that display or social had, but it did change the investment in terms of the efficiency of the channel to a point where it did justify, I guess, a bit more of the upper funnel spend a bit more. But in that sense, I don’t think attribution can be used alone. A lot of the stuff that you’re talking about in terms of it being a lot harder to track customer journeys is [chuckle] actually what I think makes it exciting because we have to be a bit smarter with the way that we’re actually doing it. And I don’t know, I think I just like a challenge but I’m not… Yeah, as Moe said, I’m in product analytics now, [laughter] so I’m not supposed to get too excited about that stuff.


10:00 MK: Yeah, but a challenge, I feel like it’s not a challenge. I feel like it’s becoming impossible. And I’m actually at this fork in the road where I’m like, I just feel like I should forge ahead and build everything based on last click, and know that we’re gonna get to a point where there is no other option, so why… I watched you spend months working on attribution model.

10:24 TW: There are absolutely other options though, right?

10:28 MK: Okay, so what?

10:28 TW: The other option is to actually do design experiments. You’re not gonna be shifting the lever right?

10:32 MK: Oh, yes, yes.

10:34 TW: Right, but that’s because we do live in that. And I’m with you, I get sucked into it as well. We live in this like, “Oh, we’re losing the view through. We’ve got these Facebook and Google walled gardens. I’m screwed.” I’m finally agreeing with you. Last click, for what it is, that’s low latency, we can see it immediately. What we capture today, we don’t have to recompute, we don’t have to wait for things to play out. But if you really wanna care about… It starts to blend into the mix modeling world, but if you really wanna know, is display contributing what the media agency, the smoke they’ve been blowing up our ass about, that they’re doing all this amazing awareness stuff that is generating oodles of money, well, fine, let’s do a controlled experiment. And I can’t do it on a campaign by campaign or on each piece of creative, but that’s a completely different… And I think this is kinda where Jim Novo was low these many years ago when we had him on, and I’m increasingly thinking, “Yeah, if you really wanna do it, you have to invest in measuring it well.”

11:37 MK: Yes, but if you are a company like Canva where you’re worldwide, you have like a gazillion users every month, I feel like that’s not as difficult. If you’re a company like THE ICONIC, where you’re just in Australia, you have a much smaller market. Some of those tests are actually really freaking hard to do. And yeah, we played around with geo lift tests and that sort of stuff. I don’t know, I think I get it, yes, and I’m probably gonna go down that path of doing some tests, but that’s also not… That’s a one-off, “Let’s run a test and see what the impact is.” That’s not something that a marketer can use every week to report their numbers.

12:24 PC: This is what I mean about the hybrid option in terms of starting out with an attribution model to fill in the gaps on what we can see of a customer’s journey. But at the retail company I working for, essentially, we had an attribution model and then applied lift test on top of that. We switched off spend on some of our paid channels across particular regions. And from the learnings of that, we could see that the efficiency of channels was different to how the attribution model was reporting on the efficiency of the channels. And the exciting part is being able to take those learnings and adjust the model based on the learnings of the lift test.

13:08 PC: I don’t think, yeah, you can have one without the other anymore, but it’s a cool space to learn at the moment.

13:17 TW: Well, so that’s… So to understand, you were saying, “Let’s do some experimentation.” And there can be… Moe, to your point, the controlled… It can be tough if you’ve got a limited region. But the other way you can do it is, say, “I’m gonna build a really strong forecasting model, and now I’m gonna deliberately drop my spend on X or increase my spend on Y,” and then use kind of a causal impact, a forecasting model, to say, “I tried to hold as many things constant as possible, and therefore, I could quantify the lift from this,” and then I think Pris, what you’re saying, whether you did it that way, whether you did it through on-off in different regions, you say, “Let’s take that and now feed that in as kind of adjustments to whatever kind of lower latency, more real-time model. Like you could have a last-click with some adjustments incorporated in it. Is that describing it well?

14:11 PC: Yeah, it explains it a lot better than I did. [chuckle]

14:15 MK: He has this annoying ability to do that. I don’t know. I’m just still feeling really pessimistic about this whole thing, and I don’t know if it’s because I’m doing marketing analytics stuff again. But there’s this bit of me that’s just like, I just feel like it’s gonna get harder and harder. And yeah, I get that there are workarounds, but still, even when you’re doing a geo lift test, it’s not perfect, it’s not even close to perfect where you’re like assuming that one postcode is going to behave like another postcode. And yeah, there’s always gonna be some degree of error, and I actually think that degree of error is very big, and we’re gonna be making huge assumptions, I don’t know.

14:56 TW: But that’s part of it, is an education of the market, right? We do as analysts, we’re such pleasers, that we fall into… Well, the marketer has been conditioned to think that we have all this data at our fingertips now, and thank-you vendors have sold that promise. And so, the marketer who can stamp their feet and be very petulant and say, “What do you mean? I was led to believe you were gonna build me a perfect attribution model.” Well, there’s part of it that says, “Well, you’re wrong, you’re not going to get that.” And then reset the conversation to say, “And part of it may be, you know what, in order to get something that’s not the dreamy ideal that you want, but to get something in that direction, you’re gonna have to invest in this, and investing means you’re gonna have to agree to vary your spend, you’re gonna have to design some experiments.” And I think that is something we haven’t… We want to say, “But this is what they want,” it’s like, well it’s good to know, it’s important information to know what they want, but they can’t have exactly what they want at no cost with just an analyst who just crunches the numbers. There could be a cost incurred to get them in that direction. And we just have conditioned our stakeholders or the way we interact with them I think too often to not recognize that.

16:17 PC: Yeah, you’re so right. I think stakeholders are so much more willing to pay for a vendor solution that tells them that they can apparently have a full view of their ROI, rather than actually switch off spend in a particular region. If you can get that buy-in, that’s amazing, but I don’t think a lot of companies at that point… [chuckle] We were actually really lucky to be able to test around with that, because to switch off a whole channel in your main regions during a sale period is not an easy win, but if you can get there.

16:53 TW: But then you at least have that discussion, say, “Maybe it’s not worth it, but here’s what you would get.” I do now do this quite often, going in, where we work with clients and say, “Let’s sit down and just do an assessment of what you have and what you realistically can achieve.” And we’ve had some crazy ones where they are a tiny market with half of their spend being in radio and out-of-home in a small region, and they’re like, “Should we be buying an attribution solution?” It’s like, “For the… No, you absolutely should not.”


17:31 TW: And they’re like, “What do we do? We need to have this perfect fidelity,” and it’s like, “Well, we can do a regression against your spend and your results, and show you that your investment… Seasonality is dominating what you’re doing, you need to massively up your spend.” And so it winds up being an education, which I get is tough, but well, it’s kind of the reality. But to Pris’ point earlier, sometimes, and we’ll say this, “What if you don’t worry about the fancy attribution models? Why don’t we just connect your online to your offline? What would that give us if we actually had a better visibility from the digital downstream to the offline?” Is there something we could do with that, that would actually be more valuable because that still requires an investment? That’s the way I kinda try to sometimes break it down, saying, “What’s the full breadth of the customer journey that you’re wanting to attribute?” You can move all the way upstream to print offline, even the view through impressions, you can move much farther downstream to the offline conversion, even to customer lifetime value. Yeah, in a perfect world, you have perfect visibility, but you’d be living in 1984, and that’s not gonna happen. Oh, my God.


18:55 MK: But so, realistically, how much… A lot of the things we’re talking about here, ’cause I watched Pris do this work, it’s like super time-intensive. Just in hindsight, I suppose, do you think it’s worth the time for the small improvements that you get on understanding your marketing spend? It’s more than a full-time job, just on trying to tweak and run tests and…

19:30 TW: Did you drive the lift in results to… That’s the perfect question, and it’s really hard. It’s like, “Really? That means we have to have… ” Yeah, I don’t know. Pris, you can answer the question [chuckle]

19:43 PC: I think I would say lift testing makes a lot more of an… It seems to make a lot more of an impact than maybe even the tweaks that I made to the actual attribution model, because you can tell straight away what the ROI is during that period. And if it is shown to have a causal impact, you can use that straight away. There are so many tweaks that you can make to an attribution model if you do it to a customer’s first order versus their recurring purchases and things like that. At the end of the day, it’s ultimately what the company is driving, what their key objective is. Is it more about efficiency? Is it more about retention and that type of thing? ‘Cause that’s where I would decide where to focus on in terms of whether I do attribution versus stitching the offline journey versus looking at customer lifetime value. Those are all questions based on the key objectives of the company at that point in time.

20:41 TW: Well, so when you first started talking about it… This is another thing that has come up ’cause I’ve been talking to some customer data platform technologies as well. And then also, this whole idea of customer journey analytics has come in, which to me, it’s very simple, that customer journey analytics is heading down the exact same path of multi-touch attribution, which is a waste of money and time, as opposed to, at least there, we’ve got customer journey mapping, where you can say, “Let’s take our research and let’s do something that has been being done for years and maybe done well, and too many companies don’t do it.” But one customer data platform I was talking to today, and sort of had the mini “Aha!” that if you could capture all of this data across devices all the way through the entire customer journey, that raw data set is the exact same… That’s what customer journey analytics platforms are promising that they’re doing as well. And I kind of wanna just grab somebody by the… Anybody who’s thinking, “Yeah, you know what? We’re not gonna try to do multi-touch attribution anymore. We’re now just gonna do customer journey analytics.” I’m like, “Well, the… “

21:55 PC: Yeah, all the same problems.

21:57 TW: The data collection problem is still the same; it’s a different problem with what you’re doing with the data, although some of those platforms actually show Sankey charts and all the things that we’ve learned, or kind of look pretty but are worthless in a digital analytics platform. I’m like, “Ugh, really? Stop buying technology.” If you took those dollars, it’s a weird… The accounting doesn’t work. If you said, “We’re gonna spend $100,000 with a multi-touch attribution platform, but you said, “Well, what if we instead adjusted… What if we increased our spend in this channel, or decreased our spend in this other channel? And yes, we would lose some revenue, but would we lose that much revenue?” We don’t think of the investment in varying our spend as we think of the investment in a software platform. I don’t see… That equation just is not… It’s different parts of the organization who control different types of budget. You’re comparing apples and oranges, but they’re both measured in dollars. So why does it matter?

23:10 MH: Yeah, it kind of brings back this central piece to me about attribution. And I think journey analytics, Tim, you’re very correct, is sort of the next kid on the block in the same vein. But there’s this sense of either deciding to have the attribution tool tell you what to do versus you setting goals and then trying to line your marketing channels behind and serve that better. So in the one hand, it’s like, “If I could get multi-touch attribution to work perfectly, it’ll tell me when to send that third email and how many impressions in display I should get before I target them in social.” All that stuff is so wishful thinking. But if you go back over to the other side and say, “I wanna grow new customers by 10% without increasing my spend overall by more than 5%. Okay, now how do I do that?” And now, you’re gonna start looking at your channels and being like, “Where can I get the most bang for the buck for that type of customer growth, or that… ” And so, your strategy then dictates how you pursue the problem, as opposed to just sort of thinking that magically, if you can have it all, it’ll just answer every question and do it for you. That’s sort of what I think people get stuck in a trap because… And again, we do this to ourselves and none of us are smart enough, apparently, to realize we do it over and over again. Half-reporting 3.0.


24:35 MK: Yeah, but see, when you say customer journey… What did you call it? Customer journey analysis?

24:40 TW: Customer journey analytics.

24:42 MK: To me, that just sounds like product analytics. It’s like once someone gets to your side, then everything after that.

24:50 TW: No, no, no, the promise is that you’re actually… “Look, you could visualize from when they saw… When they opened the magazine in the doctor’s office to the point that they… ” They literally will show Sankey charts all the way from the influencer’s Instagram account, that’s part of the journey, all the way through to… [chuckle] No, but that’s the… Yeah, we’re all laughing ’cause it’s obviously… That’s not gonna work. But that… It gets sold.

25:17 MK: But so… Okay, does that mean the only solution is to do it in-house because… It’s hard to do in-house where you do have access to all of the ‘data’ and the tools are just a load of shit?

25:33 MH: No, I’m not gonna go as far as to say that. It’s more of your approach to any of the tools or whether… What you do yourself, in my opinion, I don’t know. Priscilla, you do it for real, so maybe you could weigh in on that.


25:47 PC: Yeah, I would totally advocate doing it in-house. There’s so many times where I’ve seen companies buy tools and have the results shown to them in some flashy dashboard format but not make any spend changes off of them. In terms of doing it in-house, I’d rather spend money per se, as you said, Tim, on shifting spend across the channels and also on an analyst resource, so you can actually influence the stakeholders to actually change their decision-making because of it. Because at the end of the day, if we spend money on platforms and we’re not making decisions off of that, then it’s just another way to make our… Justify that our marketing is looking good or not good, but nothing is happening off the back of that. Customer data platforms are maybe another thing that [chuckle] I don’t know if we wanna get into, because that ties in the path as well as what people are doing in terms of product analytics, and then that’s meant to tie in the whole journey together. But is that a whole big topic to get into or…

27:01 TW: [chuckle] Well, there’s two ways… To me, there’s two ways to go… I think there’s value… If you do try to do customer journey analytics or multi-touch attribution, the person who’s really on the hook for it is gonna learn a lot about what data they actually can track and what they can’t, and they’re also gonna learn a lot about the willingness with the organization to make a change, and I feel like CDPs… It’s tough, ’cause on the one hand, you could say, “Yeah, why don’t you see what data you actually can get together? If you have it all in this centralized place, there’s a lot of stuff you could do with it.” That still winds up being a leap of faith of, “Oh crap, what if… ” It’s a flashback to data warehouses from 15 years ago. “Let’s just get it all in one place and then there will be all these cool things we can do with it.” And then, you get it all in one place and realize that all in one place isn’t quite as comprehensive as you thought it was going to be. It’s a challenge. Whereas if instead, you had somebody who really was knowledgeable…

28:06 MH: And when Michael, you were saying, “Hey, these are the guardrails we have. What can we do?” And it’s like, “Well, let’s roll up our sleeves and let’s do some experimentation.” Or, “Let’s find this dead period and in this period we expect to not have active campaigns in it, let’s ratchet up to spend on social media, and now let’s do a causal impact model, get some… The proper kind of control variables, and see if we can detect the lift and see what we learn from that. And decide whether we should double down on that or not.” That’s been a very different way from organizations typically working.


28:46 MH: Hey, show’s going great, but we need to step aside for a multi-touch moment. Hey, Josh, have you seen this new analytics platform? What I love about it is it’s all about action. So many tools force you to go through and set up reporting, and then you’ve gotta go understand what those reports mean. This tool takes all the guess work out of it. It’s called Flamealytics. You heard of this?

29:11 Josh: Oh, I’ve heard about it, Michael, and I think it’s really a necessary tool. I mean, who really wants to look at dashboards anymore? More like dash bored.

29:21 MH: [chuckle] Exactly. See, this platform is all about alerts. Listen, it just works through Slack, email and text messaging. No more PDFs downloading to your desktop and opening up. It just tells you. You put one line of JavaScript on your site, you set up the users and their communication channel, and you are done. It monitors through AI all the metrics for anomalies, and whenever it detects one, it immediately notifies everybody.

29:53 Josh: It’s straight fire, Michael.

29:55 MH: Yeah, and actually, through an integration with Emojilytics, you can even just get them as flame emoji.

30:04 Josh: Yeah, so the system indicates the magnitude of any incoming anomaly with the simple flame emoji system. One flame means an anomaly was detected, two flames means a bigger anomaly was detected, three flames means all hell has broken loose and your resume has been automatically uploaded to ZipRecruiter.

30:23 MH: It’s amazing. One, two, three and go. No more guesswork. Now you know exactly what to do, through AI, using data, no more thinking about strategy or tactics, just go, go, go, go, go.

30:38 Josh: Hey, where there’s smoke, there’s fire, right?

30:40 MH: That’s awesome. Yes indeed, Josh. Alright, let’s get back to the show.

30:47 MK: Okay, so I have a weird tangent. Big surprise. One thing that I always come back to is customer IDs because that often… I feel like that’s your trump card if you’re trying to do, whether it’s customer journey analytics, or whatever the hell it is. Having a customer ID is always going to help you stitch. I suppose, as we keep talking about it getting harder and harder and harder on the cookie side, is it kind of becoming imperative that you should force users to log in? As I say it, I feel like you’re sacrificing potentially the best interests of the user, but I reach… Literally, this keeps me up at night.

31:36 MH: I would encourage any company who doesn’t really love their customers to go ahead and force them to log in every time they visit.

31:45 TW: I really think, actually, forcing them they get an implant of a chip, ’cause really, if you force them to log in, that’s really only getting the online.

31:51 MH: And I would go further, Moe, I would force them also to download our app, so that way, we can get their phone, too. So, they’re gonna log in on the website, and they have to download the app.

32:03 MK: There are lots of sites that force you to log in. This is…

32:07 MH: Yeah, and I’m not talking about media sites, where they wanna have you log in to get past the paywall or those kinds of things, but when you’re selling discrete things to people, no, we can’t go and make poor user experiences because we need this data.

32:25 TW: We wanna collect the data.

32:26 MK: But that implies that logging in is making the experience poorer, which I don’t necessarily agree.

32:31 MH: Forcing someone to, absolutely. ‘Cause where are you gonna force them to log in? Right at the beginning?

32:38 MK: But if you’re buying something, you’d need submit a whole bunch of information in order to buy something, and therefore…

32:44 TW: I have no idea if I wanna buy something, I just wanna see even what you sell. You’re telling me…

32:51 MH: Yeah, I’m gonna checkout as a guest ’cause I buy socks from people all over the internet. It’s just what I do.

32:58 TW: If you give them a sufficient excuse to login… I completely believe that if you’re like, “Hey, it pops up, you come to a site, and say, ‘Look, if you can answer these five questions and then let us save it, we will not show you a bunch of crap on our site we know you’re never interested in,’” or something like designing an experience where… But it’s always the trade-off. There is… Yeah…

33:24 MK: Like a site like Medium, to read a blog article, you have to log in. They’re not selling shit, it’s just like… That’s part of their user flow. I don’t know.

33:33 MH: Do you know how many fewer Medium articles I read now as a result of that?

33:37 TW: Yeah, I’ve switched… Pretty much now, I rely on whatever Pocket tells me I should go read, not… Or I hop into stealth mode. They meter it. That’s a media site, so that’s kind of back to… That is the different… That is the unique model of a media site.

33:53 PC: So the data part of me really wants to capture people’s login IDs because in terms of sitting the customers’ journeys together, that’s what I wanna know. But yeah, there is the other part of me that is like, “Oh, okay, user experience,” but if login IDs are not the solution, how have you seen ways to, I guess, sit together customer journeys or even be able to get that information?

34:19 MH: So I think there’s a lot of models.

34:21 MK: Bum, bum, bum.

34:23 MH: What? No, there’s models that… Look at the grocery stores, they’ve been doing this for 25 years where they give you a little card and every time you come in, they give you discounts on different products.

34:32 MK: That’s the same as a customer ID.

34:34 MH: Yes, that’s what I’m saying. So, they’re incentivizing you.

34:38 MK: But the card has an ID that they scan.

34:42 MH: Yeah, Moe. That’s what I’m saying. Instead of forcing you to do it, they give you an incentive to do it, which is different. And as long as you’re making it seamless, or mostly seamless, then I think you can create positive outcome. I think you can do something as simple as, “Hey, we sell these products on our website. If you login and create an account with us, we will give you 5% right off the top, off of your order today. If you don’t log in… “

35:10 TW: Which happens all the time.

35:11 MH: Yeah, that’s great. “If you don’t log in, don’t.”

35:14 TW: It’s, “Give us your emails and subscribe to our newsletter and… ” yeah. But it’s a transaction. There is… You are… There’s a cost. It is gonna be perceived there is a value on the customer’s… It’s like when we talk about the value of people being able to sell their data. As soon as you’re logging in, what is the fair exchange of information or fair exchange of value, which means you have to… You don’t just have to give them, say, “We’re gonna give you a better experience.” You have to sell them that there is enough value for them to give that up when they’re logging in.

35:47 MH: Yeah.

35:48 TW: Interestingly, with Mixtiles, I’ve been… You guys familiar with Mixtiles?

35:54 MK: No.

35:54 TW: So, they’re… It’s basically $11… A very, very streamlined, simple picture printing. It’s kind of designed for Instagram people to print their stuff out. But there is no log in, they won’t let you log in. It’s a little frustrating. Every time you go, you have to put in your email address and then it’s a very, very simple clean streamlined process, but it’s jarring that you can’t say, “I’ve half setup what I wanna buy, I wanna save it for later.” They don’t give me that ability. They’ve gone so far down the path of making it seamless… Although they are always prompting me for an email address when I start the process. They’re getting an ID. I understand why they’re doing it. There’s no password to remember, but they also have that… They’re saying, “You’re heading straight into the order process,” and that’s kind of an interesting model. I can’t really poke around much and see what they do, ’cause they’re saying “Order.”

36:49 MH: That’s not gonna work with anything that has 28 different types of sleeves on it, Tim.

36:54 TW: That’s true.

36:55 MK: Genius, Tim.

36:56 MH: Throwback to our Shoprunner episode.

36:57 TW: But a lot of this falls into… And we’ve had this… With the whole cross-device debate we had whatever many episodes ago, a lot of it comes down to the business model. If I’m a FMCG, CPG company, if I’m selling pet food, what’s the value I’m gonna offer to say, as soon as you arrive… And by the way, that’s still only starting when you arrive, that’s not necessarily linking back to pre-arrival stuff. That’s what you really want and that’s what’s getting harder and harder is the view through or the offline marketing.

37:31 MK: Okay. So Priscilla, recently, you explained to me why you’ve spent your whole career working on marketing analytics, why you decided to move to the light side, AKA product analytics. I’m curious because you seem to be so pumped about the challenges that this whole landscape poses. What prompted you, then, to make that shift?

37:54 PC: It’s a daily battle for me not to be involved with marketing. The marketing team is still like, “Come help us out, we know you build attribution models.” [laughter] And I’m like, “Mmm.” [chuckle] I think now that I’m four months into my job, I don’t think I can ever escape attribution because essentially, what it is and what I’ve mentioned is that it’s all about understanding customers’ experience with us, whether that be from marketing to product, which I do now. For now, I look more at the pain points that people have when they already are on our side and have come through marketing campaigns. But essentially, I think, I don’t know, there’s a bit of a blur in what I do now, whether it’s marketing or product because it’s all gearing towards the same thing about understanding what the pain points are for our customers and how do we go about solving that.

38:54 PC: And so, my thinking of attribution has actually switched from being it to solve a problem of where we should spend our money, to how we can better personalize that experience for our customers. And so I don’t think… That’s why I don’t think I can ever escape it, because it’s still what I care about in that sense, the day-to-day of campaign analysis and that type of thing, I have moved away from, and that spend optimization, but at the end of the day, I still work quite closely with that problem, is because it’s just potentially, the beginning part of the customer journey that I still need to know about.

39:35 MK: Yeah, it’s funny. We were having a conversation here the other day and it got really involved, where someone was like, “What’s growth? And what’s the difference between understanding growth and just doing analysis on any other part of the user journey?” And oh man, someone that dropped this bomb in the pod world, the marketers say it, and all of the analytics team. And next thing you know, there was debates flying around. And I really like the way you think about it, Pris, where it’s just coming back to the user, whether you’re doing marketing analytics, whether you wanna call it growth, whether you want to call it product analytics, it’s literally just about understanding where the user came from, whether they had a good experience with you, what their pain points are, how do you improve on all of those things. Yeah, I think that’s a nice way of framing it.

40:22 PC: Oh, I just don’t think we do it enough. And I think we were trying to do this, was to actually potentially attribute the importance of on-site events and how that [chuckle] played into a customer converting. Yeah, fair enough, marketing campaigns have an influence, but what about different tiles, once someone gets on site that has an influence as well, and how should we value that in comparison to the money that we spend off-site? Should we be investing more into that optimization on-site as well?

41:00 MH: Yeah, it turns out it’s a lot easier to run a marketing campaign for an iPhone than it is for a Zune, [chuckle] and that’s something you can’t escape. The product or the product experience, then has to map back into how you’re gonna do marketing and who gets credit for that. Is Apple just the most amazing advertiser marketers, or are they just making great products all these years, and that’s selling itself?

41:29 TW: So, two thoughts. One, there is the idea of attribution within the product or on-site which is something that… Like Adobe Analytics with Attribution IQ, is actually something that I think actually is kind of useful and interesting. And even Matt Gershoff, I think two years ago, at Superweek, he was kind of talking about attribution and it wasn’t at all about marketing attribution. It was how much attribution do we give for somebody interacting with this feature or that feature, which I think is super intriguing, and that’s also where it starts to blend with potentially AB testing, of saying, “Well, let’s remove that feature or not, but we can also… ” That data is a little more controllable and collectible. But I think just ’cause I’ve only lightly kind of passed by the product analytics world from the mindset of product, I assume that a big chunk, if you’re doing product analytics, you really do wanna know the person who’s unfamiliar with the brand, this is the first time they’ve arrived on the site, and product analytics around that audience versus somebody who is a long time loyal customer, or even somebody who came through paid search, branded paid search versus came through email. Don’t you inherently find yourself wanting to kinda group those out based on who they are, based on where they came from, because their experience with the product, their likely expected needs will be very different?

43:02 PC: Yeah, exactly, so I think that’s the use of attribution in the marketing pods, how that can play a impact on your personalization on site and on the app. And I think, again, that’s another exciting [chuckle] part of attribution that… For me, that’s why it’s not dead because you still have so much to do with that data, even if we don’t have the complete picture.

43:27 MH: Great thought. Okay, on that note, maybe attribution isn’t as dead as we started out thinking, but we do have to start to wrap up. One thing we love to do on the show is do a last call. Something that we can attribute to our learnings and something we’re excited about that’s going on in our lives right now. Priscilla, you’re our guest, do you have a last call that you’d like to share?

43:48 PC: I just want to share, because I’ve made the switch to product analytics, I’ve been listening to more podcasts on product, actually. I know, I’ve gone to the dark side. If you’ve heard of Pivot, it’s pretty much just a chat between Kara Swisher, Scott Galloway, and they’re really taking down Uber and WeWork at the moment, and all that kind of scenario. And it’s just really interesting how they talk about vanity metrics and things like that. And that’s really playing into how I’m looking at product analytics at the moment. And so I definitely write that.

44:26 MH: This is gonna be rough, if between the lag from recording till this comes out. If WeWork really turns the corner and all of a sudden, by the time this comes out…


44:34 MH: They’re just rocking and rolling, amazing turnaround, Adam Neumann… Okay Tim, since you saw fit to bust in, what’s your last call?

44:47 TW: I have done last calls from this site before, but I will do one from pudding.cool again. It’s called Millennials Kill Everything. It is just a really fun… They scraped something like 30,000 headlines from different media publications that had the word millennials in it and then they did some text parsing of verbs versus nouns. You can basically go in and pick a verb like kill, and then it gives you all of the words that millennials have killed, like millennials killed the doorbell industry. And then, you can mouse over it and it gives the actual headline: Indian Millennials Are Killing The Doorbell Industry By Texting ‘Here’, up until Millennials killed cigarettes. So it is a fun little site that is just comical, which, if you’re at all interested in pudding.cool and their story, I actually, a few months ago, listened to the episode of Data Stories that had Matt Daniels, who I think is one of the founders or one of the original guys of The Pudding. And it’s kind of interesting, as he talks through how they pick projects, how they design them and it’s a deep, deep, deep blending of journalism with data and how they kinda wrestle with what they’re gonna do and how they’re gonna visualize it. It’s always delightful analytics candy, check out Pudding.

46:09 MH: Holy crap, millennials are killing marmalade, apparently.


46:16 MH: Yeah.

46:16 MK: Alright, do you wanna go next?

46:19 MH: Well, I would be happy to prefer you, Moe, but since you asked, I will definitely go. I think this kind of conversation cannot really happen without a little shout out to a former guest on this show, Kevin Hillstrom, and you can find his Twitter account at @minethatdata. He has a lot of good points of view on attribution and product and merchandising. So I think a good read. And this is also just a quick little update. Many shows ago, my two co-hosts made fun of me for recommending a book called Finish that I did not finish by Jon Acuff.


47:02 MH: Just a quick update to our listeners, I am continuing to read that book, somewhat slowly.


47:11 MH: And it’s actually turning out to be very helpful, so there’s a lot of very good stuff in that book, so I still recommend it and I’ll keep you updated on my progress as I continue on.

47:22 TW: That’s fantastic!

47:24 MH: No, I’m really, really enjoying it still, so I just wanted to let everybody know. Okay, Moe, what about you? What’s your last, last call? The last, last call?

47:35 MK: Yeah, so I have a weird one today, really weird one. I came across this article called It’s Later Than You Think, it’s by JR Storment and it was getting shared around on LinkedIn, which one of my friends has this new obsession with LinkedIn articles, so all he does is send them to me, but it actually caused me to have the most beautiful weekend, where I just did the stuff that I should do, but basically, it’s about a founder whose son dies and just how it calls him to reflect on his career and his priorities. And yeah, it’s not a super long read, but I think it’s a good one.

48:10 MH: Very nice. Alright, well, you’ve probably been listening and you’ve been thinking to yourself, “Man, this attribution stuff sounds so easy.” Well, it is, especially if you go hang out on the Measure Slack and ask questions and talk to each other. And we’d love to hear from you. And please do reach out and head over to iTunes and rate us and review the podcast and give us five stars and tell Priscilla what an awesome job she did when you see her at the next conference or web analytics meet-up in Sydney. And we also wanna give a big shout out to our producer, Josh Crowhurst, without which most of the editing and things like that would not happen. In fact, Tim Wilson just posted the other day about how much he was enjoying having some downtime, and that never happens, but now when we have a producer, he’s started to be able to enjoy life once again.

49:07 MK: Yeah, but then he just filled his down time with a bunch of other crap, like…

49:11 MH: That’s because he…

49:12 MK: Not actual downtime.

49:13 MH: He got to do this amazing TTR analysis with his downtime, which is basically Tim having fun.

49:19 MK: Have fun.

49:20 MH: So anyways…

49:21 TW: I think we should also accredit Josh for actually tracking down all these great sponsors we’ve gotten, so that’s also… That’s been nice to keep the lights go on at the show.

49:29 MH: Oh yeah, no, that’s huge.


49:32 MH: Yeah, it’s so much, and we can attribute that to something. Anyways, stay tuned episode by episode as we bring you more and more of multi-touch moments. [chuckle] Okay. Priscilla, thank you so much for coming on the show. It’s been a delight, thank you so much.

49:51 PC: Oh good, I just checked Measure Slack and there’s no attribution channel.

49:56 MH: As it should be.

49:58 MK: Wow.

49:58 MH: Yeah, they’ve asked, we’ve always said no.

50:02 TW: You can attribute that to Lee Isensee, towing the hard line.

50:05 MH: Yeah, the attribution for that is we don’t make a lot of channels on the Measure Slack, that’s a big deal. So yeah, apparently, the conversation is taking place elsewhere. Alright, well, I know I speak for my two co-hosts, Tim and Moe, when I tell you, no matter if you can get a 360-degree view of the customer or not, just remember, keep analysing.


50:33 Announcer: Thanks for listening, and don’t forget to join the conversation on Facebook, Twitter or Measure Slack Group. We welcome your comments and questions, visit us on the web at analyticshour.io, facebook.com/analyticshour or @AnalyticsHour on Twitter.

50:53 Charles Barkley: So smart guys wanted to fit in. So they’ve made up a term called analytic. Analytics don’t work.

50:58 Tom Hammerschmidt: Analytics. Oh my God. What the fuck does that even mean?


51:09 MK: Help. That was literally the place to wrap, I’m like, “Attribution is not dead.”

[background conversation]

One Response

  1. […] highly recommend the pragmatic episode about multi-touch attribution on the Analytics Power Hour podcast. Priscilla Cheung, Moe Kiss, Tim Wilson, and Michael Helbling took a critical approach to whether […]

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Have an Idea for an Upcoming Episode?

Recent Episodes

#249: Three Humans and an AI at Marketing Analytics Summit

#249: Three Humans and an AI at Marketing Analytics Summit

https://media.blubrry.com/the_digital_analytics_power/traffic.libsyn.com/analyticshour/APH_-_Episode_249_-_Three_Humans_and_an_AI_at_Marketing_Analytics_Summit.mp3Podcast: Download | EmbedSubscribe: RSSTweetShareShareEmail0 Shares