Data top marketing teams need for better decisions | Arun Srinivasan (Clarisights) | Coffee Breaks

Foreign.

So we've got a bunch of people in. We'll have a bunch of people rolling in, but we've only got 30 minutes and hopefully, as you all know, we want to keep this super efficient. So welcome everyone to the third edition of the Marketing Measurement coffee break. I'm Michael, I'm one of the co founders of Recast. I am super excited to have an amazing guest with me here today. We've got Arun Srinivasan. He here with me. He is the co founder and CEO of Clarisites. They are a, a tool that helps marketers collect data and sort of analyze it, have a, an, an operations center for all of their marketing data. He has a ton of experience working with brands on how to handle data. What marketing data analytics should people be doing? Very excited to have a conversation with him today. We're going to jump in, have a good conversation. What we really want is for everybody, everyone on the call to be able to participate. So I've got a bunch of questions prepared for Arun, but as you have questions, thoughts, ideas, please chime in in the chat and we will just really make this hopefully a really fun conversation. So with that I am going to dive in. Well, gotta find my question page here and start asking questions. All right, so Arun, kick it off. What are the biggest mistakes that you see marketing teams make generally when it comes to dealing with their data?

That's one of the things that we've been lucky to have an inside of some of the most skilled marketing teams and had the luck of actually building product with them. Now I think this is somewhat unique to very large teams where you tend to get to a very siloed view of channels. Like if you have a team that's only looking at paid social, they're only looking at paid social on a daily basis. They're not looking at what's working on other channels. They don't have an understanding of, you know, how are conversion rates holding up on other products and that you end up in a local maxima. But you're just trying to optimize a couple of channels. You don't quite see the complementary data sets. And I see the other side to it, which is basically that each channel has its own nuances. The way, you know, the kind of optimization levers are different. But there is also so much of complementary data that you could be using. And the way we've seen this is when companies look at market teams, when we see the reporting, it's usually extremely siloed with very little visibility to other channels. Etc. That I think would be one of the biggest.

That's a really interesting insight and definitely I think a thing that we see across all of data analytics where, you know, individual teams or sort of laser focused on their KPI or OKR or whatever and optimizing that can come at the expense of other things. So I guess like what's the solve there? Is it to have a centralized data team or marketing data analytics team that's pulling everything together or like what are best practices for actually, you know, dealing with that problem?

I think, you know, might have seen, we've actually seen some customers implement a pod. Structure is basically where you have teams working across different channels, responsible for certain targets. And then you have weekly review cadence meetings where you're looking at cross channel data, you're looking at the revenue target and then you're distributing it across different sources and having that collaboration on a weekly basis where you're not just relying on channel data or BI dashboards and you're actually doing review meetings every week. Understanding what's working cross channels, that I think is a very important structure, you know, to drive that cross pollination and best insights.

That makes a lot of sense. I think when I've seen these meetings. One of the things, one of the problems that often comes up is that there's a ton of different data sources, right? It's like if you look at the Facebook in platform, things look great. If you look at Last Touch, it doesn't look so good. If you look at First Touch, it looks great. Like how do you think about best practices there for marketing teams to pull that all together? Is it just like one giant spreadsheet that people are looking at together and trying to make sense of or what's best practice there?

But I think that's something that you're looking at beyond just the marketing team. You're also looking at the data team. The way I see this is the best marketing team. And data teams is where data teams is building this scaffolding. And when I talk about scaffolding, I'm talking about measurement, tracking, etc. Data teams are focused on building that level of structure so that you don't have this exact problem that you mentioned where each channel is just optimizing, adjust to their own, whatever numbers are making sense now. And that I think is very important. That should not be done by the marketing teams on their own, figuring out the efficacy of their channels, et cetera. That should be done like as a famous separation of the church and the state. You want to Be able to kind of separate that out with the data team is building that scaffolding there where I think, but I do think there is value the upper funnel metrics. Right now you're looking at upper funnel metrics like click through rate view through rate, bounce rate, etc, and the best analogy I can share for this is like when you're building okrs, you are optimizing for both input metrics and output metrics. The output metrics, the area of truth, can be your revenue metric, which is centrally governed, but you still need to look at upper funnel data, look at, you know what's working, when is how is the conversion rate going across channel and do you want to take decisions based on cost per view, bounce rates, et cetera. So having access to that level of data is important. It has its place.

That all makes a lot of sense. I think we just got a great question in the chat, like, where do things like incrementality tests fit into this view? Right. So we've got, you know, the stuff that we're pulling from the platforms, we've got a last touch view and then we can see conversion rate on the site. We have an amazing dashboard. How do we think about these other types of measurement methods and where does that fit in as well?

And okay, so this is actually comes back to the evolution of how complicated marketing has been over the last few years. Earlier we used to have the structure where you just decided on whatever MTA model that you hold on and you don't change attribution, you never change attribution, you hold on to attribution for years on end and that doesn't change. I think with iOS 14, with cookie loss, with all these other problems is kind of drives what you guys are building with MMM and incrementality test and that means that you need to have that feedback loops much, much faster to see how each of these channels are working. Do you have the data to be able to test what you're seeing on tunnels? Benchmark that with what you're seeing on MM model, the incrementality test. And this is where there is an issue with the fact that you cannot easily run incrementality test across channels because you have to run geolift studies which are specific to each channel. Right. So you need to be able to constantly keep testing this. And that's where we think the data team's work is overloaded now because this is not a set of challenges that you had a year ago, but now suddenly you have something that you earlier you just used to Define it once and hold on to it to now continuously testing your incrementality models and other sources, which I think is absolutely. It's. I would say it's more the last one year phenomenon, which is not something that is a problem earlier.

Yeah, I think it's like the way that I see it is there's both a problem for like the data teams in terms of like having to collect all of this stuff and organize it in a coherent, insane way, but then also the marketing teams to try to like interpret it all. Because you know, when I've been doing this it's like you see conversion rate going overall, but like on the Facebook platform it's going down. And we have an incrementality test from three months ago and then it's like I have to try to look at all of these things and balance them in my mind in order to try to pull the whole story together. And that's just like it's really hard to do. It's hard for even the most technical analytical people in the world to do, much less people who are marketers first and you know, analysts second. I guess. You know, you talked a little bit about sort of separation of church and state between the analytics team and the marketing team. How do you think about that? Should that, is that like a sort of an antagonistic relationship where the, the, you know, I've seen it before, I guess where it's sort of an antagonistic relationship where the analysts often coming from, like the finance team are really trying to grill the marketers about where every dollar comes from. How like what's the ideal relationship between an analytics team? Should it be like analysts holding marketers to account? Should it be a partnership? What do you think actually works well there?

I think it doesn't need to be antagonistic, but you're right, it often becomes that and I think it's largely because of workloads. Have ever increasing workloads and then you have no way of solving it. That usually leads to tension between both teams now and the way I see this is the number of things that have been piled on data teams is where you're looking at, like you said, you're building pipelines, you're maintaining pipelines, you're trying to establish a source of dashboarding and reporting for the entire plot of different functions. Plus you also have this never ending source of ad hoc requests that keep coming from marketing teams. Right. And that's where you have this tension and that's coming from it. I don't really necessarily see that Complaining marketers who are demanding for more data is necessarily a bad sign because that's usually a good sign that they're trying to be curious and you want the best. Marketers are curious and they are constantly trying and testing hypothesis. If, I mean, not to make a plug about us, but you know, way we think of this problem is if you separate reporting from dashboarding and data modeling, that's. And you let marketing teams run their own reporting while you hold this category. Scaffolding around what should be the revenue for company, how are we looking at this? Etc, the data governance. Being able to separate these two is usually the best sign because when you have a data team that is responsible for everything, usually this is what the parts that you send kind of come back. Now, I also think that this is somewhat specific to marketing data because I would say what is very unique about marketing data, which is probably not the case with everything else, other data like financial data, operational data, etc. Is the very short feedback loops that you have. So when you have, when you're testing something, you have a budget that is decided at the beginning of the month, you test it across multiple channels. There is a difference between what you're actually getting from channels and what the budget was initially assigned. Right. So you have to constantly reallocate based on learning, stress, hypothesis, etc. This is not something that fits very well if you have a structure where you constantly have to go to the data team to get answers. And that I think is where it leads to this tension.

Really interesting insights. We've got a bunch of interesting questions on the chat. I want to turn to one which is sort of along this idea of like, you know, the evolution of data maturity for a marketing team. How do you, like, what do you see as the right way for a marketing team to go from like just looking at platform reporting to adding on MTA to adding on lift test, to adding on mmm. Like, how do you think about the growth of a marketing team in terms of maturity and when is the right time to add on these different data points and measurement methods?

That's a great one because I think when you're looking at the number of variables that you're juggling right now, I think that's largely dependent on, let's say you're a single product company and you're running on one channel, you don't really need to do too much to figure out test efficacy, etc. Because there is no problem of attribution at this stage. It's a single product. There are not Too many touch points. You can kind of see it the minute you have multiple product lines, multiple channels, and then if your conversion window is somewhat long, it requires multiple touch points. That's the point in which you start thinking about all these aspects that you shared, right? So we almost think of an evolution as companies starting with a spreadsheet pulling in some data just to have daily reporting. And then as you start getting more and more complex, you start adding multiple channels, multiple product categories, etc. That's when you move to a data warehouse structure, a centralized data team building a lot of stuff around it. And at, at a certain stage, this structure starts limiting you because you're constantly growing, your targets keep changing quarter on quarter. Reporting is not, or dashboarding is not a one time build solution. It's a constant maintenance problem. And that's where the spreadsheets start coming back, where teams start going back and saying the lead. We don't want to wait for two weeks to get answers and they'll get back to this. So depending on the stage, usually you would see spreadsheets always in the mix. And we don't necessarily think that spreadsheets is a sign of marketing teams not being data savvy or data efficient. We usually think it's actually the opposite.

Makes a lot of sense. I'm curious, most companies at least that have been started in the last like 10 years will have a data warehouse that has their data. That's at least the marketing data that's easy to collect, right? So from sources like the Facebook API or the Snapchat API or whatever, like all of these digital channels that make it really easy to collect data. However, there's a whole ton of other marketing data that's a lot harder to deal with, right? You have like a local radio station in Alabama that, you know, maybe definitely doesn't have an API, maybe has an analyst, you have your agency, right? That might be at different levels of maturity that you need to collect data from. How do you like, advise marketing teams to like deal with this problem where we have, you know, some sources that's very easy to get data and then a bunch of sources that is a real pain.

You know, even, even I actually think that even for the main channels, right, Like Google, Meta, et cetera, the number of different metrics and dimensions that you can slice and dice data with. Like suddenly you have pmax and then the reporting structure breaks, right? It changes. You look at something very different. And so there is a lot of changes that happen even with the most stable or established APIs. Then you have a problem where the Google Analytics API will report in a search. You don't see that on the dashboard. So there is a lot of complexity that comes even with these main channels and so there is a lot of maintenance even for that. The way we think about the second part of the problem that you said is more of a data completeness problem where you have sources that are reporting on. You're not going to get a ton of different dimensions or a ton of different metrics. You're going to get some, but it's, it's flaky. You might get it on an email attachment once a week, you might get it daily and you need to be able to figure out when are these ads going live, did that actually increase conversions or traffic during that time in that region? There are ways in which you can get around this problem by building some structures and hypothesis around it. But yeah, I just wanted to correct that part where even for the main channels like Google Meta, et cetera, the minute you have product free darts coming in, it's a completely new reporting challenge than what it used to be earlier, right? There is constant evolution there.

Makes a lot of sense I guess. What are the. So like if I'm you know, if we're talking to marketers who are sort of early in their journey, like what are the foundations that they should be laying from a data perspective today in advance of like, you know, imagine that they're a spread or they're really just looking at reporting in platform like what should they be thinking about data that they collect today in their data warehouse to use in the future or how should they think about laying that good foundation in order to be able to do this more advanced reporting later on.

Michael, I'm actually a big fan of not over complicating things unless until you need it. Because usually if you end up into a very complicated. When you're looking at seed series A startups where you are trying to figure out get to product market fit, you're trying to figure out what products are working, etc. You probably don't want to be spending too much of time building out data warehousing, etc, it might not necessarily be something what you would want to just look at is depending on your channel mix. Are you being clear about the metrics that you're using to measure each channel? And I think the complexity is something that you want to add depending on the nature of the business. If you're an E commerce company with multiple different product categories, there's a seasonality in your business which changes purchase behavior, et cetera. That is when you want to start thinking about building a data warehousing so that you can get historical data, you can get seasonal data, et cetera. But if you don't have that level of complexity when you're starting out, you probably don't. Shouldn't be spending too much time on this. That's, that's just my view on this.

Yeah, I think the, the. I think that's right. Like don't overcomplicate things is also one of my mantras. The places where I do advise people to start to think about it earlier is on if they're doing marketing that is difficult to collect like retroactively, then you want to start collecting that today. So things like influencer programs, right. Like you want to, if you're managing your influencer program via like Instagram dms, you should probably start today by like figuring out a system to log that information, even if it's just in a Google sheet so that you will have it in the future in order to reporting and analysis as opposed to like getting a year down the road and then being like, oh man, someone has to go through all of the Instagram DMs to try to figure out what we did with all of these different influencers. That's a real nightmare.

That that is true. I think, Michael, the way I think about it is if you know your product is working and it's just scaling marketing, that's when you want to really absolutely nail down the metrics. But if you're iterating on both the product, is this product working? Is are people buying this? And you're also trying to figure out marketing, you probably want to not think too much about this is how I see it. But the minute you get to a product that's converting well, you're hitting metrics and it's just about scaling across channels and documenting your learnings quickly, then what you said absolutely makes a lot of sense.

So we just got an interesting question. How do you think, like I think a lot of the reporting that we're talking about is pretty easy for like fast moving consumer goods sold online. How do you like, what's the right data for these more complex products that have way more marketing touch points, a lot longer time to purchase? How do you think about what's the right data for those businesses to look at to be able to actually make decisions?

Yeah, I think this is where earlier companies used to focus on the multitouch attribution model. Right. That you're looking at the number of days in which you have conversion, the number of touchpoint that you have and you start adding weightage to some of those aspects right now that's obviously a lot harder to do in the current market because you don't have the. And this is exactly I think the reason where you need mta, you know, lift studies, incrementality tests, et cetera. I. And yeah, investing in the tools that you think is when you start hitting this problem where you are seeing multiple touch points before you see a conversion is where I think companies should be spending a lot more that.

So ClareSights works with a lot of these, like bigger, more complex brands that I think probably do a mix of, you know, MTA mmm and experience and experiments from, like, from your perspective and what you've seen, like how are different marketing teams bringing that together? Where have you seen success? Where have you seen, you know, failure? I guess across the different companies that you've. That you've had a chance to work with.

I would say that MM models which are a lot more nimble and then allows for this really short feedback loop that we talked about where when you start, the way I see this is in large companies, budgets are assigned by product category, by channel, mix, etc at the beginning of the quarter and based on pacing and actuals and numbers, you constantly go and reallocate it depending on what's working, what's not. And that's where I think the fast feedback loop of your MMM model feeding into the MM model, bringing that data back into your system so that it your reporting has that layer constantly. Rather than having an MM model which is updated once a quarter is really the difference. And this is where I think that not enough companies are investing in the nimbleness of this setup. Right. Michael, I guess you've been experienced, you're working on the space. Would you agree with that?

Yeah, no, I think that's right. I mean, I think obviously Recast believes that that's our whole product philosophy. But I think for, for what we've seen, the companies that have the most success are the ones that really take like a test and learn mindset to everything that they're doing. And they're always trying to think about, you know, what is this measurement method, where does it work and where does it go wrong and then how can we prove it in different ways, Right? And so like we can prove the effectiveness of Facebook by doing a conversion lift study with Facebook. We can prove it by doing a GEO holdout test. We can prove it by doubling spin next month and seeing what happens. And that sort of approach, I think, to like marketing measures, measurement and, you know, dynamic effectiveness optimization, to use a bunch of buzzwords, that is what I think is that's what the best teams are doing is they're trying to come at it from a bunch of different angles. They don't, they want to use triangulation, they don't want to take any source as, you know, truth, but really figure out what are all of the different ways we can come at this problem and then how can we pull that information together into one centralized view or way of thinking about the world that avoids analysis paralysis. Right. Because analysis paralysis is sort of the other problem where it's like, oh, we can't make any decision until we run a lift test. But that also isn't helpful. And so I think the good teams are trying to find that balance between getting as much signal as you can from different measurement methodologies while also being willing to take bets and move quickly in order to be able to dynamically optimize what they're doing.

Absolutely.

Another really good question here on cookies going away. Right? Tracking is getting harder. Cookies are going away. How are marketing teams responding to that? And like in this world where, you know, we have less faith, I guess maybe in the numbers that we're getting back, what do you think the best marketing teams are doing today?

I think it goes back to everything that we talked about earlier, Michael, by running these tests and experiments. And I think just go back to leaning to what you said. Right. I was recently talking to one of the early product managers of Optimize and about a B testing. Right. They were the first to come up with this. And there's something that they shared which was interesting. He said the best teams that look at who ran a B testing really well were ones who had an hypothesis and wanted to test it. And you're not just testing for the sake of testing, but you come up with the hypothesis and you test it with the platform and that's where they had the most success. And I think that a lot of that is true for marketing, performance marketing, etc, that you should, should be building hypothesis, you should be thinking about, you know, And I think my biggest attraction to performance marketing, digital marketing, which kind of goes back to a long time, is that it's really cheap, really fast for you to test out your hypothesis. Right. You can run a campaign, you could have your opinions on what works. You don't need to be stuck in the analysis paralysis mode. You can test it and you can test it for fairly cheap as long as you get statistically significant data to take a call. And yeah, the best marketers are curious ones. And then if you're curious, you'd be constantly testing this.

Great. I love it. Sorry, I'm picking my next question here. Okay. What are the most critical pieces of data for marketing teams to be looking at and focusing on? How would you, or maybe like talk me through. What would you think of, like, if you're a marketing team going into this weekly meeting that you were describing, like, where do you start the meeting and sort of what are the data points that you go through as you're evaluating what's working and what's not?

So, but I would start with the targets, right? So what are the targets that have been assigned? What is the, you know, assigned per channel? And when you start with that, you're assuming a certain conversion rate, you're assuming a certain number of clicks, visits, revenue, etc. And then you're, you have to report on these numbers constantly to see what's working. Which of this metric is working better in channel one versus channel two? Are you seeing a pattern in, in this behavior? Are you seeing certain product categories converting better than others? And the other unfortunate part about digital and performance marketing is the opportunity cost is very, very high. So hold off. If your decision making is slow, if you take too long to analyze data, you're going to lose that traffic. You're not going to get the traffic back. So you have that window of opportunity and a fast, really fast reaction cycle. So being able to pick up on these nuances and learnings quickly is super important. And the teams that have this weekly cadence where you're talking to each other, talking to different themes, you're looking at the kind of creatives that are working for a channel. You're also analyzing to see if there is any offers or promotions that are working better than not. Which of this can actually be replicated? Which of these are just one once in a time event? And what can you use to bake off those decisions? These are all things that good marketers look at constantly.

I love it. That's a great answer. All right, we've only got two minutes left, so I'm going to switch to wrap up mode right now. So first of all, Arun, thank you so much. This was a wonderful conversation. Arun is the CEO of Clarasites. You can go check out their website. Really, really cool marketing data product. Also, please Follow Arun on LinkedIn if you want to sort of stay in touch with him. And get his thoughts on all things marketing data related. And then additionally there were a bunch of questions in the chat about MMM and Lift test and mta. For everyone who's asking those questions, our next coffee chat is going to be with the CEO of House IO. They're a lift test incrementality testing platform. We're going to get really into the weeds on MMM lift test, how those things work together. So please sign up for that next coffee break and yeah thank you everyone so much for for making the time to hang out with us today. Really really appreciate the time. We will send a follow up email with some links and some more over the next couple of weeks and months. I will be publishing on my LinkedIn answers to all of the questions that we didn't get to today. So thank you all so much. Thank you Arun. Really appreciate the time. Really fun conversation. Have a great rest of your day everyone and have a great weekend.