The right data for better decisions: How marketers can use on-demand and centralized data to make better decisions

Edward Ford:
Welcome back to another episode of the Marketing Analytics Show, and welcome back Evan, Lead solutions engineer here at Supermetrics. Great to have you back on the podcast. We've done a few episodes together and a few months back we did an episode called Marketing Data Fast and Slow. We had a lot of people listening to that and got a lot of good responses. So this is really a follow up to that and we're going to break it down a little further and go into it in a little more detail. But before we do that, I think a foundational question, why do we bother with marketing reporting?

Evan Kaeding:
Yeah, thanks I, Edward, and great to be back on the show and I'm really excited to dive into this with listeners again today because our thinking, of course, has evolved on this as Supermetrics, as has our understanding of our customers. One of the things that when I talk with customers, especially if I'm at a speaking event with a group of, let's call it 50 to a hundred Supermetrics customers in the room, I like to ask a very important question, which is why do we bother with marketing reporting in the first place? Why do we actually put this data into spreadsheets, into dashboards, into visualizations, and why is it sometimes necessary to put them into data lakes or data warehouses, pull them into those enterprise data storage facilities so that we can have backups of our marketing data and use it for historical analysis. If we're an agency, why are we building these dashboards for our clients?

Evan Kaeding:
Is it so that they can see that we've done the work that they paid us to do? Is it so that they can see what the top performing creatives are? But fundamentally, all of these things are true and they are things that we do, but the reason that we do them, and I would argue that the only reason that we should be doing them is to make better decisions when we make better decisions, whether we're a brand, whether we're an agency necessarily, we're making decisions for our business, we're making decisions for growth, we're making decisions for profitability. But of course, naturally we're also making decisions for the benefit of our customers when we make better decisions on the marketing side, we're able to better connect with our customers, we're able to give them a better buying experience, and we're able to deliver better customer satisfaction. At the end of the day when we make better decisions, our campaigns are more efficient, that frees up our ability to then invest in our services and getting those services and products to our customers. So I would argue realistically that the only reason we should be doing marketing reporting is to make better decisions.

Edward Ford:
Yeah, exactly. So I think nothing like a good fundamental existential question to kick things off. So okay, marketing reporting is all about making better decisions. That's the starting point, but what kind of decisions then do we marketers actually make because there's a lot of different ones.

Evan Kaeding:
Yeah, great question. So at Supermetrics, we obviously have to think about this quite a bit, right? Because we're in the business of supplying data to these marketers in order to make a variety of different decisions. So we actually have to look at what kinds of decisions are marketers making. Well, it turns out that marketers have a lot of different decisions to make. They have decisions to make about creative positioning, strategy, messaging, websites, conversion rates, analytical tools, deployments, pixels, cookies. All of these different things are things that they have to consider in the full marketing life cycle of their business, especially on the digital side. Now, with all of these different decisions, we need to be a little bit more specific at Supermetrics if we're designing products that are going to help support marketers in their decision-making efforts. And so we decided to build two different ways of classifying these decisions, and ultimately these kind of came through a series of interviews, a series of panel discussions, and really a set of research and observations that we've made over the course of our last 13 years in business as a company.

Evan Kaeding:
The first emergent pattern that we find is that fundamentally marketers make decisions that vary in frequency. What am I talking about here? Some decisions need to be made very quickly. Some decisions need to be made very slowly. Some of them are made on an hourly basis. Some of them are made on a quarterly, a monthly, or maybe even a yearly basis, for example. So when we take the frequency of decision making, fundamentally we recognize that there are some decisions that need to be made more often than others. Some examples might be if I am in the middle of peak season reporting and I really need to know which products should I have live and which one should I not, those are decisions that I'm going to need to make relatively quickly with data that's on demand relatively fast. So that's kind of on the one end of the spectrum, really making decisions very frequently with high velocity.

Evan Kaeding:
On the other end, if I'm talking about making strategic decisions where I need to say, for example, conduct an attribution exercise to figure out which channels are most profitable for me, I need to do budgeting, planning or forecasting for example. Those are decisions that I'm probably not going to make every day, and I probably want to take my time with those decisions as well to make sure that they're accurate, given the fact that there's going to be a lot of weight behind those decisions from a business. So fundamentally, the first way that we evaluate the decisions that marketers are making is based on how frequently they need to make them. The second way that we evaluate decisions is based on how much data is actually required. Now, I think we can probably both agree that the amount of data required for making a decision is going to be different based on what you're trying to decide.

Evan Kaeding:
Some decisions don't require very much data at all. Some decisions require an amount of data that fits very nicely into a spreadsheet, into a data visualization tool, or maybe into a file on your desktop, for example. Some decisions require even less, right? If you're in a digital marketing environment and you just need to know what was the total cost that I spent yesterday? Great, I can sign straight into a platform, I can get that number right away. I don't need to use any tools at all for that. That's both on the side of the spectrum where we don't see a lot of data required for that decision. However, if there's something bigger and more complex that I need to decide on, there's probably going to be a larger amount of data that's required associated with that. So for example, if I'm trying to do historical trend analysis or if I'm trying to understand how the last two to three years of my historical campaign performance has been, or identify different messages that might've resonated across audiences over a different course of time, that's something that's going to require a larger amount of data. So the other way that we analyze decisions is how much data is actually required to make this decision. So between these two different ways of measuring these decisions we have, what is the frequency of decisions that need to be made and what is the required amount of data for making those decisions?

Edward Ford:
Yeah, absolutely, and I think all marketers listening are probably nodding going to be like, oh yeah, that makes total sense, and this is exactly how I work, and for me just hearing you talk about this, I can relate to that. Just looking at we're early 2024 here, wrapping up last year, there were times when we were making quick decisions, fast decisions in terms of our final Q4 budget allocations on a kind of daily, almost hourly basis using relatively small amounts of data that we could self-serve. But at the same time, we were also doing our annual planning. We were doing our budget allocations, we were doing our forecasting, looking at historical data, historical trends, so that was slower decision making using a lot of different data. So you can really see when you just look at your own work as a marketer that you're making decisions based on frequency and you're making decisions based on small or large amounts of data. Now that we know about these different decisions that we need to make, and I guess we need different kinds of data as you said. So to make these different kinds of decisions, we need fast data, slow data. This is what we spoke about in our earlier episode, and we're also using terms on-demand data and centralized data to represent fast and slow data. So I'd love for you, Evan, to kind of break this down, what are the differences between these two types of data?

Evan Kaeding:
So I like to think about this in three different levels. There's kind of the applied psychology level at the very beginning. Then you've got the concepts that we have around marketing data, fast and slow, and then you have the actual practical tools, the application of this. So in the applied psychology side of things, we see in Kahneman's work thinking fast and slow. The way he describes these two systems, he describes them as system one and system two as system one is for fast decision making and system two is really for slower decision making. A good analogy that he uses is when, for example, you're driving a car in many cases, if you're driving in your own neighborhood, you're driving in your own car, you're driving in a relatively autonomous state because you know where the traffic signs are, where the potential hazards are, and probably exactly where you're going to be going.

Evan Kaeding:
So you're making decisions very quickly without a lot of excess thought put into them necessarily. You want to be able to do that in a lot of scenarios in your day-to-day life. Now in the system two style of decision-making, you're actually seeing that you're in an environment that's going to require a lot more careful consideration. So back to the driving example, if you're driving in, say for example, a foreign country or in a state that you're not familiar with, you're going to have a lot of different things that are unfamiliar to you, and it's going to require a different amount of reasoning and cognitive load in order for you to continue driving in a safe manner. It might be that there are hazards that you're not used to or unaware of. It might be that signs are in different languages or you just don't know where you're going.

Evan Kaeding:
You're going to be thinking a lot more slowly and carefully, a lot more analytically about how you approach that situation. So from the applied psychology side of things, we can actually see that this is a concept that does exist in humans, in how we form decisions and engage with our environments. Now bring that down to the conceptual level where we talk about marketing data fast and slow. There's an entire field of research in what's known as information retrieval that backs the fact that we as humans think about how we interact with data and how we interact with the data products associated with that, that governs how we actually make our decisions on a day-to-day basis. That's really kind of the genesis behind marketing data fast and slow and the concepts that we talk about. What are the decisions that I can make very quickly based on a small set of data and what are the decisions that I need to take a lot more time with that may or may not require larger amounts of data?

Evan Kaeding:
Then we move to the third layer, which is what are the practical applications of this? And that's why we have at Supermetrics the actual tools that go along with these mechanisms for decision-making. So on the fast side, we have our on-demand set of tools on-demand. Meeting marketers can sign in directly get access to their data sources and pull whatever data they need to make their own decisions in a relatively quick and fast format, whether they're trying to find what their best creatives are by conversion, whether they're trying to figure out how many conversions they got on a site yesterday or figure out some basic budgeting exercises for example. Then the other side of things, we have the centralized set of tools in our repertoire. The centralized set of tools are great for moving large quantities of data into a centralized and managed data warehouse or data lake that can be used for some of those larger and more strategic decisions as well. Building up that base and that catalog of knowledge in the business in the enterprise, that is going to be super important for making those long-term decisions like you mentioned, if you're doing forecasting, budget planning or looking at historical trends. So the three different layers that we have, both the applied psychology side of things, the conceptual side of marketing data, fast and slow, and the different tools that we have in the form of on-demand and centralized. That's how I like to think about this applied in practice.

Edward Ford:
Awesome. All right, good stuff, and let's dig more into this, and I want to start off with something you said the other day, which was really interesting on this topic. You said marketing often moves faster than data infrastructures can support, and I think this is where having those data access models in place is going to really enable marketers to move quickly. In particular on-demand data comes in here. So in terms of decision making, what sort of decisions is on-demand data suitable for, and what sort of decisions is centralized data suitable for?

Evan Kaeding:
When we look at it at Supermetrics and the way that we talk about our products and what they are best and able to support, we actually use a quadrant, if you will. So for listeners, go ahead. If you will build a quadrant in your head that has four different squares, and in the bottom left we have a square that we call the optimization quadrant, and optimization is simultaneously a combination of decisions that are made fairly slowly with a small amount of data. We move over to the bottom and we get to the tactical decisions that we need to make. These are decisions that need to be made relatively frequently, usually on a daily, weekly, or maybe even hourly basis, but don't require a huge amount of data to support them move up to the top right quadrant. That's where we have decisions that are made very frequently, weekly, daily, hourly, and do require a large amount of data, and then we move over to the top left where we have strategic decisions where we actually do need a large amount of historical data, but we don't need to make these decisions very often.

Evan Kaeding:
So classifying our decisions like this actually opens up a whole world where we can look at the decision that we need to make and select the best tool for that. I'll give examples of all four. So in the optimization quadrant in the lower left where we're talking about decisions that need to be made that are relatively infrequent and don't require that much data, that's where we see things like creative experimentation, things like AB testing, conversion rate optimization, for example. The backbone really of growth for any efficient brand is going to be doing those brand tests, doing those studies that, again, you're not going to be looking at those every single day, but it's going to be an important set of data for you to unlock your growth and to figure out where your message resonates and how best you can connect to your audience. Those are things that don't require a whole lot of data just volumetrically, but are still crucial for marketers that want to move quickly and relatively autonomously in their day-to-day operations.

Evan Kaeding:
Now, move over to the bottom right, we have our tactical decisions. These, again are the decisions that do need to be made on a relatively frequent basis, right? You probably do want to be checking your campaigns day over day to make sure you're not blowing through the budget. You probably want to be checking for reach and frequency, and you probably want to be doing these things fairly autonomously as well, both with optimization and with tactical decisions. You want to be able to do these relatively flexibly without having to necessarily go through the struggle of loading this data into a centralized data infrastructure, building governance, and building dashboards around it. You want to be able to do these things autonomously. These are the two kinds of decisions that are best supported by on-demand tools. Like you said, not all marketing operations move at the same speed as their data infrastructure.

Evan Kaeding:
In fact, I'd argue the vast majority of healthy marketing operations move significantly faster than the speed of their data infrastructure. That said, data infrastructure is going to be super important when a higher volume of data is required. That's where we move up to the top two quadrants. In the top right, we have operational decisions. Some examples of these are outlier detection. For example, if your conversion rate crashes overnight, that might be due to something that has happened on your website. Maybe there's a form that form filled that's not going through. Maybe there is a competitor that launched a product that is at a significantly lower price and your offering is no longer competitive. You need that large amount of data to determine whether or not that outlier is statistically significant based on your historical, your historical data. The other pieces that you see in operational sides are making sure you're not advertising a product that might be out of stocks.

Evan Kaeding:
You're combining your ad data with your inventory data, for example. So those are some of the examples we see on the operational side. Moving then back finally to the strategic side, if you're doing any kinds of attribution exercises, if you're doing any kind of statistical analysis, maybe an MMM for example, long-term trend analysis or budget pacing, sorry, not budget pacing, but budgeting or forecasting, that's where you're going to need a large amount of data, and you're probably not going to be doing those things on a day-to-day basis. So these two, both the operational and the strategic quadrants, these benefit from the centralized set of tools, and these are where we do see that data governance is important. Large amounts of data coming into the system are going to be important for empowering these decisions, and fundamentally, we're going to need the help of BI teams and data teams in order to make these decisions as well.

Edward Ford:
Love a good matrix to really think about these concepts in action to help understand. So let's move on to the benefits. What are the benefits of having both these models in your organization?

Evan Kaeding:
Yeah. When we built out this framework, it was really built on a series of customer truths that we learned about through conversations with our customers, panels, surveys, and a variety of different research mechanisms, and we found that one of the key consumer truths that marketers have is that they fundamentally need to move faster and more autonomously than the BI teams in many cases are able to support. And so the benefits to your organization for being able to enable your marketers to experiment more freely, to perform their day-to-day tasks independent of a BI team are numerous. You're going to be able to run more tests, you're going to be able to stay on top of your campaigns more often, and you're going to free up time to focus on market differentiating activities. For example, talking with customers, doing research if you're in the media buying space, having more strategic conversations with media vendors rather than trying to focus too much on getting that data out of systems, potentially cobbling together reports from multiple different systems and being able to actually understand at a deeper level what it is your customers are trying to accomplish using your products or your services.

Evan Kaeding:
On the data infrastructure side, on the centralized side, the benefits here are in addition to what we're able to achieve on the marketing side, the key benefits here are that you're able to, as an organization, put data in the hands of decision makers who have the ability to use that high volume of data to make strategically important decisions. So you're able to harness the power of what organizationally has been invested in order to deliver strategic data products that are differentiating in your market, so you're able to understand longer term insights, the impact of seasonality, for example, on your business, and be able to better accurately plan and forecast what's going to happen in your business over the course of the next couple of years.

Edward Ford:
Yeah, totally. And I think marketers might not realize this, but you need both to succeed. You need that fast on demand data in combination with the slow centralized data, and coming back to what we spoke about at the top in terms of why do we bother with marketing reporting? Well, we need different reports as well for these different kinds of decisions. And when you think about the reports you look at on a daily or weekly basis for those on demand optimization and tactical decisions, it's very different to the kind of reports that you're going to get for your centralized strategic operational decisions. So I think this is really at the core of success, particularly in today where marketing is so data led and there's so much data available to you to make those better decisions that'll help you going to make the right decisions, move faster, grow outpace your competition and so forth. So I think super, super important foundational stuff for all modern marketing teams. So let's get into the tech and get a layer deeper. So from a tech and tooling perspective, what does an on-demand data set up look like, and what does a centralized setup look like in practice?

Evan Kaeding:
Yeah, certainly, and Supermetrics is, I would like to mention one of the only, in fact, I'd argue the only company in the market that has tools that really span the breadth of these different use cases for marketing decision making. So on the one side we have on-demand data, where we have the set of tools at Supermetrics where you can pull data directly into a spreadsheet, a Google sheet, or an Excel document, or you can pull data directly into your data visualization tool of choice, be it Looker Studio or Power bi. These are some of the tools that we see being used most heavily by marketers for answering those on-demand questions that they have and for making those both tactical and optimization decisions on a day-to-day basis. Then looking over at the centralized side, that's where we have the set of tools that we provide that moves data into a data warehouse, a data lake, or a database, and those are the tools that typically are leveraged by data teams, by BI teams in order to build out their data.

Evan Kaeding:
Foundations probably incorporate that with other important business data and make sure that they're able to build and build dashboards and data discovery tools on top of those. Some examples that listeners might've heard of would be Google BigQuery, for example, Amazon Redshift, Microsoft Azure. There's a couple of different tools in that space where large scale data storage is available. Those are some of the tools that we have available from Supermetrics. The nice thing of course now is that with Supermetrics, we have all of these different features ensconced in what we now call the Supermetrics Marketing Intelligence Cloud, and that's an offering that is unique and differentiated in the market that can really serve both of these audiences simultaneously, the marketers that need data at their fingertips to make those day-to-day decisions and the data teams who need a larger and more robust stack in order to build out those data products that require a higher volume of data.

Edward Ford:
Yeah, absolutely. So yeah, as you said, Supermetrics really helps marketers work with both those data models setups. Evan, this has been awesome. We covered it a lot from the applied psychology layer to the conceptual layer to the tooling and practical layer. Anything else you would want to add in terms of how marketing teams can work with both on demand and centralized data?

Evan Kaeding:
I think what I certainly when I moved into management was that I oftentimes was making decisions without even realizing it necessarily. And when you move into management, I think is actually a very good time for you to reexamine the way that you make decisions because you realize how many decisions you have to make and how many of those you make autonomously as well, because then you can start to figure out, okay, can I delegate this decision to somebody else, or is this a decision that I need to make? I would encourage marketers to start to take the same approach for themselves and try to figure out what are the decisions that I'm making autonomously and what are the decisions that I'm making that require a large amount of data? And after some self-reflection, I think you'll probably see that you're making a lot of decisions subconsciously, and if you're able to identify those and identify the methods and practices that you're using to make those decisions, you're going to be able to figure out exactly what decisions you're trying to make and what's most important for you, and then you can use the right tool for the right job.

Evan Kaeding:
And once you're able to take the decisions that you're trying to make and match them with a set of tools that are going to be best suited for that, best suited for you, then you're going to end up in a situation where you have the autonomy, the flexibility, and hopefully the backing and the power that you need to make the decisions regardless of what frequency or what data volume is required.

Edward Ford:
Yeah, absolutely, and I think there where you said the right tool for the right job, that's such a key piece of this as well, and really enables marketers to make sure they have that available to them. So Evan, I think we could carry on for the whole day, but we'll wrap it up here. Maybe we'll save it for another one and have you back in the not so distant future. But thanks so much, Evan, again for joining us on the Marketing Intelligence Show. We'd

Evan Kaeding:
Love that. Thank you, Edward. It's been my pleasure.