Marketing data, fast and slow

Transcript

Edward:

All right, I'm joined by Evan Kaeding, Lead Solutions Engineer here at Supermetrics, and excited to get into today's episode on the concept of zoomed-in vs. zoomed-out data.
And really, at the core, this is about data access models, which might not sound super jazzy, but they're super critical to your marketing success. So Evan, to kick things off and get some context, what are data access models, and why do they matter?

Evan Kaeding:

Yeah, thanks for having me, Edward. Excited to be here. So when we think about data access models at Supermetrics, the way that we break it down is how are the people in your company accessing the data that they need to make their decisions?

Are they accessing it from a centralized repository that maybe you've built around a data team or BI team or maybe some kind of product as well where all of that data is stored in a centralized place, or are they accessing it in a decentralized way is each individual person going straight to the source that they need to access that data? Maybe pulling data together from multiple sources to put together some report to answer some kind of question or provide an update to a client or manager.

So these are fundamentally the two data access models we see at Supermetrics, centralized and decentralized.

And you can think of these as corollaries to one another when we talk about zoomed-in and zoomed-out data. With centralized data access models, there are a lot of benefits for organizations that do decide to centralize their marketing data. You can store large amounts of historical data, data that might not be available at all times from the different providers for all durations of history. You can also do seasonal analysis and trend analysis that spans several years, which can be challenging to do with decentralized data access due to data volumes that are too high.

You can also make that centralized data available to data scientists, technicians, and folks who want to build dashboards using that data. So there are a lot of benefits to having a centralized data strategy. At this point in the market, there's no question that that's what we see many companies gravitating toward when they think about how their marketers use data to make decisions.

But what we find at Supermetrics is that there is a separate and distinct need for what we're referring to as decentralized data access models. Where you still need, even if you have that all-encompassing centralized dashboard, even if you have that analytics team who's making sure that every data request is handled at the ready, you still have that distinct need for individual decision-makers to be able to get that data they need directly from the source.

And Supermetrics provides a suite of tools for accessing that data in a decentralized fashion for point-in-time decision-makers to get access to the raw data they might need. They may need it fairly quickly before a client meeting. They may want to move fast and iterate on a new channel or a new type of campaign that doesn't make sense to incorporate into the centralized data model where you might have layers of governance or have to get a lot of other parties involved.

Maybe you want to move and operate relatively quickly. These are all reasons you want to keep that decentralized data access model. So many companies are going in this arc, where prior to 2010, 2012, or so, if you're running digital campaigns, all you have is decentralized data access, right? Because there's no single source of truth. You don't have a BI team; you don't have a data warehouse, and everybody's going straight to the platforms for their reporting. Then coming in with a large wave of data consolidation in the industry where we see companies investing in large data warehouses, data lakes, maybe starting with finance or operations, marketing data starts to get sucked in, and then all of a sudden, we have centralized data access where we're pushing that marketing data into that data warehouse that becomes the source of truth for intelligence when we're talking about marketing data collection.

So what we see at Supermetrics, and we've, we've had this belief for a long time, but we've only just now put the data behind it, which is for companies who invest in maintaining both a centralized data repository and decentralized data access method, we see that the query volumes, in other words, the actual amount of requests that are made from these two different sources are really about 50/50, which is pretty exciting to see because that actually validates our narrative around centralized data access models being a necessary complement to decentralized data access models. For the folks that want governed data, long-term historical data that can be used for modeling, and historical trend analysis, centralized is the only way to go. Still, for those who want rapid experimentation, quick time to insights, and the ability to move quickly and succinctly in their data journey, having a strong decentralized data access model is an important part of the process as well.

Edward:

Yeah, absolutely. And I think coming from the marketing side of things, this is the fundamental challenge of do we have access to the right data. Can we make the right decisions quickly, and are we making decisions based on the right data and so forth? So yeah, I think really interesting to dig into this world of decentralized and centralized data.

Because I'm not sure if all marketing teams are thinking about the data in this way, it reminds me of the Thinking, Fast and Slow book by Daniel Kahneman, but applying similar thinking to marketing data and having the ability to act fast and move quickly and then also at the same time make maybe longer term decisions based on your data.

So if we take a moment to look at decentralized data first or zoom in, you spoke a bit about the benefits and how it works, but where does it make sense, and for what kind of companies does it make sense to focus on using decentralized zoomed in data?

Evan Kaeding:

Yeah. So what I find is that most companies, as I said, the default for most companies is the decentralized model. And I think ultimately that's a good thing, right? Because decentralized data, that's the way that we've been doing things for a long time, is if you need data from Facebook, you sign into Facebook, you grab that data, or you sign into something like Supermetrics. You grab that data the way that you need to use it. That centralized data repository is, of course, an important part of the stack. And from a marketing data maturity perspective, you need to ensure that your team is equipped to make the necessary investments to start storing that historical data and keeping that as a source of truth for your long-term historical needs. So when we talk about the distinct use cases for decentralized data, I usually tell the customers that you probably already have a decentralized data access model, even if you don't know it necessarily, but are you investing in the tools to make sure that that is as efficient as possible?

Because I can tell you what's not going to be efficient is if you're building out a new channel or a new set of channels and you need to collect a variety of different reports from various platforms, that's going to be pretty challenging. With a tool like Supermetrics, where you can access not only the data sources that you need but also streamline the connections that you might have to that data source, share them in a managed and governed way using something like shared connections from Supermetrics, you can actually invest in ensuring that your team can experiment, iterate rapidly with those decentralized data sources as a compliment to your existing centralized data strategy.

So I talked to customers, and it depends on where they're at in the life cycle. Some have gone all in on the centralized strategy, and we think that's great, but what's important is not to forget the separate and distinct need for individual marketers to build and create those reports. Maybe it's time; maybe it's the speed that they need.I think Supermetrics declared 2021 as the year of the marketing data warehouse. You might remember Edward, and so many customers are bought into that as well. So we see these as separate and distinct needs that ultimately warrant separate solutions at the end of the day.

Edward:
Yeah, absolutely. So you spoke there about decentralized being the sort of default starting point, and a lot of marketers are using a decentralized data access model, whether they know it or not, but then the centralized strategy, building out the marketing data warehouse, there's a bigger barrier to that than say just getting your data into a spreadsheet or a dashboard. What advice would you have for marketing teams in terms of how you can get started with that centralized zoomed-out marketing data access model?

Evan Kaeding:
Yeah. Yeah. To get started with centralized, there are a couple of different things and ways you can think about it. So I think a lot of the hesitancy toward getting into a centralized data model stems from these new tools and technologies for marketers. As a marketer, why do I need to worry about data warehouses, SQL, redundancy, and all of this kind of data storage in the cloud? Why is my spreadsheet necessarily sufficient, and how much will I have to learn to actually use this data? These are all valid questions, but not necessarily questions that are very difficult to answer. And so what we see at Supermetrics is for customers who do want to pursue a centralized data strategy, it's actually fairly easy to get started, and it's fairly easy, especially when you already have that complimentary decentralized solution where a lot of our customers today, a good example would be are looking to store their historical Google Analytics data from Universal Analytics.

And some are saying, well, I'm not going to be able to store all that data in a spreadsheet, a dashboard, or something like that. So I'm just going to go ahead and bite the bullet. Let's pop into BigQuery, Snowflake, Redshift, or Azure, wherever it happens to be, and let's just go down that route. And without any special training or expertise, we're guiding marketers into actually storing this data inside of this data warehouse, and it turns out to be significantly easier than most people had anticipated. So I have advice for those who may be intimidated by getting started with the data warehouse. I'd encourage you to hopefully think that it's not going to be super difficult getting started with the data warehouse five, 10 years ago; yeah, completely different story. But now, with AWS, Google Cloud, and Microsoft being able to spin these up with a credit card and very quickly, it's simple to get started.

And certainly, if you're working with a trusted vendor like Supermetrics, we can certainly guide you along the way to ensure that you're getting started. But the key really is to ensure that you're making that move relatively quickly because what we see as well, well, of course, Google Analytics, the sunset coming up is, of course, a very topical item, but that's actually not super unique in the sense that every marketing API, every marketing data source that we work with, they're deleting data from your accounts every single day. And for example, when we look at Facebook, Facebook only retains the last 37 months of data. A lot of people don't realize that if you're a large advertiser and have been advertising on Facebook for the last 10 years, that might contain some really important trends around seasonality for you or historical campaigns that have worked well, that data is being deleted as soon as it becomes three years old.

Same thing on other kinds of ad servers like Google Campaign Manager. That data is deleted as soon as it becomes two years old. Google Analytics 4, one of the big news sources everybody's working with now, only retains data in their service for 14 months. So, as soon as that data is a year and change old, that data's being deleted. And so customers who are observant of that and looking to build a long-term data strategy are jumping in headfirst and building out a centralized data strategy. Even if they know they don't have the data engineers, the data analysts to necessarily do everything possible with that data, they know that as long as the data's there, it's saved that they can invest in those things later or bring in specialist agencies who can make use of that data. But nobody's going to be saving your data for you unless you actually go and take those steps yourself.

Edward:

Yeah, first-party data for the win, and we speak about that a lot. I think that's a huge unlock for marketing teams to have that backward-looking historical data over 3, 5, 10 years, however long it is that you've been operating, to look back at trends. Then also, that makes you more confident in forecasting based on what has happened in the past. So I think there's a huge amount in terms of the benefit it brings to go about building this centralized data strategy.

And I think it was great to hear that it's become much easier to practically get started — the democratization of technology is making it easy. But I think the other flip side is the cultural and the stakeholder management side getting buy-in, et cetera. So I know, Evan, you've spoken a lot with marketing leaders over the last few years and worked closely with data teams to build this out for organizations. So if you're a marketing leader, how do you get that alignment and buy-in around this and build the business case toward centralization?

Evan Kaeding:

Yeah, if you're in the marketing leadership space, and you're looking at your data strategy. In that case, your first call needs to be to your data leadership and figuring out, hey, when you're VP of marketing, let's discuss with the VP of data, ironing out exactly who's responsible for what. So if I'm a VP of marketing, I'll be looking to my data team to build out. Hopefully, they've got an internal data warehouse where we can start storing our marketing data. Once they've got that, then I'm going to need to very succinctly articulate what it is that the business questions my team is trying to solve are, and basically give the data team, tell them not how to do their jobs, but what it is you're looking to do and how you want to interact with the data.

That way, they can take that brief and use the appropriate tools based on the stack they've chosen to start building out those dashboards, reports, and analytical insights that you might need to focus on. Again, taking a page from our book around what centralized data will be really good for, it will be really good for long-term analysis. It's going to be really good for detailed cross-channel analysis. It will be good for analysis involving data from multiple sources, potentially from internal sources as well. So anything that requires large amounts of data and a reasonable degree of sophistication, your data team's going to be very good at doing, and I'd encourage you to rely on them for that. Now, at the same time, it's important for you to recognize that there are many things that your data team's not going to be the best at.

As a VP of marketing, we're doing a lot of experimentation on different channels with different campaign types or ad types across different platforms. Does it make sense for every single type of experiment that you're running to make its way through the request pipeline for the data engineers? Data engineering and analytics teams operate on relatively quick cycles, but they can never be as quick as a decentralized access model. So you should be equipping your team with the tools they need to answer those questions that are relatively quick in nature, that are relatively ad hoc in nature, and may not require the degree of sophistication that might be required by an internal data team. That's how I look at it if I'm a marketing leader with a data team with a relatively robust discipline and practice around marketing data. Of course, only some marketing leaders have the luxury of having a fully built-out data team.

And so for that, what I'd say as well is you're going to end up leaning probably heavily on decentralized data access, and that's okay, but of course, as we mentioned earlier, it's going to be important to make sure that you're storing that historical data somewhere. So if someone on your team is maybe a little more technically inclined, have them responsible for the data collection. Make sure that data is at least being stored away for the next two to three years so that when you do reach that degree of organizational maturity where you're ready to start making use of that data, it's there, and it's readily available. Even if you don't have the resources internally today or the money to spend to hire an external resource to make heavy use of that stored centralized data, keeping that data and keeping the flow of that data on storing that's really setting you and your team up for success in the future, even if you don't have everything you need to do, even if you don't have everything that you need to make it actionable on day one.

Edward:

Yeah, for sure. And I think this is really interesting in terms of thinking about this from a marketing and data alignment perspective. We speak a lot in marketing about marketing, sales alignment, and marketing leaders working with sales leaders and making sure the customers flow nicely through the pipeline, that all the tools talk well, and Salesforce hygiene, et cetera. But marketing data alignment is equally important to make sure that the data the marketing is using is clean, reliable, it's there, and so forth. So really interesting to open up that a little bit. It could be something we can talk about more in a future episode, but you also spoke there about the right time to go in on centralization depending on where you are on your marketing data journey and your marketing analytics maturity curve. So are there some clues or signs when it makes sense to make that jump as a company?

Evan Kaeding:

Yeah. Yeah. So let's think about what that is. So if you're in a marketing leadership position and you have a couple of different members of your team who so far are using a set of decentralized access models for getting access to the marketing data, I think it arises from the need around the sophistication of the business and certain industries are going to require this sooner rather than later. So if you're in the ecommerce business, you're probably already behind if you haven't invested in this already. But at the same time, if you're in something that's maybe industrial or forestry or something like that, something that maybe doesn't require a huge degree of sophistication, you can get by with just decentralized access models for quite a long time going straight to the platform and getting those insights.

So when I look at what are the right triggers for making that investment, I look at the questions you're trying to answer, things that are reasonably answerable with decentralized models, doing things like historical trend analysis, doing things like blending data from multiple different sources, doing things like incorporating things like first-party data or other kinds of data that you might have internally. Those are all key triggers that are important for thinking about when investing in a centralized data strategy makes sense. A big part of the triggers we see here at Supermetrics is that customers have been happily using our Google Sheets or our Looker Studio products for years and years, and that's great, using the decentralized access model to streamline reports.

In some cases, though, companies grow, and that's a good thing. And as a result, the dollar spent and the sophistication of those marketing campaigns grow significantly. And so we ultimately find that for many customers, that trigger point comes where all of a sudden, the tools you're using to access that data are no longer sufficient for the sophistication and the volume of that campaign spend. At the same time, you can also look at that campaign spend and say, well, how much am I spending on these campaigns, and does this justify the investment into tools that I can use to further improve these campaigns? If you're spending a thousand dollars a month on a campaign, it doesn't make much sense to spend any money on tools to improve that. It's not going to be a very big marginal gain.

But if you're spending a 100K a month, a million a month, all of a sudden, well, now if I look at the investment of not only time but also tools for actually improving that spend, that's where you might start to think, well, okay, if I can invest in tools and get insights that help make my a 100,000 per month marketing spend more efficient, all of a sudden, you can start to see a positive ROI there. So we at Supermetrics are always in the camp that there's no sense in doing analytics just because that's what everybody else is doing. We always want to make sure there's a positive ROI associated with everything you're doing, including the cost of buying the tools, the cost of training the people, and making sure that you're getting some value out of that to ensure that you're driving that additional revenue that's coming from the marketing or ensuring that you're cutting down on campaign waste that might otherwise be spent better elsewhere.

Edward:

Yeah, absolutely. And you spoke about tools there. So from a data stack perspective, I could just dig into that a little bit more. What do you need to enable the zoomed-in and zoomed-out data models within your marketing org?

Evan Kaeding:

Yeah, from a tools perspective, I can speak about the Supermetrics line of products and the categories into which they fall. So maybe let's go into the decentralized route first. So for decentralized products, these are tools that easily facilitate individual contributors' access to the data they need to make decisions. So in the Supermetrics lined up, our core set of products here are Google Sheets, Looker Studio, and Excel. So marketers using these tools can quickly and easily streamline access to the data they need and pull it into the destinations where it makes the most sense for time-sensitive analysis. Then when we talk about the tools on the centralized side, that's where we start getting involved with moving that data into data warehouses. We launched first with BigQuery in 2019, which is Google Cloud's large-scale data warehouse, despite the fact that it is what they build is a petabyte-scale data warehouse, which sounds pretty fancy.

It's actually one of the more easy to use pieces of technology on the market today. We later added Snowflake as a destination, and then we now have various solutions in the Azure and AWS ecosystems. So these are the variety of tools we see our customers using across our suite of products. Of course, we have competitors in these spaces as well. So we have a variety of competitors that will help you get data from Facebook into Google Sheets or a Looker Studio. We have a variety of competitors that'll help you get data into a data warehouse, a data lake, or something like that. But the interesting thing here in the tools landscape is that Supermetrics seems to be really the only tool-making company, if you will, that has expertise in both, where we can actually help you accelerate the quick and real-time decision-making inside of these decentralized tools, but at the same time, we can simultaneously help you store your historical data inside of those centralized destinations that are going to be important for your BI analytics or marketing teams to make use of.

Edward

Yeah, absolutely. And we've covered a lot here and dug into the topic, but any final thoughts, or Evan, parting words of wisdom on how marketing leaders can get going with decentralized and centralized data access models?

Evan Kaeding:

So I think your comments around Daniel Kahneman's book around marketing data Fast and Slow, or Thinking, Fast, and Slow in that way. So the way I think about it and the way that I discuss with marketing leaders is, as a marketing leader, your job is to take a set of dollars, essentially, or euros, wherever you happen to be in the world, and allocate that efficiently to drive really two things, mentally have a responsibility to drive near term outcomes. Every marketing leader, whether they like it or not, is going to be held to those near-term outcomes. And that's an important part of the job while simultaneously balancing the need for long-term outcomes. And I think when you kind of view your job as a marketing leader in terms of allocating that budget from a media perspective, the thing that races around in my mind is you want to be looking at tools, technologies, vendors, and partners that can help you achieve both ends of that spectrum.

So when we talk about marketing data, and we talk about the tools and technologies that enable it, we like to say, and one thing that's been racing around in my mind is we like to say that marketing data is something that you can use to go fast or you can go far. When we talk about centralized data access models, we talk about going far. When we talk about decentralized data access models, we talk about going fast. And these are the two different ways you should think about your employees, your marketing processes, and the data they're using to make decisions.

Edward:

Awesome, and that's a great place to end this and wrap up. If you want to go fast and far, then there's the case for decentralized and centralized. So, Evan, this was awesome. Thanks so much for joining us today, and we got to get you back on the show in the near future.

Evan Kaeding:

Sounds good. Thanks, Edward. It's been a pleasure.