Putting marketing data into action through reverse-ETL with Google, Supermetrics, and Hightouch
In this episode, you’ll hear from Henrik Warfvinge, the Technical Leader of Artificial Intelligence at Google, Kashish Gupta the Co-CEO & Co-Founder at Hightouch, and Evan Kaeding, Lead Sales Engineer at Supermetrics. They’ll discuss how to put marketing data into action through reverse-ETL.
You'll learn
- The limitations of the current data stack that block data teams from making the most of their data
- Important factors to consider when building an operational marketing data warehouse
- How reverse-ETL powers data activation
- What companies can achieve with data activation
Transcript
Edward:
All right. So it looks like we got a lot of people dialing in and joining, so welcome everyone as you get logged in. We’ll just wait a minute or two to get people a chance to get logged in and then we’ll kick off in a minute or so.
All right. So I think the participant number is kind of slowing down. So I think most people have dialed in. So what I think we can do is get started and then I think a few more people will join as we get into it. So yeah, I think we’re good to go. And I just want to start by welcoming everyone to today’s event, Putting Marketing Data Into Action Through Reverse ETL with Supermetrics, Google, and Hightouch. We’re all super excited to have you join us as we dig into the modern marketing data stack, and I’m delighted to say we have three incredible guests who will talk us through it. And then just before we get into it, a few brief housekeeping items. We’ll have about 40 to 45 minutes for the discussion, and you can submit any questions you have in the Q and A. So you’ll find that in the bottom navigation. You can also go in and read and upvote those questions, and then we’ll have about 10, 15 minutes or so at the end to go through some of the most popular questions and we’ll finish with a short poll. So I think then to kick things off, I was thinking we could do a quick round of short and sweet introductions from our panel. So if you could briefly introduce yourself and let’s go in data stack order. So we’ll start with Evan of Supermetrics first, then Henrik of Google, and finally Kashish of Hightouch.
Evan:
All right. Thanks Edward. Hey everyone. My name’s Evan. I’m a Lead Sales Engineer here at Supermetrics. Been at the company for about a year-and-a-half, but have been a user of the tool for much longer than that. A lot of experience in the marketing, data engineering, data pipeline space. So happy to join here and share my thoughts.
Henrik:
Hi everyone. My name is Henrik Warfvinge. I’m calling in from Stockholm, Sweden, the Google office there. I work as a Cloud Advisor and a Technical Leader for machine learning and AI. So my role is basically to help organizations and partners to solve business problems through Cloud, but mainly data and of course, machine learning, and helping them out on the GSP platform as well.
I’m super excited to be here, to talk about this very interesting topic that is very sort of hot right now in the market. So I have a lot of conversations around this and excited to talk to this panel here today about it. Thank you.
Kashish:
Hey everyone. I’m Kashish Gupta, one of the co-founders and co-CEOs of Hightouch. We pioneered the Reverse ETL space back in 2020, and super excited to share some of our customer use cases that involve using customer data in your marketing stack, in the reverse directions, cutting your warehouse data back into the marketing stack in order to run campaigns. So super excited to share that context with everyone and excited to be part of the webinar.
Edward:
Yeah. Awesome. It’s great to have such a stellar panel today, and my name’s Edward from the Supermetrics marketing team. And I think our starting point is essentially the current situation we find ourselves in with marketing data, and that is essentially, there’s so much marketing data available. And really, all we want to do is turn that data into insights, whether it’s understanding what and why something has happened, or forecasting what may happen in future, or identifying what we should be doing. But there are limitations with the current data stack that really block marketers and marketing teams from making use of all their data.
Evan, I know this is something you’ve seen a lot in your role. So what are some of the limitations of the current marketing data stack?
Evan:
Yeah. Good question, Edward. I think the things that I see now in 2022 lead me to believe that the limitations are largely human limitations and largely skill limitations. I think over the past really decade, we’ve had lots of transformational technologies hit the market, BigQuery of course, with the ability to pretty much instantly access petabyte scale, data warehousing and storage has massively changed the game. Kashish here with Hightouch really has changed the name of the game in terms of helping turn that data into actionable data points, customer interactions. And then of course, here at Supermetrics we’ve made it super easy to get that data into the data warehouses.
And so, in 2022 here, I think I’m kind of sitting here thinking, well, the technological limitations have really fallen, I’d say quite precipitously. I think the question ultimately boils down to use cases and skill sets. I think that savvy marketers realized this a while ago. They realized that they needed to add data to their skillset. The ability to work with data, understand data, understand complex systems that produce and also store, and also visualize this data. And those are the marketers that are succeeding in our data driven environment today in 2022.
Point being the tools and technology is there. It’s really about understanding the use cases, and I’m excited to kind of explore what the use cases are on the data activation side here today. Of course, we at Supermetrics have been pretty prepared, certainly for the last, at this point almost, decade in terms of helping turn this data into actionable insights on the marketing side, understanding campaign performance, understanding ROI, conversion funnels, those kinds of things. That is largely there and largely possible, and has been, I’d say to a large extent, for the last couple of years, but today what’s changed is really, I think the technological boundaries have fallen to the point where if you can imagine it, you can pretty much do it at this point.
Edward:
Yeah. Anything to add on that Henrik, Kashish? Are you seeing similar things?
Kashish:
One thing I’d echo with Evan’s statement is really how much the boundaries to that technology have decreased. So as Evan mentioned, the warehouse is not only much faster, but it’s also more accessible. So companies that are much smaller and earlier stage can have access to Cloud data warehouse. You don’t need a whole data team running it, thanks to companies like BigQuery making it really easily and accessible.
And then second thing is the skills required to actually work with data. So it used to be that you’d have to be like a hardcore SQL user in order to use data. And now data is kind of more easily accessible to users that understand data in a spreadsheet way. And again, that’s also because of companies like Supermetrics and Bitcoin. So the fact that there’s a lower barrier to entry when accessing data and using data makes it a lot easier.
And even at Hightouch, we have a product that escapes the need for SQL, where you can point and click generate user subsets or user segments in a natural language way, which makes it possible for someone that wants to say, “Show me all users that viewed a product in the last seven days, and then do not check out that product within 24 hours after viewing? And then let me run email campaigns and ad campaigns on all those users.”
So defining a query like that would actually be pretty tough in SQL. It would even take me 30 minutes to write, but you can just point and click… do that in Hightouch pretty quickly. So removing the kind of like skills gap there is really important too, in addition to technology.
Henrik:
Yeah. And again, echoing this and that’s something we put a lot of focus into from Google Cloud’s perspective as well, to make things easy to sort of this whole area of democratization and even going so far as being able to create sort of machine learning models in a simple way, and make use of that to do sort of predictive modeling in an easy way. This is key areas to put focus on and that’s something we try to do as well.
Edward:
Yeah. And one thing I’d love to move on to now, Henrik on the back of this, why should companies and marketing teams start using a data warehouse? And when does it make sense to migrate to a marketing data warehouse?
Henrik:
Yeah. Easily put or simply put, the whole concept of moving towards a data warehouse comes from the need and the ability to store all of this data that we talked about from different sources, and storing it in one location where you can combine these data points and these data sources together, and start to get insights from all of them in a joint matter and compare stuff in a good way.
So that’s sort of the easy answer. At a certain point, you will need to do that because there are so many sort of data points you need to manage and to derive insights from them. But I usually talk about the key aspect here and that relates back to what you actually want to do with data. And that is to me to gain some form of competitive advantage in there. And that comes from gaining understanding of your business, of the environment you work in, the industry, the competition, the products, the sales you do, the marketing efforts you do.
So first of all, gaining that understanding, you need to derive these insights from all that data. And from that understanding, you can then make data influence decisions that drives, hopefully, this competitive advantage that you’re seeking. And the key point there is the word decisions, and if you’re not making decisions out of the insights you gain from all of that data, it’s basically worthless. And for sure, the activation piece here plays into to the decision-making here, and being able to do that in an easy way. And taking it even a step forward again, I’ve already mentioned machine learning, but being able to do that in an automated way at scale, then you can really start to think about the competitive advantages you can get from using data in a better way.
So that’s all of those reasons, data decisions, automation, scale, that’s the reasons why you should be moving to a data warehouse and for sure, within marketing.
Edward:
Yeah. I love it. Pretty compelling stuff. And Kashish, following from this then, some people advocate or may argue to go with a customer data platform, but what’s your stance on a sort of classic off-the-shelf CDP, versus an operational marketing data warehouse?
Kashish:
Oh yeah, this is a good… it allows me to actually echo first of all, Henrik’s point around why the warehouse? Because that really is the reason why off-the-shelf CDPs are something we don’t advocate for. And so, one thing that Henrik, I think I would want to double-click on is that the only place where all the data exists in your company is the warehouse. So you could use a spreadsheet, you could use a CDP, or you could use something like Google Data Studio.
And that would be a good partial view of your data, but if you want everything that you have, really all the click stream data, all the transactional data, and object data in your company, the only place where that’s practically available is the data warehouse like BigQuery. And so when a marketer’s thinking of, “All right, let me activate my data and use it to make a decision,” the only place they can utilize that has everything is the data warehouse.And usually it has everything in a clean formatted way because there are teams of analysts and analytics engineers that are organizing that data and making it consumable to other business users. So it’s actually a practicality thing, where really, if you were to go to a
Cloud data platform, what you’d be missing is usually the rest of the data in the company. That’s other SaaS data, data from other marketing tools like Facebook or Google, and then data from your transactional database. And so, that’s really joining all those different data sets is only possible in the data warehouse, and they’re only really practically available there.
And the other portion is really that an off-the-shelf CDP might take a company three to six months to implement in the best-case scenario. And the reason for that is because you’re instrumenting all of the event collection and data collection as well when you’re instrumenting that process. On the other hand, we kind of believe in companies being able to A, store their data wherever they want, which means they get full ownership of it, full control and privacy control, and full control of schema.
So in something like BigQuery, you could store your data as users, as patient objects, as doctor objects, hospital objects, whatever you want, and then be able to have events and actions on each of those objects rather than on just one user’s object, which would be the kind of format that a CDP expects.
And then second, you can also plug and play with existing data that you have. So most companies at a reasonable scale has set up a data warehouse at BigQuery, and then they simply want to use that data. And that analysis has already been done by BI teams and activate that analysis. So let’s say you have a team of maybe like 10 data people in your company, and they’ve been writing sequel queries and doing analysis for you for quite a while, setting up a CDP would actually take you several months before you can use that insight in your marketing stack.
Whereas setting up something like Hightouch and Reverse ETL, you could get up and running within five to 10 minutes. And usually we have to market it as like 30 minutes because people think five to 10 minutes is too hard to believe, but truly, we even have a Looker model selector now where you can directly go and select looks from a point and click UI. And therefore you’d be able to say, “All right, go to my Looker, choose a look and send the data from that Look over to any marketing tool in my stack, whether it’s Facebook, TikTok, Snapchat, Google, Bing, doesn’t matter.”
And the ease of use there is actually why a lot of people get started with Hightouch, and then even long-term end up using Hightouch because it’s really just practically extensible for any use case versus asking people to conform to a CDP’s data schema.
Evan:
Yeah. And just one thing to add on to that, I love, Kashish, the point you made about kind of the ancillary tools in the data stack, right? The Google Data Studios of the world that are effectively ephemeral connections to different underlying data sources. We have spreadsheets playing in there and we at Supermetrics often talk quite a bit about the customer data journey, right? And that journey may very well start with spreadsheets. It might start with dashboards and any number of BI tools.
And we see these as necessary stepping stones, essentially in terms of moving toward a centralized model, right? We talk about centralized data access models, decentralized data access models. And you can really have the benefits of both, I’d say to a large degree, where you can use decentralized data access models to quickly pull in data, prototype the data, kind of see what’s even available, see what you have to work with.
And once you kind of have a sense of what the data looks like, how it can be used, then you can go in and you can push that straight to a data warehouse. You can start to centralize that and distribute it out to the rest of the platforms. So we see the merits of both different solutions and with different teams, different levels of capability. Some do like to have that on ramp and that kind of auxiliary function in order to ensure that they’re getting all of the functionality out of the final product that they want to deliver.
Kashish:
Yeah. And one thing I kind of see happening over time is that those two scenarios will merge into one. So the way you think about like decentralized data and centralized data will actually all be built on some centralized source. So like with something like Supermetrics, it’s just as easy to get your Facebook data into a spreadsheet as it is into BigQuery. And then BigQuery now has actually pretty awesome UIs that let you spin up a Google sheet from BigQuery data, or spin up a Google Data Studio exploration via the BigQuery data, or using Hightouch, get that data into some other place.
So because the interfaces on top of that data are so easy and getting the data into BigQuery is now so easy, you can almost think as just like a database that backs something like a Google Sheet. And so there doesn’t actually have to be material difference in terms of where the data’s stored if you can get a nice difference in terms of the interface with which the users accessing it.
And so, what we’re kind of seeing a trend towards is everyone is moving towards that centralized model because of how easy it is to set up, and because it’s actually just a much more principled thing to do where those transformations and that data set live in one place, because just as you mentioned, different people can come in and then consume those and say, “Hey, Evan set up a model that was really useful to me. Let me go use that and then let me set up another one too.”
Evan:
Yeah, definitely. And having the ability to centralize and govern absolutely plays a huge part in it as well. We could certainly talk about the merits of centralization and governments, and ensuring that people are reading off the same metrics when they come to the meeting with the same reports. So I think we’ve all been in rooms where the data’s not the same, despite the fact that it seemingly purports to be. And yeah, those are certainly big benefits of the centralized model as well.
And if you can use that as an area where you can rapidly prototype and do some R&D and some data exploration as well, and I’d certainly encourage everybody that feels comfortable to do so, to do so.
Kashish:
Yeah. And one thing to touch on, especially from a Supermetrics perspective is, let’s say you’re getting Facebook data, Google data and Snapchat data into some central place, right? Because you want to do attribution on which of these dollars spent resulted in leads converted? Facebook will tell you one thing and it’ll say that, “Hey, I made you a lot of money and I converted you a lot of leads.” Google will also say, “I converted you a lot of leads,” and similar with Snapchat, but only in the warehouse and with SQL can you actually identify which of those different lead sources were final touches, first touches, or middle touches? And give them actual attribution points for converting that lead, otherwise each one will take full credit.
And so I’m not saying that these companies are trying to take credit for your leads. I’m just trying to say that the only place to stitch together true marketing attribution is someplace like a warehouse and with a data team. And that’s exactly what Evan was describing that having a central definition for these things is really useful. Otherwise one ad team might be using the Facebook definition of leads converted, and a different ad team might be using a Google definition, and they both might be taking credit for the same lead, which means that you’re actually double paying or you’re actually double attributing. And in either case you’re probably making the wrong data driven decision.
Edward:
Yeah. So I think there’s a lot of really, really good background context here. So if we go forward, let’s dig into what the future solution looks like in more detail and see how we can turn BigQuery specifically into a marketing engine. So Evan, I want to come back to you as you obviously work with a lot of marketing analytics and data teams on this. So what are the important things to consider when it comes to centralizing marketing data in an operational marketing data warehouse?
Evan:
So the things that are most important, I would say are identifying your use case pretty early on, because the use case is really going to drive how you go about building out the solution that you want to deliver. And when we talk about use cases, there are really, I’d say three that stand out to me in the marketing space where the combination of Supermetrics, BigQuery, and Hightouch really stand out and not only deliver business value, but deliver it actually fairly quickly as well, because obviously, the speed of projects and the speed of implementation is pretty important here.
So the first use case is like Kashish said, is attribution, right? So there’s a couple of different ways you can do attribution. We see kind of two primarily, one of which you can do without Reverse ETL. That is essentially just kind of pulling in the data from all of your different ad sources, pulling in the data from your analytics platform, whether that’s a Google analytics, GA4, Adobe perhaps. There a couple different ways to do that, going and looking at all the individual customer touchpoints and assigning a credit, right? That’s one way to do it.
The other way you can do it is you can pull all of your customer interaction points into your warehouse using Supermetrics. You can then push a lot of that customer data. Maybe you’re pulling that data in from your CRM, from the leads that you’ve captured, or potentially a POS. You can then push that data through Hightouch to the individual APIs themselves, and actually give them data where they can say, okay, Google can say, “Yes, we’ve seen this customer. We exposed them an ad,” and they can actually enhance the ability for their kind of conversion reporting as well.
And so having kind of both of those different mechanisms for ascribing credit will give the platforms the benefit of the doubt in terms of, hey, best case scenario, what did you deliver in terms of leads, in terms of conversions, things like that. And then we can also have our reference for, okay, what did our analytics, what did our website say for conversions? And where do we ascribe the credit?
So conversion tracking and attribution I’d say is probably the first. The second is we see oftentimes in the agency side, we see customers who are, let’s just say paid media agencies. They have a whole list of customers where they want to understand, “Hey of the customers that I’m doing media for, who’s spending the most? What do those trends look like?” And they want to see that not in a spreadsheet, not in their BI tool. They want to see that inside of Salesforce, where their actual account managers are doing the work.
They want to able to take their account, sort by ad spend, and that’s actually a pretty unique combination that you can achieve with Supermetrics, with BigQuery, and with Hightouch. Streaming the advertising data from the different companies that you’re working with, pull it in and be able to actually see at an account level, how much are my clients spending? What do those look like? And that can be a really unique way to understand which accounts within the agency or media portfolio are performing well, are maybe going down, might be in risk of churn? So that can be really important as well for the agencies that might be tuning into the webinar here.
The third that we see is primarily around audience building. So using Hightouch, you can push your customers essentially to these platforms. You can push their conversion activity, any kind of events that might have taken place that you’re aware of. We pull the data from the CRM, Supermetrics does. We store that inside of BigQuery. You can build those audiences based on your SaaS data, based on any kind of customer interactions points you have. Maybe you want to add them to exclusion lists after they purchase so that they don’t see advertising anymore. Maybe based on their website activity, you’ve identified that they are likely to be interested in a certain product that you’re marketing. You want to add them to an audience group so that they can then start receiving ads there.
So you can actively use the combination of Supermetrics, BigQuery and Hightouch to really accelerate a lot of the audience development that you might be doing with other potential onboarding platforms. So this is a really powerful use case where you’re actually really putting your team, the marketing team and the data team in charge of how you actually ascribe different audiences to different campaigns. And that’s, I think in my opinion, one of the cooler pieces around building an operational warehouse, right? We talk about what is an operational warehouse? It’s something used to actually drive the marketing operations in the day-to-day efforts here.
Kashish:
Yeah. And I think one thing that is uniquely possible in the stack is building segmentation off of your first party data, rather than off of third party data that those ad publishers have. So for example, as I’ve been describing, you could identify users that are higher spending or higher likely to buy based on some traits in your app. So for example, let’s say you’re selling glasses online, the people that fill out your survey online for which types of glasses they like are much more likely to convert than people that just visit your website. And so you’re probably going to want to send those users higher spend ad campaigns and potentially even personalize those ad campaigns based on the results that survey form. So if I choose metal glasses that are for women, I should receive ads for women’s glasses that are metal frames. And that’s the kind of personalization that you can do if you send this data back to Google and Facebook, otherwise you’re just segmenting based on things that they have, like demographics and zip code.
And then the reason the stack is pretty important is because in tools like Google Ads and Facebook Ads, the segmentation builder that they give you is based on their third party data like demographics. And even if you send them all of your conversion events and your first party data, they won’t give you an audience builder for that first party data, and that audience has to be built upstream in BigQuery.
And whether you do that with SQL, or whether you do that with Hightouch’s audience builder, doesn’t really matter. Both are really valid solutions, but you do have to do it upstream in the place where that third party data and first party data is merged, which is the data warehouse.
Evan:
Yeah. And I think that to put some words and some labels on that, we see that for audiences that are used in certain campaigns, you can build them using demographic targeting, as you suggest Kashish, with kind of the native audience builders in the platforms, or you can do behavioral targeting. And I think that most marketers in 2022 know that behavioral targeting is really the way to go in terms of delivering higher ROI campaigns, in terms of just really delivering efficiency. If you can capture that user activity, use it to inform the interactions that are being ascribed in the website inside of your app. All of those are more likely to lead to more efficient marketing budgets, more leads, more sales at the end of the day.
Kashish:
And one really cool example of this that we see, I think this is the most advanced version of the marketing campaigns is where in those campaigns you include a product skew for which skews reviewed. And then also to Google and Facebook, you upload a pick list of product skews, product names, and product images. And that way in your marketing, in your ads, you can actually have a carousel of the actual products that were viewed, and then Facebook and Google become aware of, “Here’s the exact skews I should market to this specific user,” and update those skews in real time as that user reviews more and more skews.
So most websites are selling many different products, and as a result, you might have tens of different skews that you want to market to that consumer, and then each consumer would’ve a different skew. So the only way to really build automation around that is to have for each user, a list of skews that need to be marketed, and then be maintaining that constantly in the warehouse. And when it gets that complicated, you really can’t imagine a spreadsheet being a good solution for the problem, because as you can imagine that spreadsheet gets outdated in 24 hours or less, and it’s not going to be able to handle the complexity of all the skews and also maintaining, updating those skews in Facebook and Google.
Edward:
Yeah, absolutely. Oh yeah. Go ahead, Kashish.
Kashish:
Well, to Evan’s point, the cool thing about running campaigns this way is that you’re not spending ad dollars unnecessarily. You’re spending them in a very personalized way, and it not only increases your ad efficiency and return on ad spend. It also increases the user experience because sometimes as marketers, we’re thinking, “All right. Well, we want to spend as few dollars as possible to get as many leads as possible,” but as that end consumer, I would also prefer to see products that I’ve actually viewed or categories that I actually viewed.
So if you’re a sporting good store and you sell ski equipment and basketball equipment, I’d like to see ski equipment if I’m buying ski equipment. And so personalization actually also results in much better content distributed to end users.
Edward:
Yeah, absolutely. And to that point, we also see a lot of marketing teams, as Evan mentioned, migrate from spreadsheets into data warehouses to enable all these use cases. And Henrik, you spoke earlier about some of the benefits of marketing data warehousing earlier on, but what are some of the benefits of BigQuery specifically as your data warehouse?
Henrik:
Yeah. Yeah. And I would say, and we mentioned that sort of area already from an ease of use perspective, that’s something I talk about with our partners and organizations and companies interested in moving towards GCP and BigQuery. That is, I would say the key benefit of using BigQuery and for everyone small or large as a company. So BigQuery is completely serverless. It has no infrastructure to manage, no clusters to sort of provision or size or scale, and plan for the future, or sort of even relate to sort of what queries you’re running to a certain degree.
So you take away all of that complexity that usually is there, for sure in the on-prem environment, but also compared to others, and you just focus on what’s important, and that is sort of the insights you can get from the data and querying that data and building these use cases that we talked about now. So that is for sure, the key aspect of BigQuery and why people use it. Other, there’s tons of benefits for sure, related to GCP and BigQuery as well. To highlight a couple is definitely the unification of sort of building a data lake and a data warehouse as well.
So that’s something we now sort of handle within BigQuery. You can store data in files, in cloud storage, in databases, like outside of BigQuery, you can query data in other Clouds, stuff like this. So we are sort of unifying that sort of whole space, making it easier again. That’s one area. Another area is sort of the built-in machine learning tools that exist in BigQuery, simple SQL query. And then you have a model that could possibly predict lifetime value or propensity, or clustering of customers in segments based on data as a base.
And you can use them directly on top of your data, those models in BigQuery. You don’t have to extract the data and do it somewhere else, and then implement it somehow. It’s very sort of, again, coming back to ease of use and all of that, but functionality wise as well. And then you can talk about, we handle streaming data. We have sort of very good integrations with BI tooling from GCP, but also other ones, for sure, we open up sort of to run whatever BI tool you want.
And what else? For sure, the ecosystem around BigQuery, Proofpoint here today, like Supermetrics and Hightouch, building the ecosystem around it through APIs, providing the ecosystem, the opportunity to do that, add components to build this sort of decoupled sort of approach to building a marketing platform. And yeah, so main point, ease of use. You can talk about scalability and all of that stuff, but it just comes into play when it’s a really big organization. Most companies are smaller and will never sort of come into that area. Just the ease of use, I think is main key issue here to talk about.
Edward:
Yeah, absolutely. And Kashish, I want to focus in on the third and final step now to close this loop and take that data from your data warehouse and put it back into your marketing platforms, which is achieved through reverse ETL, but Kashish, I know you prefer to think about this from a data activation perspective. So can you elaborate on this term and talk us through how marketing teams can actually activate that data?
Kashish:
Definitely. So data activation is all about taking data that you already have and putting it to use in your marketing stack. And that can be for lifecycle marketing, for ad marketing, or for really, any sort of personalized marketing that you want to run, and it’s meant to be omnichannel. So the whole idea is that you have insights on your users. You have things like which products they viewed? Things like how many times they log in? Things like have they filled out a demo form or some sort of email form? So now you can directly target them. Have they downloaded an iOS app? And all of this kind of behavioral data that Evan mentioned earlier, and based on the behavioral data, you want to run different marketing campaigns on those users and different personalizations on those users. So data activation is really about giving marketers access to that data, and then the ability to turn that data into different automations in their marketing stack.
So for example, you might have a life cycle marketing tool, and that marketing tool has set up different campaigns for things like product viewed, product checked out, things like one year since last order date, one year since last birthday, et cetera, and you want to power emails in that marketing tool. And there’s two worlds that we think data activation can help with. One is getting all of your data into that marketing tool. It could be something like Iterable or Braze, or some other email tool. And that way, you can then build those audiences and those customer segments and those marketing campaigns in the ESP or the email marketing tool.
And the whole idea there is that whatever data set you have in your company and whatever insights you have such as LTV/CAC et cetera, sync those out to your marketing tool. That way, your marketing tool can be LTV aware. Like how many people have LTV in their email marketing tool and can actually say only send this email if the LTV is greater than $50? So running campaigns like that is actually not that easy, unless all the data is available in the email marketing tool.
And so we think of data activation sometimes as basically data proliferation and send all your data sets to your marketing tools, and therefore be able to run marketing campaigns on that data. And the cool thing there is that you’re putting that data directly in the hands of marketers in the tool that they’re already used to working in. So they’re experts at building campaigns in things like MailChimp, Iterable, Braze, Google AdWords, and so on. And so let them build those campaigns there by giving them all of the data and insights that you’ve achieved in your warehouse in those email tools. So that’s like one mode of operation, which is really sharing that data set with the marketing team in the tools that they’re used to using.
And then the other tool of mode operation is allowing marketers to build user segments themselves directly on the warehouse. And so there’s two tools that we have for this. One is called Hightouch Audiences, and the other one, which we’re announcing in a couple weeks is Hightouch Audiences Traits. And the idea for Hightouch audiences is to give marketers a point and click UI on top of their data set that lets them subset their users. So you can do simple subsets like, “Show me all users in a given state or all users that have spent more than this much money?” And those are like pretty simple filters to run.
And then more complex ones such as, “Show me users that logged in and then viewed, or viewed and then purchased, or viewed and then did not purchase?” And those are more funnel queries, but any sort of user steps that you can imagine for your data set, you can build in Hightouch audiences. And the only setup required is that your data team share with Hightouch the schema for how the data is related to each other. And then the marketer doesn’t have to worry at all about things like joins or any sort of sequel query language.
So Hightouch Audiences lets you filter data you already have, and Traits lets you actually create net new metrics on those users. So I could say something like, “All users who have viewed more than three items in the past week are now likely to buy users,” and create a true or false category for likely to buy, and compute that trait automatically every few hours. And now using that trait, many different marketing teams can take this likely to buy user concept and run ad campaigns, unlikely to buy users, or you could create things like categories viewed.
So you could compute a trait for which categories that user has viewed based on different product categories that they’ve viewed most commonly, and then write that back to the warehouse. So the cool thing here is that you’re actually generating net new data based on ideas that marketers have. And one thing that we like talking about with marketing teams is experimentation. So experimentation happens when the barrier to entry for doing experimentation is low.
So if I can easily experiment on, let me just figure out which categories my customers are viewing. And then based on those categories, run a different ad campaign per category. If that’s going to take me six months to run because it requires a lot of development and code, then I’m not going to run that campaign because I’m only going to run it when the ROI is very clearly high. If it takes me something like two hours to experiment with and test, then I’ll probably be very likely to dedicate a few thousand dollars testing it, seeing if it works and then scaling it out.
But because the barrier [inaudible 00:39:53] right now is pretty high for experimentation without tools like Hightouch and Supermetrics, that doesn’t happen as often as it should. And that’s really what we like to encourage with being able to build these campaigns and being able to build traits on users. Oftentimes, the marketer actually knows what user subsets they want to run before the data team or the analyst knows. And it is just a really good way to share that definition with the analyst team, or to be able to create that definition yourself and then be able to share with the rest of your marketing team.
So as far as use cases go, it’s really to echo what Evan was saying around sending conversions to your ad tools and then sending user audiences to both your ad tools and your email tools. And usually it’s for the purposes of either increasing conversion and therefore revenue, or decreasing cost of acquisition. So being able to better target users and therefore send them more targeted messaging, that results in them converting quicker rather than through more and more dollars spent.
Edward:
Yeah, really good points. Evan, Henrik, anything to add before we move on to some audience questions?
Henrik:
I think that was good. Covered everything.
Edward:
Cool. So we’ve covered a lot, gone through the stack, several use cases, a lot of really good discussion points, but we’ve got some great questions in through the Q and A. So I think we could spend the next 15 minutes or so on going through those. So let’s take the first one on privacy. “So is now the time to invest in a data warehouse, looking at all the challenges related to data collection?”
Who wants to take this one first?
Henrik:
Yeah. So I can comment, and for sure, it depends on what you sort of look as data collection and challenges that for sure, this whole concept of first party data is key here for organizations to have an advantage on for sure, the marketing side or any side using data. So creating sort of a valued exchange with your customers, a way to easily share data and they trust you in that sense is getting more and more important, and you need to put efforts into this.
So that’s the first piece, and from a sort of technology perspective and data collection privacy matter, all the technology is there to make it absolutely private and sort of handle all of the challenges related to this. It’s just a matter of doing it correctly in a good way, and for sure, put efforts into it, but the possibilities are there for sure. And yeah, if not doing it, I think you will lose a lot of benefits coming out of it.
Kashish:
One thing to add is back on the point of first party versus third party data, with IDFA first party data becomes a lot more important because Google and Facebook’s pixel that you put on your website can’t track all of the information that it used to be able to. So there’s a lot of information and identity resolution that you can only do in the warehouse, that we then have to manually send back to Google and Facebook because they can no longer kind of collect that information.
And so whether it’s a data warehouse or not, it does have to be some central way of collecting that information as a first party and then sending it to those ad tools. And currently, we believe the data warehouse is the most future forward and long-term correct way to do this, but it’s okay if you start with something else in the beginning.
Evan:
And I think the thing to add there is, I think classically, doing this would require a CDP. And that’s one other leg in the value chain that you need to then include in your terms, in your DPA, in your privacy section, in your website. And essentially, if you can then be the processor of that data, and Google is fully SOC 3, ISO 27k certified. So you have the ability to store that data and to secure it, assuming you have all the proper legal checks and balances and you no longer need to actually worry about having additional vendors or different additional points of risk in your data custodial retention.
So those are some other key points to consider, is that to a large extent when you’re dealing with customer information, yes, you do need to be very careful. And one of the ways you can minimize the rate of risk is to be working with fewer platforms or fewer companies who are actually touching or managing your customer’s personal information.
Edward:
Yeah. Really good points, and got some more questions. So let’s see how many we can get through. So this one is from Christian on GA. “So since Google Analytics Universal is going to be shut down in a year or so, we would like to download all historical data from GA into search metrics, and/or BigQuery in order to preserve the data for later analysis and comparison. Is that theoretically possible with the free universal GA version? Do you know if we lose any data since GA, API and the free version cannot export data in ROM files to BigQuery?”
Evan:
So I can take a crack at the first part of that. So luckily here at Supermetrics, we’ve been exporting data from GA really since day one. It was the first connector we ever built. Our founder, Mikael Thuneberg, that was kind of the first thing that he did. And so, we support pretty much every metric and dimension on the original GA, API. So that is certainly something that many new customers have come to us for in terms of helping them create historical exports for their GA data, saving that for benchmark purposes, especially because with having the COVID epidemic over the last few years, for a lot of companies, 2019 was the last time they had a truly normal year to benchmark against. And so having that historical data as an analytical reference is very important.
On the GA4 side, there’s a few different ways to approach it. We have a connector to GA4 as well that leverages the Google Analytics for API. We can load that data into BigQuery, into the spreadsheet of your choice, and that works just fine. I know that Google also has a native integration as well, where the data comes in and is loaded automatically into BigQuery. So you’ve got a couple of different options on the GA4 side of things, but in terms of universal analytics, that’s certainly an area where Supermetrics can help.
Edward:
Yeah, good stuff. And then we have another question. This is from Camilo. “So what do you recommend for small businesses with limited budgets to measure marketing data and get insights from it?”
Evan:
Maybe one thing we haven’t really touched on in the webinar at all is the cost of all of this, right? Everything that we’re talking about is going to be fairly nominal in terms of setting this up. Google has a very compelling free tier for BigQuery, and I’d estimate that most of our customers, probably 60% of our customers stay within the free threshold for BigQuery. I forget what the exact limits are, but I want to say up to your first five terabytes are totally free for querying, and I think it’s up to 10 gigs of storage.
And for those of you who don’t necessarily know how big the marketing data that you’re working with is, that’s quite a lot, at least in our experience. Larger advertisers will certainly surpass that but for smaller organizations, I think you’re probably going to be on the free tier there. Hightouch also has very compelling pricing as well. That leaves us on the Supermetrics side, we have pretty flexible packages with our sales representatives who can basically build a custom tailored solution to whatever it happens to be that you’re looking to accomplish.
So overall, the cost of spinning up a lot of this technology is fairly low and fairly approachable even for the smallest organizations.
Kashish:
Yeah. And as mentioned before, from the BigQuery GCP side, it goes from zero to almost infinity. So there is no sort of minimum level of size you need to be able or to use BigQuery. So if you’re small or if you’re large, it still works and the free tier there for sure fits into this whole thing as well.
Edward:
Yeah. Awesome. Then we could move on to the next question, a good follow up. I think Evan, this one is for you. This is from Miguel. So the next on the list, “How does Supermetrics differentiate from ELT tools like Fivetran?
Evan:
Yeah, certainly. So this one is one of my favorite questions to answer. So we differentiate and I’d say probably kind of one of three ways. So depending on who we’re talking to. Fivetran specifically, I would kind of put them in the same group as Stitch, where they’re kind of generic ELT or ETL vendors, depending on how you use them. So those are companies who essentially, are really good at moving data between different databases among each other and companies that have a lot of connectors to different SaaS platforms.
Whereas when you come to Supermetrics, you’re coming to a specialty shop. We specialize here in marketing data and advertising data and sales data So you’re really coming to a place where we are supporting really the highest level of metrics interventions from the different APIs that we connect to. We are prioritizing in our go to market strategy, newer and potentially more… I don’t want to say fringe, but maybe lesser used advertising platforms where we’re really trying to serve the marketing data function as best as possible. And we believe that by specializing in the area in which we operate, that essentially delivers us the best result.
That’s one way. The second is we have a handful of competitors on kind of the BI tools spreadsheet side. Supermetrics, I think there we differentiate and they might be specializing in marketing data connectors as well. I think there, we differentiate in terms of we’ve been doing this largely for the longest. We have the largest number of customers among many of our competitors, and as a result, it’s given us the resources to really invest in connector depth, connector quality, and now expanding that into connector breadth as well. So there’s quite a few compelling reasons to choose Supermetrics over those other smaller competitors.
Then we have other competitors who do specialize in marketing and they specialize in the data warehousing side. And that’s where I would say we differentiate in terms of being able to serve many different customer use cases across the data life cycle. So getting started in something like Google Sheets or Data Studio, very quick and easy on ramp for many in the organization where you might have someone on your marketing team using Supermetrics today, and you might not even necessarily know it. And we can essentially help them take those queries that they’ve built out in Google Sheets, Data Studio, Excel, and translate those into BigQuery fairly easily.
And so we provide that on ramp in a way that it is not only accessible to get started, but it’s easy to start moving this in the direction of a marketing data warehouse.
Edward:
Yeah. Great stuff. And then following from this, we have a really good question from an anonymous attendee but, “To build an actionable data stack, how should we structure a team? What are some of the skills we should look for?”
Kashish:
Well, I think a couple things. One, at least having someone on the team that can set up resources like BigQuery. And again, given that there’s Supermetrics and GA4 can be added directly to BigQuery with click a button, the technical barrier there is going down, but upon doing so, you do still need someone that can write SQL to transform that data into something that’s usually unusable. And so at least one person in the company that’s good at writing SQL would be useful to set up the stack in the way that we’re recommending.
And then otherwise having someone on the marketing side that wants to make easier driven decisions. We see a lot of marketing teams that are looking at more broad-based marketing and just increasing spend. And so your marketing team at least has to have some intent to run personalization and segmented campaigns, which not every marketing team does, but as Evan mentioned, we’re seeing this become a lot more common now.
But I think this talk is really about how the skill set needed is really, almost really just intent and no hard skills given how easy these tools are to set up. And I think that’s something that all of us are really proud to say that over time, everything is becoming more accessible in terms of actionability.
Evan:
And Kashish, correct me if I’m wrong, but this at some companies, of course, depending on the size and scale, this can be one person. Doesn’t necessarily have to be, doesn’t mean it should be, but it certainly can be. I know for example, in previous roles of mine, I was gathering data, preparing it, putting it in the dashboard, and also making marketing decisions using that data.
Of course, this was for a very small company with limited resources, but I was able to get it to work. And so, because these tools have become so easy to use, you have a lot of operational leverage essentially, where instead of having to hire two, three, four, five employees in order to get this to work, you can pay a couple of vendors a nominal fee per year and really stand up something that’s pretty impressive until it delivers a lot of value.
Edward:
Yep. Good stuff. And then we have one final question from the audience. Kashish, I think this one is for you. “So user audiences tracking and creating within Hightouch, what data setup is needed from a website tracking perspective? For example, information in the data layer or information in GA, or something else?”
Kashish:
Yeah. And Alec, our subject matter expert on MarTech actually beat me to it, but he’s totally right. You do need both customer data and behavioral data in the warehouse for that if you want to do it right. But luckily, the OTB feature from GA4 will ask you automatically to sync this down to BigQuery for you. And that’s actually how I would’ve answered Henrik’s question around why BigQuery integrates so well with the Google ecosystem, that you actually get event tracking for free out of the box, which means you’re potentially skipping quite a big step with GA4.
With the original GA, you didn’t get the same level of granularity for those user IDs. And so, having those upfront with GA4 actually makes it a lot easier and it’s really just, you just build the tags in GA4 and then those users will pop up in BigQuery automatically.
Edward:
Cool. All right. So that was our last audience question and we are coming up to time, but before we wrap up, are there any other final points or closing comments you’d like to add based on what we’ve discussed today?
Evan:
I think one additional use case that didn’t come to mind, but is I think an experience we can all relate to is negative targeting. And for those who don’t know, this is when you go online, you buy something, and then you see ads for it for the next week or so. I don’t know about you, but as a marketer, that frustrates me. I see this product and I’m like, I already bought this guys, spend your add dollars somewhere else. I don’t want to see it. I don’t want to think about it anymore. This is one other thing that can be achieved using this tool.
Kashish:
Yep.
Edward:
Cool. So, all right. I think we are coming up to the hour. So this is all we have time for today. We’d like to thank everyone here so much for joining us, and if you’d like to learn more, you can obviously go check out Supermetrics, BigQuery, and Hightouch. Feel free to reach out to all of us and myself. We’ll also follow up via email and send you the recording most likely tomorrow, but I think by the end of the week, so look out for that.
Edward:
And finally, I’d like to thank our guests, Evan, Henrik, and Kashish. Thank you all so much for joining us today and see you at the next one.
Turn your marketing data into opportunity
We streamline your marketing data so you can focus on the insights.