Breaking the silos: Why you should own your marketing data

Transcript

Joy:
So should we start with you, Jim? Tell us about you and Snowflake.

Jim:
Yeah, thanks, Joy. Really excited to be here. So I'm Jim. I'm the field CTO for Advertising and Marketing at Snowflake. I have the pleasure of helping our customers solve their super interesting advertising and marketing-related challenges on top of Snowflake. Prior to this, I was co-founder and CTO of a company called Upwave, which is an advertising measurement company that does reach and frequency as well as brand lift for cross-channel campaigns. So linear TV, CTV, digital. So that's where my deep knowledge of advertising and ad tech comes from.

Joy:
Great. Over to you, Jack.

Jack:
Thank you, Joy. Yes, I'm Jack Bitcon, a Solutions Engineer here at Supermetrics. I've been here for about a year now, but I was a former product user before actually hopping on board. As I alluded to, I used the product in an agency-type format beforehand. So familiar with the different implementation techniques and potential use cases that we see on a day-to-day basis. And super excited to be here. Appreciate the opportunity.

Joy:
So yeah, and finally, my name is Joy. So we're three Js in the house today. Just noticed that. I'm from the Supermetrics Marketing team, and I'll be your moderator throughout the discussion today. So to kick off, why are we here today? It's the current challenge, the current situation that many marketers here today, including myself, face. There is so much data available across different platforms, and there is always the challenge of proving the ROI of certain campaigns, channels, and marketing overall.

And then, on top of that, the platform and the tool that we use, we love, for years are phasing out. I don't know if you noticed, but Google started cutting down to GA Sunset this week on their social. So we have less than 100 days for marketers to migrate everything from GA to GA4. So doing marketing in 2023 is pretty hard. So with all of that challenges, Jim and Jack, why is it important for brands, companies, and marketing teams to take control of their marketing data?

Jim:
Aside from that, you mentioned Joy, just the macroeconomic situation we're in right now. We're in another advertising slowdown, which happens every once in a while. And when the ad market slows down, we always get a lot of pressure to know where we're spending money, to make sure that the money that we're spending is well spent, to trim. Probably a bunch of marketers on the call are faced with a situation where they're going to have to trim their marketing spend a little bit. So in that environment, I always say, you can't know where you're going unless you know where you are now.

And so it's super important to have the data to know where you are spending. How much are you spending? By what channel? Putting everything into a common language so you can compare apples to apples. That's the first step. And everything else that's more complicated has to come after that, but the first step is to know where your money's going. It's as simple as that.

Jack:
Absolutely fantastic points there. And to your point of the current macroeconomic climate right now, I'm building off of this question, not only why companies should take control of their data, but why they should take control of their data in an automated fashion. There are so many man-hours, so much time, and so much work spent on hopping into Facebook Ads, downloading that CSV, and either copying and pasting that data into somewhere else or uploading it into an in-house data warehouse.

Now, with something that helps automate that process of taking control of your data, you can dedicate that time, those man-hours, those resources to actually analyzing that data, drawing insights from what you're actually collecting, and can draw different conclusions based on different things that are going on. So between automating the process and actually trying to gather where we're currently at in today's climate, I think taking control of the data is the first part of that situation.

Joy:
Yeah, I think you guys pretty much sum up all the challenges marketers face today. A lot of problems with marketing attribution, proving the ROI of your marketing dollars. So then, Jim, I'd like to ask you this question because there's so much data that the marketing team is collecting, and they have their ads running across all channels. What would you advise them to start collecting first? What kind of data should they collect?

Jim:
Yeah, so I'd start with, "What are the channels where I'm spending the most money?" So if I'm spending the most money on Facebook, I should probably start with Facebook data. If I'm spending the most money on Google, then I should probably start with Google. So I'd just start there, "Where is the most money going?"

The second thing I'd ask myself is, "What's the problem that I have today that I'm struggling with?" And I think there are so many uses for this data, and there are so many things you can do with it, all the way from the most basic things all the way up to the most complicated. And I think you have to think about what that use case is. Maybe it's if you're on the marketing team and your finance team is constantly bugging you about where your money is going, and you just want to have a dashboard that you can share with them so that they can just look and they don't have to ask for updates all the time, maybe that's the problem.

Or maybe you're working with an ad agency, and again, you're trying to simplify the communication. Maybe you're trying to unify data that's coming from different ad agencies and that you're using in different countries. There are so many sorts of uses for it. The first question is, "Well, where's my money going? 'Cause I should probably get data from where my money is going first and foremost." And secondly, "What's the use case? What's the pain point I'm trying to solve?" And I always think it's better to go deep, get all the data you need to solve that one problem, and then you can broaden and cover more.

And then, I'd think about how hard it is to get the data. Aggregated data is easier to get if you're using a tool like Supermetrics, to data that Supermetrics can easily pull for you. Start with the easy-to-get data. Some of the hard-to-get data may take more time. Always start with the lower-hanging fruit.

Joy:
Yeah, exactly. Exactly. I can echo the need for a shared dashboard like you said — for example, if you work with an agency, we work with an agency on our campaign. It's just so important for our communication to have a centralized report and dashboard to communicate what's happening, what's changed, and discuss the suggestion or how we can improve the campaign. So it's a two-way street. Anything to add, Jack?

Jack:
Yeah, to build onto Jim's point, honestly, from my perspective, as much data as possible is always good. You don't know what the future's going to hold for what you're going to need. There could be a use case that's important for you all to address right now, but who knows how that use case is going to evolve over time? And sometimes, using a more robust tool like a data warehouse, like the data cloud provider from Snowflake, you can use, you can move more of that data into the destination for maybe whether it be a current use case or a use case that you see in the future.

With the whole GA to GA4 migration that we're seeing right now, we see a lot of clients on a day-to-day basis who are like, "I just want everything from Universal Analytics that I can get," which is a reasonable ask is that data's going to be going away eventually. But essentially, if you want to do something like that, you're not going to be able to pull that into a Google sheet or an Excel document or analyze that data in a visualization tool. You're going to need some sort of middle ground to store, process, and analyze that data before passing it on.

So in my mind, as much data as you can collect is a good thing, but to Jim's point, narrowing in on what the current, what's the word I'm looking for here? Use cases are the current priorities for what we're trying to do. And then, looking into the future, what could be cool to look at in terms of your ad spend, your analytics platforms? There are a million different data points on a lot of these APIs. And the big KPIs that we all work with are consistent across the board, but some of the insights we can draw from some of these more niche data points are pretty neat.

Jim:
Yeah, that's a great point, Jack. Definitely err on the side of pulling more data rather than less, I think. You know, don't want to find yourself in a situation a year from now where you're starting some MMM analysis, for example, in a year, and you go, "Oh, I really wish I had this data from a year ago, but when I started my journey, and I started pulling the data, I didn't pull that data, I left that data where it was, and now I don't have it, and I'm trying to do some analysis, and I don't have the data." That's the advantage of using a tool like Supermetrics. You can just take the data that's available and drop it into Snowflake, and it's not going to do any harm even if you're not using it in the immediate term.

Jack:
Exactly.

Joy:
I want to expand on this because, Jim, you wrote an excellent article about data ownership. And you mentioned a lot about the idea of marrying first-party data with some of the apps, so I want to get your thought on that and dig a bit deeper into that point.

Jim:
Yeah, I think one of the really powerful things about bringing ad data into Snowflake with a tool like Supermetrics is that it is there sitting in the warehouse in the data cloud for you to join it with other data you're bringing in. We probably have a lot of folks from the marketing department on this webinar, which is super exciting. But one of the good things about Snowflake is that it's also trusted by your IT team. And so there's probably lots of data getting into Snowflake today. Point of sale data is getting into Snowflake, and CRM data is getting into Snowflake. All that data is making its way into Snowflake. And so when you bring your advertising data as well, it really allows you to break down silos and join data sources that you may not have thought about much before.

So one thing that's super important for me is, well when you're doing paid social, there's also all this free social that's happening. There are influencers, and there are all the people that are engaging with your brand organically, and how do you unify those so that you can leverage the learnings altogether? So that's a great use case. But as all this data migrates into Snowflake, it really enables you to break down those silos and not have, "Oh, this is where I look for ad data, and this is where I look for sales data," and all those different channels having their own their own data sources.

Jack:
We see it all the time when we've got a client, "Oh, here's my Facebook Ads dashboard, here's my Google Ads dashboard, here's LinkedIn, Twitter, whatever else, paid advertisement platform." It's like when utilizing some very point-and-click tool sets, and you can join that data together. So instead of individual ad platforms and dashboards, you get your complete paid ad spend picture in one dashboard or break that out into multiple use cases. But consolidating that into one place, again to your point, is step number one.

Joy:
Yeah, definitely. Data, if they're analyzing data individually by channel, is good, how that channel performs. But the real power comes when we combine different data sources and get more valuable insight from that. Very cool. Jack, you briefly mentioned why you want to own and consolidate your data in a data warehouse. So what are the common use cases that you've seen from working with a lot of customers, a lot of marketers, and brands?

Jack:
Yeah, I think that's a fantastic question. There are a few that we see very often on a day-to-day basis, especially over the last three to six months; we've seen a lot of GA to GA4, things I've been alluding to all during this entire call so far. But it's essentially, "Hey, how do we get enough Universal Analytics data, whether it be two years or five years or 10 years or whatever it may be, into a warehouse so that when we implement something like GA4, we can compare maybe not apples to apples, but red apples, green apples kind of between data sources, and utilize old Universal Analytics data as a guiding rod for how we want to implement GA four?"

Everyone knows we're not able to pull the same data points, and there's a little bit of bridging the gap that's required with tag manager and custom conversion events and things like this.

So utilizing historical data to implement some of these new platforms we're seeing is a very common use case, at least at the moment.

Another one that we see often is attribution modeling. "I see X, Y, and Z conversions on my platform, on my conversion platform. How do I know where those came from?" We see that very often. And it's difficult to join those data sets together. The points would be required to analyze that data and perform that sort of analysis or gain those insights — it's difficult to do inside of a Looker Studio or a Sheets or an Excel. So, generally speaking, those workloads are pushed towards a data warehousing, Snowflake type of environment. "So, where do my conversions come from?" It's a huge use case that we see on a day-to-day basis.

And then I'd say number three, you could do something like budget pacing, where you see, "I'm going to have $50,000 to spend on ads for the month across all of my platforms. How should I spend that data, or how should I spend that money? How do I get the data in from the investment I'm pushing, and how do I know how to pace that budget so that I don't get to the third week of the month and I'm out of money?" You can utilize this combination of products for all kinds of different things. Those three are the ones that we see on a day-to-day basis. And things that are very, or I'd say relatively simple to implement and things that can be very powerful in the long run.

Joy:
All right. We have a very timely question from the chat that I'd like to call out. So Andy was asking, "Now that GA4 is sunsetting, what are you seeing users do with their data before it gets fried?" Thank you, Andy, for the questions. Very timely. That caused a lot of panic among all of us. So Jack, please calm us down. What should we do with GA data?

Jack:
Absolutely. We should stash it; it's kind of the short answer to the question. We need to stash that data somewhere. It's going to go away. Google is sunsetting the API. There will be a period of time when we can access that data from the API. We don't really know what that period of time looks like. It's been anywhere from six months to 12 months. But essentially, at the end of the day, Google's going to get rid of that data at some point.

So stashing it somewhere that's not a Google Sheet or an Excel document but in a place like Snowflake. Because it all sits there, it's not going to get deleted from a data warehouse — it's not going to get deleted from Snowflake. It'll be there for use in the long run as we go through this experience together.

Jim:
I think it's an opportunity. It's an opportunity for people to think about their analytics more deeply than before. Google Analytics has just been there. And now that we're going through this, sometimes these sorts of events give us an opportunity to rethink how we're handling these analytics entirely. So maybe you have been using a tool like Supermetrics, but you haven't been storing it in the data cloud, for example. So maybe this is the time that you get to start doing that. You get to start experimenting; maybe you start trialing different tools, try to change your definition of a session, and see what that does to everything that you're measuring. I think sometimes these things give us a chance to do things we've known we wanted to do for a while that we just haven't quite gotten around to doing.

Joy:
Yeah, love your positive way of thinking of things. We always find this silver lining in these challenging times that maybe this is a sign for us to find a better way to store and take control of our data instead of letting them sit in this platform where they can be like, "Okay, I'm not going to store your data for more than 12 months, 24 months," and then you start panicking. Yes. All very good points there.

We spend a lot of time discussing the background of the challenges, the problems and then more on the use cases. So if we talk a little bit more about the solution, let's say, okay, marketers, now here, they might be interested in starting to consolidate all this siloed data. Jim, what should a marketer do? How can they get started?

Jim:
So I think you have to pick your platforms first. So we talked about figuring out where your most important data is because it's where you're spending the most money. We talked about picking that one acute use case that's a problem right now that you want to make sure you're addressing. Once you have those out of the way, pick your platforms, Supermetrics and Snowflake. And once you have that, you can start tackling the problems. You want to ensure that you have a tool that gets data from all your data sources. So make a list, "Okay, here's all the data sources that I have," and then compare that to the list for Supermetrics and make sure that they cover. They probably do.

And then, you want to have a platform to store the data where you can connect any BI tool. Whatever your preferred BI tool, you want to be able to connect that. You want something that can handle the scale of data that you have today and the scale of data that you'll have tomorrow. And then, make sure that you're picking a platform that will stand the test of time in terms of processing power, performance, and security, which is super important. Given that we're talking about Google Analytics and what's been driven by privacy concerns and everything, picking a data platform with data sharing where you're not going to be moving the data around when you're sharing data is important for futureproofing.

Joy:
Yeah, thank you. Anything to add there, Jack?

Jack:
Honestly, yeah, it's just to identify the use cases, like what Jim said, what's pressing right now. And a bit of thought towards what would be cool to do in the future? I always think it's a good problem to have. We're outgrowing the tool sets that we're using right now. We're looking for more enterprise suite tooling, identifying the use cases, and getting the data in, and then that's the fun part.

After you get the data into the destination, into Snowflake, everything from there is a bonus. How do you want to work with this data? How do we want to transform it? How do we want to gain insights from this? If we've got data from X, Y, and Z data sources, what's the best way to blend this? That's really where the process becomes fun, in my opinion. But essentially, step number one is identifying the use cases and deciding the data we want to collect and the platforms we want to use to produce that data.

Joy:

Awesome. I think we have some pretty good questions from the audience. Anything else you both want to add before we move to Q&A?

Jim:
Yeah, there's a question by MMM, which I love. And yes, MMM is expensive. So probably, there are some marketers on the webinar who use MMM, and it can be really expensive. Still, one of the reasons it's expensive is because gathering the data and preparing it's a really time-consuming process. And so we have companies doing MMM maybe yearly, maybe twice a year if they're lucky. And in a world where attribution is getting more difficult because of the loss of cookies, MMM is getting more and more important.

So what do you think? Do you want to use Snowflake as the base for your MMM? What do you do? Step one, you set up the pipeline with Supermetrics like we saw Jack do. Get all the data from all your channels, and bring it into Snowflake. Pick a timeframe. So do you want your spend daily, weekly, or monthly? How do you want your spending aligned? And then you really want to get spend by that timeframe per channel, and then you want to get sales per that same timeframe frame and then pick an MMM tool. And some of the modern MMM tools are able to pick up the data directly from Snowflake, and that'll save you the weeks of preparation of data that people are going through for MMM today.

Joy:
Yeah, that's a very good point you just mentioned there. A lot of marketers are interested in the topic of MMM in general. Do you guys see any good questions that you want to take on?

Jim:
Yeah, so I'll mention the data sharing about how the data are not being copied — it's a live share. That's specific to once it's in Snowflake. Obviously, the data sitting in Facebook or Google or Google Analytics, whatever, has to be copied to be brought into Snowflake. But once it's brought into Snowflake, when you do data sharing, like Jack showed with the example of an agency sharing end data to the end client, that data does not get copied. It's actually a live data share.

Joy:
Yeah, with UA, it's a bit tricky because Google will stop processing new data coming in. So you won't get any new data from GA essentially anymore, but rather GA4. If you start migrating to GA4 already, that's where you want to get your data. Okay, I have a question here. "How long does it take to implement this solution, and what maintenance is required?" I guess the audience is referring to setting up a data warehouse in Snowflake.

Jim:
Yeah, so Snowflake is a fully managed tool with really little maintenance required. Compared to legacy systems, it's going to be a lot simpler to get up and going, and there's as little maintenance required. It heavily depends, as I'm sure it does for the same question with Supermetrics. In this case especially, how complicated is your marketing? If your marketing is mostly Google and Facebook only, people can get going really, really quickly. If you're a huge marketer in many countries with multiple agencies and dozens of ad channels, it will take a little bit more time. So it obviously depends. But in the simplest case, a handful of ad channels in a single country, I'd imagine that somebody could be up and going quickly.

Joy:
Right. To that point, Jack, in your experience, what kind of the shortest or the fastest project that you ever take on at Supermetrics when setting up data away? What's the longest, just to give people an estimation?

Jack:
Yeah, so that's a good question for a couple of different reasons. And one of the points I wanted to address is what kind of logic is behind your marketing setup. Is there campaign naming logic? Is there UTM tag logic? How easy will it be to tie point A to point B once that data's inside the warehouse? From a Supermetrics perspective, we try to be as point-and-click as possible and set and forget like, "Okay, I've got my transfer for Facebook Ads — it's going to keep going." And then, the only reason you would go back and edit that transfer is if there are new data source fields that are available that you wanted to include. Or maybe if you're following the agency model, even brands with different accounts, if you've got new ad accounts, new things like that, they'll go through and add that data. But essentially, from a Supermetrics implementation perspective, we try to be pretty set and forget.

When it comes to project timelines, that's a fun one because you could go anywhere from five days with a very focused team of people with a very clear end goal, and that gets accomplished by Friday at 5:00.

And then you have some that to get the data in, build the data model and push those end views to the visualization layer can take upwards of three or four months, depending, again, on Jim's point on the complexity of what you have going on if you've got multiple ad agencies over a dozen different channels, that adds complexity to the actual project.

With the campaign naming conventions, UTM tag conventions, and things like this, things can be made a lot easier. But then I'd say, from implementing a Supermetrics or full-blown data warehousing product solution, I'd say you're looking anywhere from a week to get all that data in there to six to eight weeks to actually getting in, analyzing the data, making some sort of it and then pushing that to a visualization tool. That's the timeline we use internally.

Joy:
So basically, I'd say anywhere from a week to four months, three to four months. And then David has a very relevant question about the solution as well. "Is this something small businesses can use, or is it more of an enterprise thing?" I think he's referring, again, to the Snowflake solution.

Jim:
So Snowflake is obviously used by some of the largest enterprises in the world, but we also have a lot of small customers that are using an on-demand account and paying as they go on their credit card. So we probably, like Supermetrics, have customers at all levels.

Jack:
To add to that, Snowflake, as well as Supermetrics, are built to scale. So if you're a smaller type of business, you won't be paying enterprise prices for the Snowflake solution. When I was going through some of the educational material on Snowflake, you could start with an extra small warehouse or bump that up to a six X warehouse. It can be scalable up and down to the required needs and really what you're trying to do with the data. So it can start very reasonably, and then as you expand your analytics capabilities and the insights you're trying to drive, it grows with you.

Joy:
Yeah, excellent. Okay. Miracle, if I pronounce your name correctly, "Can you give a super high-level overview of Supermetric and explain the benefit in a nutshell to someone new to the subject?" Okay, Jack, it's your time to shine.

Jack:
That sounds good. I wasn't going to make Jim answer that one, but essentially, I like to think about Supermetrics in its current form, and this gets into a whole other topic about what Supermetrics will look like in 12 months. Still, I think about Supermetrics in its current form: we're the best people on the planet at getting your data from point A, being the different data sources, to point B, being your destination Snowflake in this case. We build and maintain those API connections.

Anytime an update is made to Meta's API or another API or something like that, we have a team of engineers who will ensure that that connector is up-to-date in the most recent form possible. So again, we're like a fire hose in my mind between the data sources and the destination, whether that destination be Snowflake or whether that destination is a, Looker Studio or a visualization tool like that. There was a second part to that question that I didn't address.

Joy:
Yeah. And then the benefit in a nutshell.

Jack:
Yeah. So in my previous life, I was a copy, paste Excel junkie. And I spent 20 to 30 hours a week going into different data source platforms, Google Ad Manager, Google Analytics, Facebook, all these different places, and putting together KPIs in Microsoft Excel. It was literally my full-time job to go do this. And I figured, "This is driving me insane — how can I make this an easier process?" And Supermetrics essentially automates that process.

So if you've got an employee or a team of employees who are spending dozens of hours, if not hundreds of hours, manually going and scraping all this data through, exporting all this data into CSVs or raw text files, and then uploading them into either a KPI sheet or an on-premises warehouse, it's very, very time-consuming. So the time savings and the efficiency that you can find with a Supermetrics product, a product that will automate that process for you, leaves a lot more hours for actually analyzing that data, drawing insights from that data, building out the dashboards and the views, rather than just trying to ingest it into the destination. It shifts the resource need from getting the data to what we do with this data.

Joy:
Yeah, that's a very good answer, and I want to add that not only Snowflake or other data warehouses, but we do push data into, for example, Google Sheets, Excel, or Looker Studio. And that's actually led me to the next question from Kama here. I'm sorry for the late response; I saw your question. So Kama asked if we have any Microsoft Excel templates connected by the plugin. We do have those. Maybe our team, Ha, from our marketing ops team, can drop a link there to our template gallery. We have Excel, Looker Studio, or Data Studio templates, and then Google Sheets templates. So if you're new, try them out. I think a template is the easiest way to get started. But then, if you want to customize your dashboard or build one, then just reach out to our team, and we're happy to help you with that. Okay.

Let's see, is there any other question here? I think, yeah, Jim's been amazing in answering all of these questions. Is there anything else we missed?

David has a follow-up question for you, Jack. He's asking you specifically for this one. "Jack, if I have a small business, can I just save my data on my server, or do I need a cloud server like Snowflake?"

Jack:
That's a fair question. Many clients that we talk to on a day-to-day basis use our SFTP product or our API product just to ingest data onto an on-premises data warehouse or an on-premises database somewhere. I will say that the risk associated with having something on-prem vs. something in a cloud solution is a little bit greater. Generally speaking, with Snowflake, there are backups upon backups, and that data's always going to be there. Depending on how you're implementing your own in-house server, things like this are a bit trickier. It can take a little bit more to implement an in-house warehouse than it can be to say, "Hey, I've got Snowflake right here. I'm going to start an extra small warehouse. It's going to cost me X number of dollars per month," and move from there.

Joy:
Yes. Thank you. We have time for one more question from Annie Kath. "Are you saying that we can tie a user who came six months ago and not purchased to the same user over time using Snowflake?" So it's about a customer journey. If you visited the store and left without purchase, but then you come back and purchase something, if you can't match, stitch that data together.

Jim:
Yeah, that's a complicated question that would probably take some time to do justice to, but we do see a lot of folks doing customer journey orchestration inside Snowflake. It's going to come down to what kinds of identity you have and how persistent that identity is. So if the user's logged in and you have their email address, you're probably going to have more luck than if it's made. We do have native identity inside Snowflake, so we have partners like Live Ramp and Experian who can do identity resolution inside Snowflake. That gives you a better shot. I'd say that having that person's identity persists for six months. But unfortunately, there are no guarantees.

Joy:
So I want to wrap it up and thank you so much for anyone who joined us today. And Jim and Jack, thank you so much for joining me today. It was really fun to do this.

Jim:
Yeah, thank you, Joy and Jack, this is great.

Jack:
Yeah, it's been great to work with you.