Maximizing Return on Ad Spend Through Reduced Latency and Silo Removal
Data latency and data silos are two of the biggest culprits that negatively impact return on ad spend. Both can lead to delays in gaining real-time insights, resulting in inefficient campaign optimization, inaccurate analysis, poor targeting and ineffective budget allocation.
So what can digital analytics professionals do to reduce data latency and eliminate data silos?
Anthony and the team from Snowflake and Snowplow explore Digital Virgo’s story in more detail and explain how you too can replicate this success.
Cynthia Ramasaran (00:00:04) – Welcome, everyone, and thank you for joining today’s webinar, Maximizing Return on Ad Spend Through Reduced Latency and Silo Removal, presented by Snowplow and Martech. Before we begin, if you have any audio issues, click the audio icon on your screen to enable your audio. And if you have any viewing issues, you can use the Q&A section to communicate with us at any time. You can also send questions and comments directly to our speakers by using the Q&A. Now, let’s get to our presentation.
Cynthia Ramasaran (00:00:37) – Joining us to moderate today’s session is Mike Maloney, Chief Data Officer, Field North America at Snowplow. Welcome, Mike. I will turn things over to you.
Mike Maloney (00:00:48) – Thanks so much, Cynthia. So like Cynthia said, maximizing return on ad spend (ROAS) through reduced latency and silo removal. So just a little housekeeping, and I’m not going to read the slide word for word. We really encourage you to ask questions during the webinar. So pop those in the Q&A function, and we’ll do our best to get to as many of them as possible. At the end, we’ll be sharing some useful resources. So keep an eye out for all those in the handouts link. And finally, this session is being recorded.
Mike Maloney (00:01:19) – So if you have to leave early for any reason, recording will be available right after the webinar. So without further ado, a little introduction of our panelists. Today, I’m proud to be joined by David Wells. He’s an industry principal of media, entertainment, and advertising at Snowflake. David is a data and identity executive with over 15 years of success in data strategy, data sales, and global management of partnerships across agency holding companies, programmatic platforms, and publishers. Welcome, David. Thank you.
Mike Maloney (00:01:55) – We have Jordan Peck, principal data strategist at Snowplow. Jordan has been working in data and digital analytics and digital marketing for over 10 years. Prior to joining Snowplow, he worked with clients of all shapes and sizes across some of the biggest digital agencies in the north of England, was a web analytics consultant, headed up a data team, and was also a Snowplow user out in the wild before we converted him. He’s seen data and analytics set up of every description.
Mike Maloney (00:02:27) – And now he uses his experience in the industry to help our customers with the most interesting use cases and most complex setups to get the most value from Snowplow possible. He also helps with them strategy-wise, what their data journey looks like into the future. Welcome, Jordan. Hi, everyone. Great to be here. And lastly, Antony Gia Astazio, head of analytics at Digital Virgo. Antony started with a master’s degree at Chernobyl Management School in marketing and is passionate about technology.
Mike Maloney (00:03:03) – In 2011, he joined Digital Virgo and got a lot of different marketing positions before becoming head of the analytics department. With his team, he deployed a global analytics stack across 50 countries and over 5,000 web properties, first on Google Analytics 360 and then on Snowplow BDP. Now, we are a global data analytics team. And Antony is using Fivetran, Looker, Snowflake, DBT, and Snowplow to keep on with the mission to democratize and leverage data and make it available in a self-service platform across all organizations. Welcome, Antony. Hello.
Mike Maloney (00:03:48) – Thank you very much. And last but not least, I am Mike Maloney, chief data officer field North America here at Snowplow. I’ve had a number of different executive IT roles, analytics and operational positions at a number of Fortune 500 companies. When not working, I like to spend time with my family and also teaching and writing at the University of New Hampshire. So like I said, this will be a very conversational style today. We’re going to ask some questions, do a little bit of a roundtable.
Mike Maloney (00:04:24) – And again, I encourage questions from you guys as we go in the audience. So starting with Antony, what was the impetus for Digital Virgo to make changes in its MarTech solution stacks? Specifically, what were the business motivators for you to make changes?
Anthony Gianastasio (00:04:44) – So, in fact, we were using our previous analytics stack for the last six years and, in fact, we were, at the beginning, very happy about it because it allowed us to kickstart, I would say, because we started with nothing and we started with a base analytics stack. But this data was still stuck in the tools in which we were collecting data.
Anthony Gianastasio (00:05:10) – And little by little, when we started to bring this setup across all our countries, organizations, companies, this setup started to be a problem because we had quite a lot of difficulty to ensure governance in terms of tracking design, especially. It was… and define rules across all this setup. And we were also lacking some. But this, we only found it way before, like years after starting, because we increased in terms of maturity, we increased on the way we were using data in the company.
Anthony Gianastasio (00:06:05) – So little by little, we started to find all these weaknesses in our setup. And therefore, basically, I was looking for a tool that would allow me to enforce this common language across the organization. And also, especially, and this is a very important topic, is how do you make sure that the developer who tracks the events and the marketing team who is asking the event and the data team who is building the report on these events is aligned on all the things that we want to achieve.
Anthony Gianastasio (00:06:39) – And this was extremely difficult in our previous setup because sometimes you may have developers implementing events without saying nothing to anyone. Sometimes you had marketing people saying, okay, I want to build a report on this event and this analysis, and it was very difficult to align all these people together. So we were looking for a platform that would allow us to align all these stakeholders in fact. So this was one of our key motivators.
Anthony Gianastasio (00:07:20) – The other motivator was also about the silo, because for years, we were using marketing analytics team was using GA, with great success, but we were like in our own silo, in GCP, in BigQuery. So we were doing our stuff. So this was great. But on the other side, you had the data team managing all the company’s data with the payment data, with the advertising data.
Anthony Gianastasio (00:07:48) – And there was some issues here because it was very difficult to bring all this data in one silo and put it in Snowflake, for example, in order to have the full picture, the full vision of the user journey in fact. So it was also for us a key motivator to have a native data stream going straight to Snowflake in order to centralize and consolidate all our data into one place. And this was one of the key motivators of switching to Snowflow BDP in our company. So we have a lot of motivators.
Anthony Gianastasio (00:08:33) – And I could also say, and what is funny is that it came way before because we had all these issues in terms of regulation, compliancy in Europe, especially. But in fact, we had already taken the decision to move to Snowflow way before all this happened, in fact. So this was another, let’s say, another point that said, OK, we made probably a good choice to do that. Yeah.
Mike Maloney (00:09:00) – So if I can play that back a little bit, it sounds like previously you were using a combination of GA, BigQuery, but you also had different data teams advertising other data teams who were concentrated on Snowflake. So you had a lot of silos between business, between even IT departments and business departments. So probably governance lineage was a little bit of a problem, right? Because you, as we mentioned in the beginning, 5,000 web properties, 50 countries.
Anthony Gianastasio (00:09:40) – Yeah.
Mike Maloney (00:09:41) – Getting quite a bit. Yeah. Is that something, David, Jordan, that you’ve seen other companies struggle with?
Jordan Peck (00:09:52) – Yes, definitely. At a previous organization, I worked with a global brand who are headquartered in France, who worked across 52 geographic regions, so 52 different markets. Each had their own Google Analytics setup connected to their own Google Ads or AdWords setups.
Jordan Peck (00:10:16) – The Centralized Data team tried to get a single Google Analytics property across all of those different regional websites with limited success at start because you have all of the different specific digital marketing teams and web development teams all set up individually and working on their own priorities. And as Anthony would mention, you would come across, sometimes the event was like sign underscore up, sometimes it was sign space. And then the pain would But then you’d end up with multiple rows in a report, which are actually the same thing.
Jordan Peck (00:10:54) – And then you would have to go and manually try and clean that up. And you’re just continually building up this technical debt over time, which means when you come to do an analysis, you have to remember to do, oh, yeah, I need to merge sign underscore up and sign space up. Every single time you come to do an analysis. And it becomes really, really difficult when you’re doing it across that many teams. And in Google Analytics, there is no connection in the universal anyway, between different accounts.
Jordan Peck (00:11:24) – You can’t set up different event rules between different accounts or custom dimensions between different accounts. So yes, I have come across that kind of issue before.
David Wells (00:11:34) – And I would iterate the same. I mean, the idea around Snowflake is that you’re able to centralize your data in one place and various teams within your org or partners are able to access that data with rules and governance. It’s Snowflake is all about that shared data architecture. There’s an ongoing joke that we want to put on a T-shirt at Snowflake, which is silo intolerant. We really like this idea of Snowflake as a central repository to land data, any type of data, whether it’s own and operated, own media, analytic data, whatever you have.
David Wells (00:12:16) – No, it makes a lot of sense.
Jordan Peck (00:12:18) – I suppose another thing I would maybe mention as well is, so Anthony, you mentioned that the marketing analytics team were focused in the BigQuery, the Google ecosystem. And then the other departments or teams in the business were already using Snowflake, I think. Did you try or did you not bother coming up with some sort of ETL process to get the data from BigQuery into Snowflake or did you just operate? So these silos exist. It’s too hard to fix. We’ll just live with these for now.
Anthony Gianastasio (00:12:56) – Yeah, in fact, for years we stayed like this, but at some point we needed to consolidate this behavioral data, this web data generated by GA and all our other data that we… Snowflake. So what we had to do, of course, is to build some ETL pipelines to copy this data on a nightly mode inside to bring this from BigQuery to Snowflake and feed this table into Snowflake. But the issue with that is that at the end you had one load per day. So you had to wait the other day to see the data inside Snowflake and be able to do your analysis.
Anthony Gianastasio (00:13:35) – So this was a key weakness of this setup. And also the fact that, of course, it requires some engineering resources to build the pipeline. When a new country is being launched, you have to say, OK, I have a new table in BigQuery, so I have to replicate this data into Snowflake. So you have all these kind of issues that arise when you are in the middle of your operations. And this is the kind of thing that can, yeah, you are losing time at the end.
David Wells (00:14:06) – Well, I was going to ask, was there, you know, you said you operated that way for years. Was there like a seminal moment where you remembered, like, OK, this is not the answer. The answer is Snowplow on Snowflake. Just for other people who are potentially considering the same thing in the audience who are maybe deliberating over that moment or that switch. Like, is there any advice? I mean, you talked about it a little bit, about why it was so powerful, that shift. But was there a seminal moment where you’re like, this is definitely…
Anthony Gianastasio (00:14:37) – I wouldn’t say there was a key moment that say, OK, now we have to switch. But it’s more like an accumulation of…
Anthony Gianastasio (00:15:26) – We need to know what happens here when we switch on the campaign in the next 15 minutes. I want to see in terms of conversion rate where we are and how the companies behave on this kind of thing. So little by little we saw all these weaknesses and at some point we had to launch because this project was quite a big project for us because we had to migrate a lot of properties. So it was not an easy task to be honest because it’s quite big and a lot of teams were involved in these migrations, IT teams, front-end teams, a lot of people were involved.
Anthony Gianastasio (00:16:08) – So yeah, that’s why at some point it was one year and a half ago we took the decision.
Mike Maloney (00:16:16) – So if I can play back a little of that discussion. So, you know, a lot of our clients come to us with questions around data latency and that’s, you know, the title of this is reducing ad spend or increasing return on ad spend and a lot of that has to do with reducing latency. I think you’ve told us something about a 90% reduction in latency. So, and when you’re talking about matches, we’re talking about you guys had some tremendous success at the World Cup 2022.
Anthony Gianastasio (00:16:51) – Yeah, exactly.
Mike Maloney (00:16:53) – Kind of we’re going from this overnight monolithic process to, you know, how fast are you loading data now and being able to react to it, if that makes any sense. And you don’t have to be exact.
Anthony Gianastasio (00:17:09) – Yeah, that’s exactly. In fact, what we had until now is that we were able to see the amount of conversions we were generating every five minutes, but we were not able to see, for example, the cost in Google Ads because the Google Ads data is not updated at this frequency. You know, you have to wait at least one hour or maybe a little bit more to get the clicks and to know your conversion rate, your real spend during this hour, you know, on which you were live on the campaign. So, we were lacking this information.
Anthony Gianastasio (00:17:38) – So, and it’s a key point because our team needs to know whether or not we are good in terms of performance in order to, during this point of time where you don’t have… are good or not good in terms of performance. And with this NoPro data coming in real time, we were able to know that, okay, on the last five minutes or the last 10 minutes, I got this amount of sessions and I know that this amount of sessions were costing like 10 cents per session and I know my conversion rate.
Anthony Gianastasio (00:18:12) – So, I can take a decision to, okay, either I have to decrease my investment or I can increase my investment. So, this was like, for us, it was like the most tricky edge case, in fact, because this is the point of time where you don’t have this information available and you have to rely on your own data collection process to know what happens at this level of time. So, and yes, NoPro and all the new stack was able to answer this kind of requirements, in fact.
Mike Maloney (00:18:44) – Yeah, and I think that’s really key, right? That time where you don’t know what’s happening with your spend, and, you know, I think we’re all under a tremendous amount of pressure with the way the world’s changed. Everything’s almost digital ads, not everything, but pretty close. That’s how most of the people interact with the world now. Jordan and David, like David earlier, you brought up breaking down silos, right? And having everyone work off of kind of this more accurate, low latency data. Have you seen that a lot of other places, right?
Mike Maloney (00:19:23) – Like Snowflake’s a pretty great place to decentralize and break down those silos.
David Wells (00:19:28) – Yeah, I mean, we think that’s part of the differentiation of Snowflake, you know, the idea that we work atop, you know, with all the public cloud providers in all the regions throughout the world, the ability to harmonize and land data in Snowflake, using a partner like Snowflow, Snowplow to pull that data and land it in Snowflake, and then to action on it in, you know, near real time, to basically look at a segment and say, hey, I’m going to increase or decrease my ad spend based on that.
David Wells (00:19:57) – You know, when we talk about AdTech and Martech and the relation, you know, they have operated in silos in many orgs and in the industry separately. You know, they’ve sort of their own silos and a lot of those teams didn’t talk. We’re now seeing a convergence of AdTech and Martech. We think a lot of that is happening in Snowflake. And so we’re super appreciative of partners like, you know, Digital Virgo, who have this vision, have taken these steps, as well as partners like Snowplow that are enabling it on Snowflake.
David Wells (00:20:30) – You know, part of the value of Snowflake, just at a very high level, is not only the technology and the shared data architecture and the separation of storage and compute, it’s also the partner ecosystem that have products and services that unlock value for large marketers, Snowplow included. And so what you’re getting with Snowflake is not only this next level cloud technology or cloud data platform, you’re also getting the ecosystem of partners like Snowplow that just add value and unlock more value of Snowflake, if that makes sense.
Mike Maloney (00:21:08) – No, it makes sense. Jordan, I just want to get you here in here for a second. So, you know, you’ve been doing kind of digital marketing advertising for 10 plus years. You’ve probably experienced some of those endless or sleepless times where, you know, you’re not entirely sure what you’re doing and paying off. How has this changed, you know, this dynamic of kind of latency and speed and the pressure on advertisers?
Jordan Peck (00:21:41) – Well, it’s interesting because, and this might be a relatively controversial thing to say, but not in my experience, majority of people use cases that people say are real time and not necessarily real time requirements.
Jordan Peck (00:21:55) – But there are more and more and more that are and ones that, you know, Anthony has been describing for Digital Virgo, going through a sporting event where you have such a small amount of time to ensure that your spend is, especially as something as competitive as the World Cup, when there’s so many competing bids and brands trying to outbid each other to get the attention and, you know, users’ attention these days is so much shorter than it ever was.
Jordan Peck (00:22:26) – In some use cases like that, which can really make or break, you know, your performance, you only have, you know, World Cup, you’ve only got four, six weeks. You’re going up against every other advertiser on the planet. You need to know within minutes if your ad spend is going effectively and be able to change course if it isn’t.
Jordan Peck (00:22:48) – On top of that, not only are more use cases genuinely real time than they ever were before, that’s not to say all of them, there are still like BI use cases, which are still plenty good with nightly ETL jobs in my experience, but with more and more real time, genuine real time use cases, it’s more of an enabler because the technology was never really there.
Jordan Peck (00:23:10) – And if it was, it was extremely expensive or extremely complicated to run or you needed to hire very, very experienced and very expensive data and Java engineers, data engineers, things like that in order to actually put something, you know, to be able to put it in the hands of people who could make use of it. In this case, you know, the advertisers and the marketers who are running those campaigns. Nowadays, there’s lots of technologies now that are much easier to use than they ever were.
Jordan Peck (00:23:39) – Snowplow is obviously one of them, but there are lots of others in other spaces. And that kind of general, our co-founder, Alex Dean, I saw him on a webinar sometime last year. He said, I’ve been saying it’s the year of real time for the last 10 years, it feels. But I think we’re genuinely getting to the point of where real time technology is easier to set up.
Jordan Peck (00:24:02) – It isn’t going to cost the earth and it’s going to put, it’s going to enable activities and enable use cases or enable individuals or teams to do things that really will, you know, deliver return on investment or return on ad spend far more so than they ever were before. A, because there’s genuine needs now because people act and react and respond in real time. And the technology is actually there and usable and affordable to do so.
Mike Maloney (00:24:34) – Yeah, no, I think that’s a great point. Right. So.
Mike Maloney (00:24:39) – over the last five, 10 years, I think we’ve moved to this expectation that the companies you work with get you, right? And that takes personalization, which is built on segmentation, and that’s a lot of data. So Anthony, I kind of want to come back to you just to kind of walk through, you’re working with another partner of ours, HiTouch.
Mike Maloney (00:25:03) – So kind of that snowplow, snowflake, HiTouch centralized data setup, and how has that helped benefits of improving monitoring, detecting issues, how has that helped with activations, as we like to call them going forward?
Anthony Gianastasio (00:25:22) – So yeah, this is in fact the last addition to our stack, and it’s very recent, it’s not yet fully deployed, and we are still learning in terms of how to use it and how to leverage this new kind of tool. But the idea here is really to say, okay, you have all this data already in Snowflake, you have your event data, you have your advertising data, you have your transactional data, and then you want to take decision.
Anthony Gianastasio (00:25:51) – Of course, you have a reporting, you can analyze lifetime value, you can analyze marketing spend, that’s great, but now you want to take actions, you want to bring this information, and you want to put it in your campaigns, you want to put it live, you want to alter the Google Ads bidding algorithm, you want to use something like this.
Anthony Gianastasio (00:26:13) – And in order to do that, this is where iTouch enters into the stack, and this is where, thanks to iTouch, we can create, for example, an importation of data for a lifetime value, and bring this into the ad system, and this is very, very rich, because you get new insights that you were not able to have before, in fact. So this kind of thing is, yeah, really putting all your data collection analysis into operational, you know, and that’s changing a lot of things, and really help bridge the gap between the analysis and the action.
Mike Maloney (00:27:00) – That’s interesting. So, Jordan, you know, again, lots of experience, and we talked about kind of the near-time, real-time, quote-unquote, use cases. How have you seen kind of the change in activations, and that, you know, with the onset of all these new activation technologies like iTouch, how has that need affected people?
Jordan Peck (00:27:27) – Yeah, I mean, for a long time, well, it’s kind of like what I was saying before, like, for a long time, it didn’t matter if you needed to react in seconds, just because you couldn’t. The services, the tech vendors weren’t able to allow you to do it anyway. And people can churn on a single click nowadays.
Jordan Peck (00:27:48) – People can churn on a single, you know, ad that they see that they don’t like, or, you know, a YouTube pre-roll or something, and if you get that wrong, and you, or you target your ads incorrectly, or you don’t update your, you know, your, what’s not the term, block lists, you know, like, so you don’t keep showing the same ads to the same people over and over again, even after they’ve bought.
Jordan Peck (00:28:09) – If you don’t keep those updated in close to real-time, then, especially when you’re running, like, digital Virgo, where I have several hundreds or thousands of transactions every few minutes, then, actually, because, like, back in the day, people didn’t really mind, and now people do, and, you know, the unsubscribe rates or emails are going up. There are applications to help you do this. Ad blocker usage is going up, and it is extremely easy, and more likely for people to abandon your brand so fast and so quickly nowadays.
Jordan Peck (00:28:43) – So, for certain use cases, you can probably get away with daily, maybe even, like, a few times a day, but for certain types of use cases, for certain sort of environments, then it really is important to get real-time things right. Like, if you’re in sport, or if you’re doing things around events, or television, or, you know, you’re coordinating with TV shows, it’s something you have, you cannot be, you know, you cannot be doing with hourly batch jobs.
Jordan Peck (00:29:12) – Like, as good as tools like, you know, Fivetran and DBT and Hightouch are, they are inherently batch tools. You need to be reacting in real-time as much as possible for those types of use cases.
Mike Maloney (00:29:26) – Yeah.
Mike Maloney (00:29:27) – And David, I don’t know if you have anything to add with your experience in media, especially.
David Wells (00:29:31) – Yeah, I mean, for sure. I mean, we’ve seen, like, the same way that we’ve seen adoption of SaaS analytics tools like Snowplow, we’ve seen the adoption of reverse ETL connected app partners like Hightouch and others. And, you know, we were talking about near real time and the importance of the temporal nature of your, as a marketer, communication with an individual consumer. You know, I used to work in an identity provider.
David Wells (00:30:01) – I still work with them today at Snowflake. They talked a lot about people-based marketing. We’ve seen more and more people, and I can’t remember who originated or coined the term, talk about moment-based marketing and having a really strong SaaS analytics tool that you’re able to action on in a moment, right? We used to talk about David or his household and, you know, marketing to David as an individual in the various iterations of David and how important that was.
David Wells (00:30:33) – But more and more, we’re seeing marketers talk about, well, you need to actually capture David not only as an individual, but in that appropriate moment. You know, between the hours of nine and five, he’s working. He’s probably not receptive to, you know, a marketing message or any type of communication. It’s understanding that moment when he is available and he is a captive audience and sort of targeting him. That’s the difference, I think Jordan was alluding to it, between churn and retention and seeing value in that communication.
Mike Maloney (00:31:04) – Yeah.
Jordan Peck (00:31:05) – I think that also touches on, you know, some of the things about, like, the technologies that I was saying. So, like, now, you know, the advertising platforms do allow you to, like, refresh ads, like, in real time, which they never really allowed you to do before. They’re easy to use streaming platforms available nowadays. And I guess as well, like, Anthony, so you were, David was mentioning, you know, that kind of contextual information about the users, like this person’s at work now or this person’s in their commute or something.
Jordan Peck (00:31:39) – That’s the kind of contextual data that you may have on, like, your CRM or, you know, depending on your business, which, in your example, was living in Snowflake, whilst all your behavioral data was living in BigQuery. Now that everything’s in the same silo, sorry, Mike, I’m taking your role here. Now, all of your data is in Snowflake.
Jordan Peck (00:32:01) – Do you find it easier to do that dynamic, you know, that segmentation work and generate cohorts and user segments based on behavioral data connected with other data as well, like CRM, payment, demographic, that kind of stuff?
Anthony Gianastasio (00:32:18) – Yes, indeed, because we, in fact, we were already doing this kind of work before. But we have all this behavioral data available. So now we are able to build this kind of cohorts, like user subscribe and use the service,
Anthony Gianastasio (00:32:41) – We know how to do that.
Anthony Gianastasio (00:32:43) – And that’s why we have a lot of ideas for the future and ideas to leverage the high touch and how we can use it in the ad system in order to improve… overall our usage rate, you know, on the way people interact with our services. All this kind of project will rely on all these tools… the last year. But at the moment, we are still in a very early stage at this stage. So we are using for now just like basic importation of conversions or things like this.
Anthony Gianastasio (00:33:21) – But we haven’t invested yet on all this advanced activation on segment or specific segment of user. But of course, all the technology is there to do it now.
Jordan Peck (00:33:33) – It’s interesting what you mentioned as well about the identification strategy. And David mentioned that he was in the identity resolution space. That’s really, really things that are really going to go into the next. In the digital advertising space, I think, you know, this idea of like advertising for on Google Ads advertising for people and audiences, not for search terms, which was what you did historically for the last 20 years, you actually now target individuals and audiences rather than search terms.
Jordan Peck (00:34:07) – or something.
Anthony Gianastasio (00:34:08) – And I think this is what is interesting here with the SnowPro project is that when you start on SnowPro, you say, okay, I want to build my tracking system. Okay, great. But once you do that, you do more than just rebuilding your tracking system. You implement the basis, the foundations of more even use cases you have not even thought about. And I think that’s very interesting because all the things we are talking about today, maybe two years ago when I started on SnowPro, I did not even think we would do that.
Anthony Gianastasio (00:34:40) – So yeah, I think these capabilities here are built together in order to support very long-term goals and strategy. And I think that’s very important to mention.
Mike Maloney (00:34:54) – That brings me to kind of a key question, Anthony. So how long, I think one key is like you just spoke about the ability to implement new ideas and research now that you have something like Data Cloud and premium behavioral data going in to Snowflake with all the other data, you can kind of mix and match and make new recipes and try new models, do A, B testing, but also do like C, D, E at the same time with your models that you’re segmenting.
Mike Maloney (00:35:29) – But like at the end of the day, what was like the time to value from initial, we’ve decided to change the way we’re going to do things, we’ve made our decision, like what was the implementation like for Digital Virgo?
Anthony Gianastasio (00:35:44) – So the implementation of the migration, I would say it took around six months to do the core work on the migration. If you take the other actions required, like some bug fixes, it took a bit more time. But in six months, we were ready. Between the moment where we started with Snowplow, the pipeline was ready in three days. But then we had to work on networking, we had to work on data modeling, we had to work on tracking implementation. So all this, yeah, take approximately six months on our side. And given a very big size of…
Mike Maloney (00:36:33) – 15 countries.
Anthony Gianastasio (00:36:34) – Yeah, exactly.
Mike Maloney (00:36:34) – 1000 websites.
Anthony Gianastasio (00:36:36) – Exactly.
Anthony Gianastasio (00:36:37) – If you have a single website, or even a couple of websites for your brand, I think in one month, you can be ready.
Anthony Gianastasio (00:36:46) – In fact, it’s
Anthony Gianastasio (00:36:53) – a lot of them, we had more than 5000 domains worldwide. So it’s a lot of web properties. So we had a lot of works to do in order to fulfill this migration. But yeah, six months was more or less the time we needed.
Mike Maloney (00:37:12) – Yeah, considering what you had to do and the global reach of it, that’s pretty good.
Anthony Gianastasio (00:37:17) – Yeah.
Anthony Gianastasio (00:37:18) – And I think the most, let’s say, pain point at the beginning was the data modeling stage, because it’s something that we never did in the team before. Of course, when you use a package analytics tool, you get the data that the tool is giving to you, and that’s it. And you don’t start to say, okay, I want to define my session, I want to backfill my session data based on the late arriving data in the events table, these kind of questions.
Anthony Gianastasio (00:37:46) – And I would say it was the main pain point at the beginning for us, is to say, okay, how I move from, let’s say, pre-packaged tool with a given data set, which is ready for analysis, versus having this big amount of events arriving in Snowflake, and then I have to build… my page view tables. So this was, let’s say, I would say the most pain point for the team, because we had to learn a lot of concepts and guidelines
Anthony Gianastasio (00:38:18) – But at the end, it was super interesting, because you are putting your hands into the matrix, you have to learn, you have to know what you are counting, what is my session, where it starts, where it stops. And this, at the end, is very important, because you are defining your matrix.
Anthony Gianastasio (00:38:41) – And I think it was a very funny part.
Jordan Peck (00:38:48) – I think it’s quite, so defining metrics is quite a common thing in, say, traditional data warehousing, defining your dimensions and metrics. I think a lot of people who come from Google Analytics find it a, you know, incredibly, you know, a wonderful opportunity, but also potentially a little bit, like, overwhelming that they can finally now define everything, rather than have it dictated to them. So, but it is an enormous opportunity that people, once they realize the freedom that provides.
Anthony Gianastasio (00:39:17) – Yeah.
Anthony Gianastasio (00:39:18) – At the end now, even GA, because what I’m talking about is more the previous version of Google Analytics.
Anthony Gianastasio (00:39:26) – It’s not the new one.
Anthony Gianastasio (00:39:27) – In fact, the new one is more working in an event-based data model. At the end, you start to have the same kind of issues.
Mike Maloney (00:39:34) – Interesting.
Anthony Gianastasio (00:39:35) – Yeah.
Mike Maloney (00:39:36) – I think I got two more questions and I know we have a little bit more time. You mentioned this shift in mindset and moving away from the technology for a second. How has Digital Virgo changed how they work internally after putting this new data solution in place? Has it changed or impacted how your teams around the globe work together? I know you mentioned breaking down silos and things like that.
Anthony Gianastasio (00:40:07) – Yeah. Of course, the Snowpro project was not the only motivator Because in fact, what we decided to do one year ago is to build a single global data team. We no longer had the data analytics team working with the GA data in BigQuery and the other team working with all the other data inside Snowflake. We decided to say, okay, we are a global data team and we built a global team that is working with all those tools available.
Anthony Gianastasio (00:40:45) – We have a department in the team which is called the Tracking and Analytics Department, which is my department, where we define tracking, where we define the JSON schema, the data we want to collect and we want to implement on our different properties. In the same time, we are working with the platform team, with the BI team in order to make sure that all this data is fully utilized and is leveraged as it should be in fact. This is a key change because we are now just a single team.
Anthony Gianastasio (00:41:24) – Yeah, it’s much better in terms of collaboration, in terms of language, in terms of sharing all this knowledge. It’s much, much better in fact.
Mike Maloney (00:41:35) – Interesting. David, you talked about how Snowflake wants to make a T-shirt. I’d probably buy it, breaking down silos. In media, you talked a little bit about this earlier, and we had a question come in. Have you seen anything about TV campaigns, linear, CTV, OOT? Yeah, absolutely.
David Wells (00:42:04) – When you talk about the conversions of AdTech and MarTech, and having a single central repository where you’re managing owned media and paid media, CTV is obviously part of that paid media opportunity. How do you manage that owned media presence and information, and then activate accordingly across CTV or plan your data-driven linear spots? It’s something we’re seeing.
David Wells (00:42:35) – I mean, we partner with a lot of very large streamers, and so that’s a huge opportunity as we move more from linear to more streaming capabilities, and there are more addressable audiences in TV as a channel. More and more of our marketers in these lean times where they want to target more effectively and do it in a privacy-compliant way and do it where they’re maintaining a relationship with customers, they’re leveraging more of that owned media analytics with paid media with CTV as a destination, for sure.
David Wells (00:43:17) – Yeah, and then it’s not only, hey, who’s gone to my website and how do I deliver them a message at the right moment across CTV, but what’s the impact of that CTV campaign then on additional traffic to my website? That’s ultimately what we want to do is power that convergence in a privacy-compliant way where it all lives together in Snowflake, and it’s providing value to the marketer and to the consumer in that that marketer is only communicating with them with relevant messaging in a time or moment when they want to get that message.
Mike Maloney (00:43:59) – No, that makes sense. Anthony, I know we talked about this a little bit, but something David just said and I think you mentioned earlier.
Mike Maloney (00:44:13) – across 50 jurisdictions, different countries, different laws, in the US, I don’t know, we got like six different states with six different laws and I don’t even know how many bills on the floor. How big of a point was privacy for you guys, right? Cause like we’ve kind of beat around the bush, like we’re really talking about first party data ownership. I like to talk about as ownership, right?
Mike Maloney (00:44:39) – The trackers and the cookies are one thing, but being able to govern your data in the way you need to for your customers, it’s like big of a point, decision point. Was that for you guys?
Anthony Gianastasio (00:44:51) – Yeah, yeah, absolutely. Because today the data is hosted in our AWS infrastructure. In Europe, we have full control over it. We are able to implement any kind of data masking inside Snowflake or directly inside Snowflake. If we want to mask some informations, we… redactions of a person’s data. We have a lot of capabilities, in fact, in terms of privacy. And of course, it’s changing completely. Having all the control of this amount of information versus checking some reports in a package analytics tool, you know, it’s totally different.
Anthony Gianastasio (00:45:39) – So for us, it was also a key lever, this part, because this data is, of course, very important as we take decision on top of it. So controlling this information and being able to bring this information to the different peoples the way they need it is the key point. And it was, of course, a good motivator to move to Snowflake.
Mike Maloney (00:46:03) – Yeah, and question, I think maybe this is best for Jordan, came in, you know, what’s been your experience, both in the industry and working with some of our clients, you know, how, you know, important and hard is privacy to get right nowadays? That’s very wide open.
Jordan Peck (00:46:29) – I know. It is. I mean, there’s a couple of, well, there’s a lot of things in there, but first thing that’s really hard is keeping up with all of the regulatory updates. So that, I mean, I won’t touch on that in too much detail, but like, technically, very much when I first got in the industry, it was collect everything, figure out what to do with it later. And just, you know, when you come up with an idea for what you’ll use this data for, you’ll have a load of it ready to go ahead and do that analytics.
Jordan Peck (00:47:05) – If there’s anybody out there who still thinks that that’s the way to go about it, whether they’re in the Europe or in the US or anywhere else in the world for that matter, that really doesn’t fly. That’s really not an approach that you can take anymore. The best way to protect yourself from user and customer PII is not store it unless you absolutely have to, or redact it, anonymize it, pseudonymize it, encrypt it, hash it, whichever. Be aware of the ramifications of hashing or encrypting or truncating data. It can still be useful for analysis.
Jordan Peck (00:47:43) – So Anthony mentioned, you can actually, with Snowflake, you can hash values in real time. So you can say this particular field that’s coming in is potentially sensitive user data, hash it with SHA-256 on the way through. So it’ll be, before it even gets to Snowflake, so you don’t have to do those kind of like schedule, hash this column in SQL in Snowflake. So it does live in your warehouse in unencrypted format for some time. I’ve seen people do that before.
Jordan Peck (00:48:11) – But that means it’s still usable for joins and count distincts and things that will still allow your analysis to be useful. And Snowflow and other tools also support things around anonymous tracking or cookie-less tracking, which is great and you should do it to some extent, depending on what your legal team or your data protection team says in your business. But I think be aware of what, again, the ramifications for the usability of that data are gonna be once you’ve done it.
Jordan Peck (00:48:46) – If you say, if you think, right, we won’t collect any kind of session identifiers if the user hasn’t granted consent, great, more power to you. But please make your marketing team aware that you ain’t gonna be able to report on sessions for users who haven’t granted consent. And if all of your reports are session-based, then there’s gonna be a lot of confused dashboard users.
Jordan Peck (00:49:09) – So there’s a few things to consider like what should we do, what the regulators tell us that we should do, what’s our particular organization’s position and stance and risk tolerance pretty much. What will the technology, whichever technology I’m using like Snowflake with data masking and column and row level access controls, whether it’s Snowplow and anonymous and cookieless tracking and pseudonymization, etc. Then what are the repercussions that we put in those particular measures in place to encrypt or mask any of my data?
Jordan Peck (00:49:45) – How is that going to affect my analytics? How is that going to affect my ML models? How is that going to affect my recommendations, etc? But figure out what you’re allowed to do first, then figure out what your particular stance is, find out what your tools can do, and then figure out the downstream repercussions of it.
David Wells (00:50:03) – Yeah, I think that’s a good point. I think once upon a time, or at least still to some point today, some marketers look at regulation or legislation as highly restrictive and impacting revenue in a negative way. What we would say is, no, it’s in the interests of consumers and there’s, to put it another way, what Jordan said is, there’s a way to still use data where you’re respecting the privacy of individual consumers, and you’re still interacting with them in a relevant way. We think Snowplow and Snowflake are able to do that.
David Wells (00:50:39) – I mean, we place a huge level of importance on consumer privacy and tools that maintain relevance when you interact with those consumers yet still respect the privacy of those individuals. I think the ones that get that right, it’s not a deterrent, it’s not restrictive, it won’t impact revenue in a negative way, it’ll actually unlock more revenue. I think consumers respect that when they’re messaged the right way and relevant way.
Mike Maloney (00:51:09) – I think I saw one question and maybe this is for Anthony. How did with privacy, with the switch, and talking about increasing return on ad spend, how did the data change from accuracy standpoint or trust standpoint, switching to the new platform?
Anthony Gianastasio (00:52:06) – What we have decided to do is to implement some server-side tracking in order to collect really the first server requests when the user is clicking on the ad in order to log the sessions of the user. This implementation allowed us to make sure that we had almost a one-to-one combination between a click and a session. This is really key because you are able to switch from a data source to another. When you analyze Google Ads data, when you analyze Snowplow data, you see the same thing.
Anthony Gianastasio (00:52:38) – Because we made sure that we were aligning our session metrics with what we had on the click level. This is really great because we are able to get too much on the data. We are really staying at session level. Most of our analysis is based on session data. We are not looking at user data. We are really trying to understand what happens on these complaints. We get 1,000 clicks, now I can see 1,000 sessions. On this 1,000 session, I know that 200 sessions left and didn’t do anything. This is really key.
Anthony Gianastasio (00:53:21) – This capability of mixing the client-side tracker with back-end code, back-end SDK, like in our case was PHP. But we also did it in some Python-based back-end system. All this implementation is changing the way you see your data collection. This is very important.
Mike Maloney (00:53:50) – No, it makes a lot of sense. Better accuracy, better reporting, better actionability. You can be data-driven a little more confidently. As usual, I’ve done a great job moderating us to almost being late. I did promise everyone 30 seconds to talk about and give you a chance to talk about the next big thing in 2023 and going forward. Anthony, maybe you want to share what’s next for you guys or what you’re looking forward to technology-wise in 2023.
Anthony Gianastasio (00:54:24) – So, in fact, for this year, we are working. Basically, we have a lot of nice projects. We are working, for example, on personalization. In fact, we would like to be able to adapt our user experience based on some specific user behavior. For example, if a user had an error on the page because of whatever reasons, we could propose him something different, another product or another way to subscribe to the product, with So we are trying to work on personalizations.
Anthony Gianastasio (00:54:57) – We are also trying to invest time in multivariate testing in order to test… page and choose the best combination of elements, and all these kind of use cases will be supported by the Snopo data and, of course, the Snowflake data warehouse. So, yeah, we have a lot of exciting projects for this year.
Mike Maloney (00:55:30) – Cool, david, I’ll give you 30 seconds too. Yeah, sorry, I promised you guys more time, but no, you bet.
David Wells (00:55:38) – I mean I would say where we see the trend going is more of what we said just to iterate. It is the convergence of AdTech and MarTech and more unique solutions like Snowplow on snowflake that are unlocking value with owned media, and then technologies that complement snowplow that allow you to sort of action on that data.
David Wells (00:56:01) – We have a heavy emphasis on marketing workloads on Snowflake this year and we’ll just continue to sort of introduce more and more solutions with partners like yourself and others to sort of provide value to large marketers that are on Snowflake. So we’re super excited about that and we think we’re only at the start. Like this will move really fast and it’ll scale really quickly and, Jordan, I know you’re fighting a cold so I appreciate you being with us today.
Mike Maloney (00:56:32) – I don’t know if you have anything you’re looking forward to.
Jordan Peck (00:56:37) – So there’s a couple of things I think. The first one that I’ve been thinking about a lot quite recently is metadata management. I think old school data catalogs don’t cut it anymore because lots of data assets aren’t just tables and columns in a data warehouse anymore. From a Snowplow’s perspective, it’s your custom event schemas, it could be your DBT models. You know column level lineage is very popular these days. But what kind of assets do you have like what version of your app has? What version of tracking goes to?
Jordan Peck (00:57:08) – Which version of your snowflake tables then goes into which version of your data models and which version of your visualization tool? And for discoverability. I think that that’s a world that needs to be improved a lot and some really nice vendors doing a lot of good stuff in there. And the other thing that I think is we talked about activation and the you know companies like high touch reverse ETL tools have done a wonderful job so far, but I’d love it even more if some of the visualization tools, some of the BI tools, look.
Jordan Peck (00:57:36) – I’ve done this a little bit in the past, but I’d love to see new players like Mode and Holistics and Hex do things where you can generate analysis from an interface and then sync those you know metrics, features, cohort, segments- to an activation point from the point of consumption rather than from a dedicated tool. I’m sure all of the vendors in that kind of space, in the data activation space and the visualization space, are probably already thinking about this.
Jordan Peck (00:58:09) – I know Salesforce and Tableau are definitely thinking about this, so I’d like to see that come out a bit more this year, cool, well, so first of all, thank you to you three- Jordan, Anthony, David- on the screen- a few calls to action for you all to think about.
Mike Maloney (00:58:27) – If you want to learn more- and I definitely encourage this- register for the Spring Martech conference right. We have our customer case studies available online at Snowplow, you can request a demo. And lastly, if you want to learn more about marketing attribution and return on ad spend, you can download our 101 guide. So with that, thank you to Martech, I’ll hand it back to Cynthia and everyone. Have a great day.
Cynthia Ramasaran (00:58:57) – Thanks so much, Mike. I did pin those resources for our audience to go ahead and download it. Thank you all on this panel for this engaging presentation and if we didn’t get to your question, we will be sure to pass it along to the panel. Thanks everyone. We hope you join us again soon.
All (00:59:14) – Take care, thank you, thank you, thanks everyone.