The following is a transcript of a DMN Tech podcast featuring Velocidi CEO David Dunne. During the course of the 15 minute podcast, David and Kim Davis, the host of the podcast and editor-in-chief of DMN, covered topics such as the application of AI to marketing data, the trouble with data quality, and how brands can use data and technology to get closer to their customers. Listen to the podcast on soundcloud here, or read on below for the full text.
Kim: Hi, this is Kim Davis and welcome to another One-on-One. And with me today I have David Dunne, who is CEO of Velocidi. Welcome, David.
David: Thanks, Kim. Thanks for having me.
Kim: And in case anyone’s looking it up, obviously it’ll be on the website but it’s Velocidi with an “idi” at the end not “ity”. A New York based marketing intelligence platform.
David: Indeed indeed. The intersection between speed and digital. Velocidi.
Kim: Got it, got it. And we were just having a chat before we started the podcast and one of the reasons you’re here is to talk about the current release of the platform which has some very interesting AI related enhancements, but just to set a bit of a background for that, can you talk a bit about AI in the marketing technology space because we hear about it all the time these days. It means different things to different people.
David: It’s really a spectrum that we’re seeing now, from solutions that I would categorize as “AI-lite” where they're really just doing statistical analysis work, generating some auto outputs. To the other end of the spectrum, something like Watson which is true AI. And you know when we looked at that, we didn't’ want to come out with an “AI-lite” solution. Frankly that kind of work is too simple. It doesn’t really help a client or an agency to understand what is actually happening and ease the burden of all that manual work that they are otherwise doing.
So we went and looked at two things. First of all we want to launch a strategic product. Something that will actually be able to understand the priorities that they face the choices that matter and how to truly optimize their media spend. And at a later point, tactical AI. So tactical AI being where you then operationalize those insights, perhaps change a search buy for example. So starting with the strategic we also focused on the problem in two ways. The first was data quality. So we wanted to make sure that customers could be really confident about the quality of their data. And that’s everything from the sources, the way in which it's brought in, the data joins, unions, etc that occur within the data when it's brought in, looking for basically issues anomalies outliers within the data, and surfacing them for the analyst so that in the morning when the analyst comes in they’re literally coming into a set of insights that they spend their first part of their day on which assure them that the data is ready to be analyzed.
Kim: Ok just to make it concrete, what kinds of customers and what kinds of data are we talking about?
David: So the kinds of data typically are search data, social data, ad data, usually coming in large volumes from large brands. We work with global agencies and many large advertisers so they are brands you would know in the marketplace for whom there could be hundreds of millions of lines of data pouring in from their various ad campaigns. Particularly we’ve taken a global perspective. And so with that volume of data it's extremely difficult for analysts to - you know the human eye can’t find those kinds of issues.
So what we’ve done is built an AI quality tool that essentially combs through any kind of marketing data that you would ingest and looks for all of the known issues that might occur, and looks for anomalies. So for example, “Hey we missed some data for a couple days here, what happened?” and surfacing that so the analysts can go look at it and find out. Outliers - ”Hey you’ve got these really freaky performance things happening,” you might want to know about those, good or bad right? So then that data quality initiative that we launched in June really assures analysts that the work that they are about to do, which is deriving insights, is going to be of the highest quality possible and that they can then get those insights very rapidly and then turn those into perhaps recommendations for the media team or could be a planning team or a buying team that’s gonna implement those changes.
Kim: I think there’s a sentence I wrote yesterday or today that hasn’t been published yet which is that really bad data is almost as bad as no data at all. AI has an enormous appetite for data. You have to be able to feed it to make it work. But if you’re feeding it bad data you can really go astray can't you?
David: I totally agree. Again that whole centerpiece of our whole data quality initiative is to say that most data is not in good order when it comes in. The data ingestion is one part of the puzzle for sure, and that’s a good challenge to solve for. But data ingestion without great data transformation is useless. AI to improve that is terrific. Specially when you are dealing with massive volumes of data which, by the way, data volumes are growing all the time so we know that’s not going to change. So the more new sources that become available and the greater the volumes associated with each of those sources, this problem compounds itself every year.
And so, that was our rationale for starting off with data quality. But once you’ve got the data to the point where you’re confident, now you want to also speed up the analysis. So the second part of the platform release which we are just pushing out now, we call “Impact” which is essentially as the word would suggest, helping you to understand what’s working and what's not working, where when and how it's working not working so you can really tune your spend again. And so these tools are not intended to take away the work of the analysts. What they’re intended to do is to take away all that mundane work, or the work that frankly you and I just can’t get to. And so if you think about that overused term, “paradigm shift”, many analysts have been using Excel or PowerPoint in order to get their work done, or a myriad other technologies. And essentially here we are automating that process so they can come in the morning to a set of results that tell them the quality of their data and tell them the impact of their marketing performance. And then they start to work.
Kim: This is impact of all kinds of media spend is it? It could be social, it could be, you tell me, mobile, desktop?
David: Absolutely. So the first release is for search. This release that we are doing this week is for search and then every six weeks are are pushing out for the various other ad formats. Like display and social advertising and so on and so forth. So those are the first three that we are doing. And so by the end of Q4 we’ll have released all of those.
Kim: Another thing we talked about earlier is the quest of the 360 degree view of the customer. So I’m wondering when you’re analyzing this data, when you’re putting together and providing the feedback on it, are you appending the data to profiles? Are you tracking identities through the data and pulling it together that way?
David: Good questions. So there are really two levels of data that are being analyzed today. One is the aggregate level data. So how many customers came, what actions did they take. And that data is typically the data that brands and agencies are using to report their activities. And then there's the event level data which is to your point looking at individual customers and looking at them and their actions through your website or your apps or ecommerce or whatever ways in which you're gathering data from your customer. Like most platforms ours deals with both sets of data. Large, typically structured data sets, but typically aggregate and event level. And that part of our business is changing pretty significantly. We’ve done a lot of work around data ingestion and transformation to enable that data to be more useful in more settings.
Kim: Is this on a route which will lead to providing your customers with enough information to the to target ads individually?
David: Yeah, we have a number of things in our pipeline, or product roadmap I should say, which is primarily targeted for Q4 around this. So the answer is in short, yes, we are very much focused on how to help our clients get closer to their customers, and to their potential customers. And in order to do that we are accelerating development and release of our tools that do just that. And those tools are intended to pair up with other tools that we know brands use like media planning and media buying tools. But we’re not building those tools but rather integrating with them making it easy for brands to create better, smarter segments, that kind of thing.
Kim: Now in a sense, someone who follows the space or reads about it they might think to themselves, didn't we have this solved already? We have big data. We can track people across devices and channels we can talk to them precisely. What’s so new about this? Didn't we figure this out already?
David: Well the funny thing about marketing data is, unlike almost any other data source that you can think of within an enterprise, we think human resources data, sales data, inventory, finance, all those data sets are in a big data warehouse and they have been for years. But marketing data is the exact opposite. It's actually scattered. So the more agencies you employ, the more geographies you operate in, the more brands you have, the more data sources there are and by the way the more channels you market on, what ends up happening is what I often talk about as a “small data problem.” You’ve just got many, many parts of small data that are in many different places and are hard to get at. And sometimes for a global brand that could be an analyst in Kuala Lumpur, in Excel who is maybe getting manual uploads of information, could even be getting printouts from a local media provider.
So not all things are equal in the media world. We get a bit spoiled here domestically. Things tend to be in digital format and be sophisticated. But as you go across the world that changes very rapidly. And yet when you think about it from a global brand perspective some of their biggest markets are overseas in these locations so one of the things we are working on is to make it easy to bring together the all those dispersed marketing sources. So that's why we talk about “small data” problems but when you bring those things together.
There tends to be more issues because the work hasn't been done to normalize them, so in the marketing intelligence business the hardest thing to do is to bring together all the data from the dispersed sources and then to do the data transformation, to make the data really useful. And that contrasts very significantly with any kind of business intelligence, which is why we coined the phrase and launched into marketing intelligence. We felt within business intelligence there exists this particular niche and really it requires a specialist skill to get at it.
Kim: It's clearly an enormous challenge. I mean I’ve spoken to companies that use 200 agencies around the world and tracking the digital assets is enough of a struggle without tracking performance. Are you reliant on your clients to tell you where their data is or do you align with them, sit down with them and talk about where their data sources is?
David: It's really the latter. Clients are working on the data problem, be they agencies, be they brands, and both in concert are working on their data problems, typically coming at it from different angles. But when we talk about the single source of the truth it feels like an elusive goal to some. And one that frankly even today with all the sophistication that exists out there many brands have yet to achieve. We’re leaning really hard into that and helping brands and agencies get there faster. But also this is a change, they will have to change the way they work, they have to want to do that. But I think from the viewpoint of the brand and pull-through that occurs with their agency partners, this just makes a ton of sense.
So if you're sitting in the C-suite, lets call it the CMO of a large multi national brand, and it takes you a month, two months to get a view of what happened worldwide, that rearview mirror is not very helpful to you. Business does not operate that way anymore, business are much more cynical based on consumer patterns and purchases and that's really important today more than ever. It's not like the old days where you set a campaign plan it out in a years advance buy out all the media in advance and like they used to say, “Set it and forget it.”
But it's quite the opposite now. You can change your whole media strategy next week when you learn something in a particular market. That's the kind speed of data that is super important. And at the same time, as we were talking about earlier, there’s an ever increasing volume of data. But volume doesn't equal quality or anything else. It's [important] to sort of to sift through the volume and, not make the volume a problem, but to solve that and set it aside so that people can focus on always getting to that insight really quickly.
Kim: Okay and just to wind out, I know people don't want to always call out customers by name, but give us a sense of the types of clients that you are looking for. What kinds of business are you choosing to serve?
David: Historically we’ve worked with a lot of the global agencies and large brands, but as our product has been substantially automated and self served as was released this year, now we’re able to bring those same solutions to the whole audience of mid-market brands that we wouldn't have even talked to 2 years or a year ago. That’s kind of an exciting development for us in addition to the AI piece. For brands, any kind of mid-market brands, but typically speaking the companies that benefit the most, if that they are spending a couple of million dollars or more in advertising they can start to utilize these tools to get very rapid returns. If they’re spending tens of million then it pays for itself in a month.
Kim: That's the kind of thing a CMO likes to hear.
Kim: Okay. Thank you David, very much for talking to us about Velocidi.
David: My pleasure. Thanks for having me.
Kim: And everyone look out for next One-on-One Podcast.
Many thanks to Kim Davis and DMN Tech for orchestrating this podcast!