Last week, MediaMath announced that it had secured a new US$175 million senior credit facility, led by Goldman Sachs, in partnership with Santander Bank. The new line of credit will support MediaMath’s growing scale of the business, including the refinancing of existing debt facilities and to fund its ongoing growth objectives. Campaign talked to Rahul Vasudev, the company's APAC managing director about how the independent programmatic company differentiates its approach in an ecosystem where DSP services are becoming increasingly commoditized.
Last week, MediaMath was named as a market leader by Forrester. In your view, what are the attributes that led to that recognition?
Being named in the latest report really gives us a lot of pleasure and excitement because the state of the industry today, a DSP platform is expected to have a minimum US$100 million coverage in order to be considered for the Forrester Research report, says plenty about how the ecosystem has evolved since MediaMath started. And within that report, the kind of things they are looking for—such as ‘do you have a strong identity graph?’, ‘do you have a strong ecosystem play?’, ‘are you able to bring more tools to your clients who want to solve so many problems?’—there are so many things that marketers need to worry about. And so the report by Forrester speaks of the need that a platform be able to handle all these different things, to make programmatic marketing genuinely good as a response mechanism for the clients. The fact that there are only 11 significant or noteworthy players in the ecosystem is good, and within that, MediaMath has a good reputation in the market. People have heard of us. So being one of the largest ones and being mentioned in the report is great. Forrester to their credit actually conducted thorough diligence, going in and taking presentations from every vendor. This was to understand the capabilities, the future product vision, what releases are forthcoming, what clients say. And it was great to see that we like our clients as much as the clients like us back.
What is unique about the approach MediaMath takes in order to serve its clients?
So, a large part of our business comes from agencies. And MediaMath was one of the first ones to introduce The Triumvirate Advantage whereby the advertiser, the advertiser's agency, and the technology vendor as a whole play the greatest role in making anything happen. Because not every client or brand has the capability in-house to make it happen. And on the brand side, the focus is on creating products and striving to create better products for the end customer—as opposed to getting involved in the media space and the learning investment it requires. I come from a media agency myself and people there are constantly in learning mode with regards to what Facebook is doing, Google is doing. There is so much to learn and it's unfair to expect that someone sitting in a silo somewhere will stay up to speed with the industry. So the brand, the agency, and the technology together—we consider it to be the ecosystem for us.
With so many players in the DSP ecosystem offering services that are not exclusive to any intellectual property laws or patents, how does MediaMath differentiate itself from the commoditised remaining players?
A few things set us apart. Obviously, our internal company culture, our beliefs, our philosophy. What makes us come to work every day. The original intent was as John Wanamaker famously said, “50 percent of marketing is wasted, I just don’t know which 50 percent.” Our CEO Joe Zawadzki remarks that this is a terrible tagline for an industry to have.
So that drives us every single day to help make sense of what is and what isn’t working. We come to work thinking, ‘what can we provide now as a set of tools, capabilities, ecosystem integrations, as a set of identifying who your customers are, who your best customers are. The best customers are those you should be willing and able to pay more for and knowing who you should be willing to pay less for. Knowing who gives you what kind of return on marketing. Some will give you a $1 return and some will give a $100 return. Your sweet spot lies somewhere in between, where there is the largest gain for the largest return. And...people are just not able to do it.
So that’s kind of like the driving philosophy behind why MediaMath was formed, why we come into work, what problems we are trying to solve. Obviously, we are trying to solve these with lots of small problems like, there is a lot of inventory that is being bought and targeted at bots, fraudulent traffic and bad players in the ecosystem. How can we prevent that? So that’s one way in which we solve that to remove 2 percent of that 50 percent that’s not working.
So I think that is something critical in driving us. Our clients [advertisers and agencies] care about giving a great customer experience. They want to know that you, the customer, are viewing communication at the time you are most receptive to it. And that the viewer is only seeing content, ads, and offers that is relevant to them. If we step back a few years ago, companies had to deal with a plethora of mobile companies, video companies, and social networks. And the aggregators just added to the noise. Marketers with time came to grips with it, accepting a reality wherein they would deal with 17 vendors to reach one customer. But if you are the customer, they are having a terrible experience because three mobile companies are not talking to each other and they are each showing one person a brand ad 20 times when on the client side the frequency is capped at five impressions per unique viewer. So marketers want a tool that is sophisticated enough to allow their brand message to reach the customer as an individual, as a person who has genuine tastes, and be able to serve them the right content at the right time, and at the right price. Because if I represent a shampoo company and you represent a luxury car company and both of you are trying to reach Priya, you are probably willing to pay S$100 for her and I am willing to pay S$0.50 for her. We should be able to distinguish the two, else it's an unfair right. So, as MediaMath, how do I become intelligent enough as a platform to enable our advertiser to understand 'at what user, at what price?' is relevant.
And you're able to go that deep with every single campaign, every single vertical, and category?
Exactly, so that's where we have arrived. We've been in tools and our focus on machine learning is I think is one of the things that sets us apart. So the first one was like omnichannel, like truly omnichannel—reach the customer across all channels irrespective of walled gardens, siloes, individual strongholds that might have been built.
So you might have heard that MediaMath is starting to do Facebook, which is great because Facebook understands that there are clients that have their own data and they want to be able to reach their customers with their data and control that as part of the broader marketing function that they want to achieve, as compared to 'here's my Facebook plan'. So this is great. We believe Facebook is one of many such walled gardens that will come in together to give the marketer control of their overall marketing ecosystems.
So, now you have access to all the media. Now I can serve an ad on mobile, serve an ad on Facebook, serve an ad on video—now what? How is my entire set of marketing systems that have been created, how are they learning and getting better over time? So we actually invested very heavily in putting together a team of data scientists who are working on improving our algorithm constantly. The machine, every time it sees an impression, every time it serves an impression, every time it sees a response to the impression, every time it sees a transaction on the client's website—the algorithm is learning.
Any CRM connectivity for post-transaction frequency capping?
One of the principles is that because we are so focused on business outcomes for the customer, the algorithm needs to learn that this customer has transacted. The algorithm sees that yesterday she transacted and has to determine how likely she is to transact today again. If she bought a car, probably not—she's not going to buy a car again. But if it was something else, then maybe it is relevant. Or maybe another kind of communication: if she bought a car, maybe she needs [child] seats.
The algorithm has to learn and become better with every transaction, with every exposure, with every bid. The first thing that happens with any programmatic transaction is the bid opportunity. So an exchange, a publisher like Yahoo, and a customer. A banner ad space makes a call to Yahoo's server, Yahoo, in turn, calls AppNexus, and they, in turn, give us a call. Now we have to decide, is that bid opportunity worth it? Who is the user behind that opportunity? Which means you must have a very strong identity graph to know who that user is, what our past experience with that user has been. All this happens in 100 milliseconds, to tell Yahoo 'yes' or 'no', and if the former then 'at what price'.