In his four years as Havas NA’s SVP of marketing analytics and data consulting, Leon Barsoumian has worked to marry programmatic and analytics.
From a programmatic standpoint, the French holding company has two key offerings, Affiperf and Artemis. In 2010, Havas debuted Affiperf, a global trading desk that Havas calls a “Meta DSP” and is designed as a control center that lets clients operate programmatic campaigns across multiple DSPs.
Havas inherited its DMP, Artemis, following its 2001 Hook Media acquisition. Artemis has since been enhanced with analytics, acquisition sizing, attribution modeling and engagement scoring. The platform processes more than half a billion events per day and houses 100-plus terabytes of data.
AdExchanger checked in with Barsoumian about Havas’ programmatic efforts, and the challenges an analytics chief faces as automation evolves.
AdExchanger: What are Havas’ core programmatic offerings?
LEON BARSOUMAIN: You can either outsource to a third party or you can build it in-house. From a holding company perspective, we started a trading desk operation [Affiperf]. The thought was, if you have traders at a trading desk sitting at each of the local offices, the value is that they’re colleagues and they can be part of the bigger plan. It’s more of a collaboration and ultimately means results are stronger.
What’s the value of centralizing Havas’ programmatic offering?
It’s not fully centralized. Programmatic is a pretty general term, but when I talk about Affiperf, it’s an RTB environment using a DSP interface. We also work with other people outside of Havas that buy programmatically. So it’s not fully centralized, nor are 100% of our programmatic relationships in-house. It’s a hybrid approach.
The space is evolving and everyone is getting more sophisticated in how they pitch their offerings to media planners and buyers. Sometimes there are clients on a plan that may be competing against our in-house programmatic offering in terms of inventory. What ends up happening is that we have conversations with clients about what they’re trying to achieve, and if there’s a better way to achieve those objectives. That can mean consolidating or recommending that clients do work with a third-party firm if it serves a particular niche.
How much of Havas’ programmatic business does Affiperf account for?
I can’t go into specifics about that. It’s a function of what the campaign objectives are as well as the insight objectives, and who is best equipped to handle it.
What is Artemis?
Artemis is Havas’ proprietary DMP [data-management platform]. I hate to use that acronym, but that’s essentially what it is. We’re getting daily data from ad servers at the cookie level, and we use that for reporting to clients and analysis of consumer journeys. Some of that information is passed along to Affiperf for targeting or for insights on how to target. Essentially, it’s a reporting, data aggregation and activation platform, and it’s managed in-house.
How do you handle transparency and programmatic?
Marketers are having a hard time in general with getting transparency around programmatic. It can be a function of a bunch of things. What we get by having people in-house is access from an analytics perspective. That helps us provide as much transparency as possible about what types of data sets, third-party data or segmentation we use to target folks.
We’re doing a lot of education with our clients around how programmatic works, the information that we have access to within the programmatic space, and what limitations we face. After we run a campaign, we offer detailed reporting on what happened during the campaign versus saying, “Here are your results.” For each programmatic campaign, we spell out what type of targeting we did, the corresponding performance and what we can change to better optimize.
Is data quality a problem?
Third-party data is a challenge for everyone in the industry, whether it’s being purchased programmatically or not. Think about the MRC. It was established by an act of Congress to ensure things are done in a standardized and credible way. But there isn’t any sort of accreditation for data. Every day there’s a new company that claims they have certain data or access to certain information.
Right now, the way we’re vetting it is to essentially test things out. Our clients focus on direct response, so if they don’t see ROI, we move on to the next data partner. Until there are regulations to validate the quality of third-party data and guidelines around what constitutes valid data, it’s going to continue to be a wild, wild West.
How does that tie into viewability?
The challenge in the ecosystem is that when an ad call comes it, and you’ve got to decide whether to bid or not, there isn’t necessarily a signal attached that tells you if it’s going to be a viewable impression. There’s pre-bid technology that can determine if a bid is more or less likely to be viewable, but it’s a likelihood. Even if the impression is likely to be viewable and you bid on it, it can ultimately be a non-viewable ad. The technology isn’t there yet from a measurement perspective, so you don’t know what people will ultimately see or not see. Unless we start changing the way ads are rendered, that’s a limitation we’re going to see for the next few years.
How do you solve that problem for your clients?
As an agency, we’re trying to address viewability in a different way. We’re rolling out something in the next couple weeks. We’re trying to figure out how to help clients make decisions based on their comfort level of uncertainty with viewability thresholds. We’re not perfect and we’re on the same learning curve as everybody else. But more and more, there’s a formal relationship between programmatic traders and analytics. We’re trying to formalize that even further.
From a strategic level, we have to first define campaign objectives and the strategies, and then decide on the tactics to test and optimize campaigns. Often times traders and buyers are good at their craft, but they may not be really good at telling the story, doing analysis or teasing out additional insights about audiences. That’s where the marriage of programmatic and analytics needs to happen.
