We recently had a chance to speak with Kiril Tsemekhman, the chief data officer and senior vice president at Integral Ad Science. He shared his insights on the latest trends in the advertising industry, including Big Data and mobile ads.
Q: How do you define Big Data, and how does IAS put it to use?
A: Big Data is an extremely broad term, and there is no common definition of it. On the one hand, it reflects an amount: When you say “big,” you imply that there is a lot of data. It encompasses both structured and unstructured data. You generate, collect, store and make it accessible for various applications and analyses. That’s when it becomes Big Data: it’s the data itself, the data stack of frameworks and applications designed to make this data accessible and analyzable and processes developed to make sense of the data. If you can’t make sense of the data – it’s not Big or Small – it’s useless.
The data Integral Ad Science collects and analyzes falls into this category, both because of the amount of data and how we only collect and store the information we can use for analytics and insights to create our data products. We use the data to understand the environment, the circumstances, and the way different advertising units are rendered and seen by users. We collect as much useful data as possible on each and every ad that is being shown and we are able to track. There are many data elements that we intentionally don’t collect today that we will probably collect in the future once we know how to put it in use.
Q: What kind of things do your clients learn from your Big Data analysis?
A: Our data has many different applications. One is client analytics which is essentially the aggregate-level view of how their campaign performed. We can report on the viewability of their ads on different sites and pages and how much fraud was present – if any, what the ad environment was like, how many risky pages were hit by their ads, etc. Building off our data, we also provide insights into how to improve campaign metrics.
The other application is what we call data products, or scores. Scores are statistical models representing different variables that we collect and analyze. For example, the average viewability on a specific placement (page, location, ad size, etc.) or on a particular page, that can be used both for analytical insights for clients as well to contribute to real-time decisions for clients who need to price the impression (bid), according to how much value it is likely to produce.
We also use our data to generate and apply fraud detection models, which are trained on large number of events and then applied in real time.
The same data is used to prevent ads from being shown on undesirable websites. We have a product called Firewall that combines real-time signals and placement level scores, using them to decide on the client’s behalf whether to show the ad or not.
Finally, the most advanced application of our data is the analysis of causal impact of advertising campaign on advertiser goals – which is the holy grail of any marketing. We determine how many incremental purchases (or interactions with the brand) were generated due to (‘caused by’) the ads shown to users.
Q: Why is media quality so critical in digital advertising?
A: To me, it’s like driving a new expensive car on a terrible road, or watching a great movie in a rundown theater. Here, when you create your message to the consumer, you hope to show it in a good environment. You try to do your best to engage the consumer with the ad – to make sure they see the ad and respond to it. To do this, you need the ad to be as viewable and as attractive as possible. You want the environment to be clean, giving the user a chance not only to see the ad, but to pay attention and engage with it.
When a person sees an ad, they tend to associate what they saw with the environment, so if the environment is bad, it is likely that that perception will be transferred to the brand as well. Given that so much attention and effort has been paid to finding an audience, the environment component has been somewhat overlooked, and that is potentially very detrimental not just to brand reputation, but also to the effectiveness of the campaign. In poor environments, ads have virtually no chance of producing a desirable effect on the audience.
Q: What is the True Advertising Quality Score (TRAQ) and can it improve ad placement strategies?
A: We treat it in some sense like a FICO score for the media industry, except that FICO is more of a risk management score. The TRAQ score is a combination of risk management and assessment of positive potential of ad placements. Not only does it include components that help eliminate risk, but it also allows you to select placements using the TRAQ score to improve the chances of the message reaching and engaging the user. Consider, for example, elements like clutter: if you have three ads on a page in front of a user, it is most likely that the ad is less effective than when it is the only ad on the page. Still, in either case ads are not totally wasteful. You just need to price them differently – and TRAQ will guide you in your pricing and bidding strategy. So, having a TRAQ score allows you to evaluate the quality of the placement that you are offered and to decide whether to choose to show the ad or not. If you decide to show it, TRAQ also serves as a guide to how much to pay for it.
Q: Your Company recently acquired Simplytics, a U.K.-based mobile ad service and analytics platform. How serious is mobile fraud, and how are you protecting mobile buyers?
A: Mobile marketing is already picking up a lot, and mobile analytics are becoming more fashionable and popular. There is little knowledge of the presence of fraud there yet, and the common belief is that it is very nascent at this moment. As the market grows there will be a lot of incentives for fraud to penetrate the mobile space. I think so far this is not a major market for fraud, but it will definitely be there sooner rather than later. The way that mobile devices will be compromised is also very different from how it’s done on desktops, and fraud detection and protection methods will have to adjust accordingly. With Simplytics technology, we are already working on various ways to detect fraud.
Q: So the goal of the acquisition is to create mechanisms to protect mobile buyers in the future?
A: It’s to expand the same services to mobile clients that we now provide to our clients in display advertising, and it includes not only fraud detection and protection but other media quality metrics such as brand safety for mobile applications. Furthermore, there is a huge concern about ads not being viewable because of lack of compatibility with different systems due to the industry fragmentation and concerns about media quality. We will also be looking into cross-channel and cross-device campaign effectiveness. We are talking about much more than just fraud.
Q: Your Company spearheaded video ad monitoring. How is this capability beneficial, and how does it enhance your ability to analyze the quality and usefulness of video advertising?
A: It’s a very successful new product. There is a lot of traction and interest in the online video advertising space both on the buy and sell side. It is also very important for our clients, particularly brand clients, for whom video represents a significant piece of their digital media budget and is an extension of their TV advertising. However, it is more measurable, less explored, and more risky than TV advertising. To make clients more comfortable with video we need to be able to measure user interaction with the ad from inside the player, and that’s a new technology that we developed. That’s what attracts brands, buyers, and sellers to work with us; we protect them from risky environments and help optimize their video advertising for viewability and better interaction with users.
Q: Finally, in your opinion, who’s doing a better job in leveraging Big Data – buyers or sellers?
A: It depends. I wouldn’t say it’s the buyers or sellers that take better advantage of Big Data. Among buyers, there are companies that are much better prepared, and on the sell side there are companies that have been in the business of data analytics for a long time, so I wouldn’t say that on the whole buyers or sellers are better at it.
Traditionally platforms are better prepared because they are technology players; you can find sellers – such as traditional publishers – whose core business is not analytics. They are becoming much more savvy. Platforms like exchanges, SSPs or DSPs are traditionally technological players, and they are the ones that push the boundaries and create new technologies and approaches to Big Data. So it’s not about whether it’s a buyer or seller, but who is more technologically advanced.