Niklas Karlsson

Chief Scientist and VP of Research and Development at Oath

Dr. Niklas Karlsson is the Chief Scientist and Vice President of Research & Development for Oath Inc‘s Programmatic Ad Tech Demand Platform, where he creates and implements the research vision around AI applied to online advertising. Niklas received a Ph.D. in Engineering with a focus on Control Theory, Dynamic Systems, and Robotics; an M.A. in Statistics and Applied Probability from UC Santa Barbara (UCSB); and an M.S. in Engineering Physics from Lund University. He is also an alumni of the Stanford Executive Program. Niklas holds 29 patents and received the Distinguished Alumni Award from the Department of Mechanical Engineering at UCSB in 2015 in recognition of “outstanding application of systems engineering principles to the field of online advertising” and the Oath Master Inventor award in 2017 (the highest technology/science recognition within Oath Inc.).

Demand-Side Platforms & The Future Role of Marketers

Programmatic advertising is the cornerstone for companies such as Oath, Google, and Facebook to grow and drive revenue. The strategy uses advanced algorithms and trusted data to automatically place relevant ads that drive sales for advertisers while creating a positive user experience. From a marketer’s perspective, it provides a means for designing and running a marketing campaign with less guess work.

A Demand-Side Platform (DSP) provides a particular venue for programmatic advertising. It is the middleman between an advertiser and one or more open impression exchanges. An open exchange is an open marketplace where buyers can bid either manually or programmatically to purchase ad impressions. In this context, an impression is an opportunity to show an ad creative (e.g., banner ad, text ad, or pre-roll video commercial) to internet users. The price of an impression purchased through a DSP is determined in an auction held by an ad exchange. Hence, it is dictated by the supply of similar impressions and the demand from competing advertisers.

The goal of a DSP is to optimally manage an advertiser‘s marketing budget. For some advertisers, the goal is to drive branding and reach, while for others it is to maximize the number of clicks or sales.

Business Model

To appreciate how a DSP can serve marketers, it is important to understand the underlying business model. It is based on the principle “winner takes all,” which is not totally accurate but is directionally true. In particular, since a high percentage of all impressions are sold on open exchanges, competing DSPs (representing advertisers competing for ad impressions) have access to roughly the same internet users. A rational advertiser brings their ad budget to the DSP that can generate the best return on investment (ROI) on the ad spend. For example, if DSP A can generate slightly more conversions than DSP B on the same advertisement budget, then the advertiser is likely to transfer its entire ad spend to DSP A. In practice, it is more complicated. An advertiser may not want to create an exclusive dependency on a single DSP. It would also take into consideration what is offered in terms of customer service, reporting and insights, ease of use, etc.

In other words, the focus of each DSP must be placed on producing the best possible campaign performance for the advertiser given their ad budget. The business model is illustrated in Figure 1 as a flywheel. Superior performance is enabled by feeding the optimization algorithms with the best and richest data about internet users. Larger volumes of rich user data, on the other hand, is obtained by data deals1, mergers and acquisitions2, and by managing larger advertising budgets. Finally, larger advertising budgets are incentivized, as explained earlier, by generating better performance than competing DSPs.

Oath’s DSPs, powered by leading ad optimization engine AdLearn, are a good example of the above business model put to work3. Oath brings together trusted data from a variety of sources, which enables better optimization, and AdLearn implements state-of-the-art AI and machine learning with the simple goal of serving the right ad to the right user at the right time and place.

The Future Role of a Marketer

If the role of a marketer is to manage online ad campaigns, then the most efficient way to do that job has changed. Before DSPs, the job likely involved lots of planning, manual A/B testing, and plenty of guessing on where to spend ad budgets and what audience to target. With an access to programmatic advertising in general, and DSPs in particular, it is far more appropriate to perform the job as follows:

1. Find a trustworthy DSP

This is critical since there are many DSPs with a questionable reputation.

2. Carefully define the campaign objective and the relative importance of multiple objectives.

Note, if marketers want to maximize the total number of conversions, then they should aim to minimize the effective cost per conversion. Since impressions are priced differently, according to their market price, it does not make sense as an objective to maximize the conversion rate.

3. Monitor and evaluate the ROI to reconfirm campaign objectives are met and that the DSP is trustworthy.

Most importantly, know that success is a combination of technology and human assessment. Algorithms process data and identify complex patterns and predictors of clicks and conversations. However, these algorithms need human input on what to optimize and good judgment on how.

Written by Niklas Karlsson, Chief Scientist and VP of Research and Development at Oath.

Edited by Shiying Yao, Medill IMC Class Of 2018


1: A DSP provider may sign a data deal with a third-party entity to access audience data that makes it easier to determine what users will respond favorable to ads from a certain advertiser.

2: Mergers & Acquisitions in the ad space typically result in a larger pool of rich data that improves the prospects of audience targeting and campaign optimization.

3: Oath Inc. is a fully owned subsidiary of Verizon Communications and consists of internet icons AOL and Yahoo, which were acquired by Verizon Communications in 2015 and 2017, respectively.

© 2020 Northwestern University

1845 Sheridan Road
Evanston, IL 60208-2101