Dentsu Inc. × Google Pioneering the New Programmatic Market: The Dentsu Inc. Private Marketplace (PMP)

At station gates, tickets have been replaced by Suica cards; soon, cars will be driven by machines rather than humans; and eventually, even padlocks will be automatically secured via smartphone operation. While some of this remains untested, such a "mechanized and automated" world certainly seems highly convenient.

The "programmatic" advertising transactions Dentsu Inc. and Google are pursuing fundamentally stem from the "mechanization and automation" of advertising transactions. However, what we are tackling as a project is neither mere automation of advertising transactions nor a cost-cutting measure. It is the development of new markets.

Dentsu Inc. has established an unprecedented collaborative framework with Google and is now fully engaged in market development. This market is the programmatic advertising market, specifically focusing on the Private Marketplace (PMP).

Following the release below in October 2014, we received significant feedback. Starting this February, we began sales in beta form and have successfully achieved certain results. If progress continues smoothly, we will finally commence full-scale sales starting in July.

Dentsu Inc. PMP Press Release: http://www.dentsu.co.jp/news/release/2014/1006-003837.html

 

Dentsu Inc. PMP's Goals and Key Factors

For the time being following this article, we intend to start from the basics of what programmatic and PMP actually are, and then delve into specific descriptions within the programmatic domain. However, at this stage, the goals of Dentsu Inc. PMP and the key factors involved are already clearly defined.

Dentsu Inc. PMP is the initiative to achieve the above. To this end, we have formed teams both inside and outside the company and are conducting various trials daily. In this column, I intend to summarize as much as possible the current efforts and the market formation we are challenging.

Furthermore, within this project, we are deeply collaborating with Aegis companies, particularly amnet, which possesses strong capabilities in the programmatic domain, to advance product development. We plan to add details about this collaboration with amnet at a later date.

Globally, the WPP Group, centered around companies like Xaxis and GroupM, is the most advanced in this field. Its defining characteristic is setting PMP as its primary focus, rather than the typical programmatic approach of "open auctions (RTB – a purchasing method allowing real-time bidding for each impression slot with no participant restrictions)." Last year, a GroupM executive's statement, "We want programmatic to be 100% PMP (RTB 0%) by the end of 2014," was received with great surprise.

A major feature of their PMP approach is the "upfront" (meaning advance investment or prepayment) purchase of premium ad placements in advance. They bulk-buy the placements they define as premium, sometimes over a span of a year, effectively managing the risks associated with such bulk purchases while delivering client benefits. Why do they take on the significant risk of large-scale upfront bulk purchases to secure premium slots? While arbitrage (trading to profit from interest rate or price differentials) is certainly a goal, it's also because they understand how much of a variable site and ad slots are in relation to performance.

Performance Dependent on Ad Slots

Recently, under the banner of "from slots to people," audience-centric ad delivery has dominated the market. Among these, retargeting, capable of achieving overwhelming ROI, has become one of the leading ad delivery methods. How skillfully retargeting can be segmented and expanded is a critically important factor for performance. Despite its immense popularity, our research has uncovered a surprising fact: while retargeting is the quintessential "person"-based approach, its performance is heavily dependent on the advertising "slot" (site).

Below is the conversion analysis result for a retargeting campaign. While it's labeled as "person"-based, the results show conversions concentrated on specific sites.

Out of 20 cases we investigated, a full 60% were heavily dependent on specific sites (the top 20 sites captured over 50% of all conversions). If you can access delivery destination data, we strongly recommend reviewing your retargeting delivery destination results list.

As noted above, despite the significant impact on performance and the high value of sites and ad slots, they are often undervalued during media planning.

What I wish to convey here is not that audience data holds less value than sites and ad placements. Rather, the phenomenon above suggests that when viewed through the lens of audience value—a buyer-driven metric prevalent today—the high quality of audiences visiting these media sites positively impacts performance. This highlights that "the value of media itself should be reevaluated." In other words, we are entering an era where we must not only refine and expand retargeting list segmentation using tools like DMP and GA, but also re-evaluate the element of media/advertising space as a variable within audience segmentation to optimize ad delivery.

What is programmatic advertising?

In advertising transactions, programmatic refers to the automated, machine-driven execution of ad deals through systems on the ad delivery side (DSP) and the ad supply side (SSP/exchange). A common question is how it differs from ad networks. The key distinction lies in whether two systems exist on the delivery and supply sides, and whether players (regardless of whether they are advertisers, agencies, or DSP vendors) can choose participants at each layer. In other words, the distinction between ad networks and the programmatic domain can be made to some extent based on whether the model is vertically integrated or horizontally separated, as shown below.

In programmatic, variables like DSP, SSP, and targeting are selectable by the advertiser. For example, if there are 10 types each of DSP, SSP, and targeting, there are 10×10×10=1000 possible combination patterns, and the optimal combination must be selected from these.

Ad networks, on the other hand, offer a selection of pre-packaged menu options for delivery formats, regardless of the underlying ad delivery or supply system. Because they are packaged, selecting players at each layer is generally not possible.

However, this distinction is purely for convenience and isn't inherently clean-cut. For example, a system might use programmatic delivery mechanisms but be packaged as an ad network during sales. Conversely, a fully automated optimization DSP with AI features might utilize both ad delivery and supply systems, yet offer almost no customizable operational variables, effectively functioning like an ad network. (This could conversely be considered fully programmatic.)

In any case, the current "programmatic" transaction involves using two systems—the demand side (DSP) and the supply side (SSP)—to conduct ad transactions "automatically and mechanically." This definition will likely continue to evolve around the concept of "automation."

Indeed, the definition of "programmatic" in advertising transactions is currently undergoing significant change. Until now, "programmatic" was largely synonymous with RTB, as defined by Rocket Fuel. That definition was: "Selling remnant inventory in open auctions with the added value of targeting." While this approach achieved high performance, gained widespread support, and formed a large market, concerns have recently begun to surface regarding advertising credibility. These include issues like ad verification (tools/systems to prevent brand damage from ads) and ad fraud (tools to prevent invalid impressions, clicks, or conversions from bots, etc.). To address this, a new transaction model called the Private Marketplace (PMP) is gaining attention. It aims to maintain the flexibility of ad delivery while ensuring the credibility of the ads.

This article concludes here for now, but the next installment will delve specifically into PMP. While it uses systems on both the demand side (DSP) and supply side (SSP) for delivery, much like RTB, PMP is fundamentally different. We will focus on explaining what PMP actually is and how its mechanisms can provide benefits to both clients and publishers.