[What You Need to Know Now] The Latest on DMP (Part 1)

Two planners leading Dentsu Inc.'s data marketing division visited Legoliss, a company headed by Mr. Sakai (center) who possesses diverse backgrounds including engineering, marketing, and business development.

How will the advertising and marketing industry, rapidly converging with technology, change in the future? Young planners tackling digital marketing at Dentsu Inc. Digital Inc. explain the latest insights centered on "Data × Marketing."

Part 3 focuses on the integrated environment for data marketing: the DMP (Data Management Platform). While DMP adoption has grown rapidly amid high expectations, its practical implementation and success know-how often remain highly dependent on individual expertise. Even among pioneering companies, voices frequently express that results haven't met expectations. This time, three key figures—Katsuaki Sakai, CEO of Legoliss (a leader in DMP implementation and utilization in Japan)—and young data scientists from Dentsu Inc. Koichiro Kondo and Taku Arakawa—discussed their on-the-ground trial and error experiences and future outlook.

※Part 1: "The Latest on Audience Data (Part 1): What You Can't Afford toMiss"
※Part 2: "The Latest on Audience Data (Part 2)"

Pioneering Data Marketing from the Advertising Domain

Kondo: Today's theme is "DMP Utilization: What You Can't Afford to Miss." While we'll defer the basics of DMP to previously published articles, I hope we can share insights unique to those actively involved in the field.

First, Mr. Sakai, who's in high demand for DMP projects—what has your career path been like?

Sakai: I've worked across various industries. I started as a systems engineer at a major electronics manufacturer before moving to a pure-play online advertising agency. I then served as a business development lead at a startup and as a marketing lead at a corporate entity. At that company, I handled hotel marketing, engaging in a wide range of activities: determining location, price points, and added value like amenities; local hiring; and offline customer acquisition strategies. This experience was pivotal—I learned that digital and data alone have limitations; you must execute integrated marketing encompassing offline efforts to achieve results.

After that, I moved to a full-service advertising agency and spent six years strengthening their digital division. In 2014, I joined my previous company, Modulo, as a director.

Arakawa: And then last year, you became independent and launched Legoliss. Could you tell us about your current work?

Sakai: In a nutshell, it's "helping clients with data-driven marketing from various angles." It's not solely about handling data, nor am I fixated solely on digital marketing. My greatest strength lies in my background: starting as an engineer before entering advertising, having experience at both pure-play internet agencies and full-service ad agencies, and also having worked on the client side. So, I want to help with cross-boundary marketing activities, such as Technology × Creative, Data × Communication, and Digital × Offline.

Legoliss, though newly founded, handles a wide range of work from creative production to DMP segmentation and database visualization. Furthermore, we assist with all kinds of marketing initiatives centered around data, including ad operations and event planning. What is your background, Mr. Kondo?

Kondo: I joined the company in 2010. My first assignment was evaluating the effectiveness of digital ads for a major SNS platform, so I was focused on analyzing numbers from the start. Later, I worked in the creative department, gaining experience as a copywriter, among other roles. Now, as a data scientist, I consult on DMP implementation, identify data challenges clients face, and propose solutions.

Sakai: And Mr. Arakawa?

Arakawa: I joined in 2015 and currently handle projects for major telecom carriers and newspaper companies. I'm involved in the day-to-day execution of initiatives using data, including DMP operations, coordination with various media outlets, and work related to owned media.

Companies that implemented DMP but aren't seeing results...

Arakawa: Since we're focusing on real-world DMP experiences today, I'd like to share some candid insights. First, what's the most common industry topic surrounding DMPs right now?

Sakai: Actually, most clients who come to us for consultation have already tried using a DMP.

Kondo: They tried implementing it but didn't see results...?

Sakai: Or that it doesn't translate into marketing effectiveness. They have solid knowledge of the tools and features, but they come to us asking, "What's the truly effective way to use this?"

Or they say things like, "We have the data and the tech, but we need help translating that into communication scenarios," or "We tried to spread our wings, but it turned out to be pie in the sky. Where should we realistically start to succeed?"

Kondo: I get it. I often hear about companies that implemented a DMP but have now completely scrapped it. The common reasons seem to be that when they decided to implement it, the DMP was marketed as some kind of "magic box," leading to unrealistically high expectations that weren't met, or they couldn't define the scope properly.

Sakai: Absolutely. When first implementing a DMP, people tend to paint a big picture. "Cross-functional, organization-wide!" they say. They envision "dreams" like integrating all the data from five or six different departments, or connecting this data with that data... But balancing the organizational challenges needed to make that happen with the short-term ROI requirements is often extremely difficult.

That's why I often recommend starting at the smallest unit: "Let's first integrate the data held solely by this department with cookie data. Let's see what we can do there." For organizations where a top-down, simultaneous rollout isn't feasible, adopting a strategy of building small successes often leads to a more successful DMP implementation.

Kondo: Otherwise, the hyperinflated expectations before implementation could plummet, potentially leading to the entire DMP initiative being scrapped. While aiming for company-wide impact long-term is essential, we must first identify tangible, measurable outcomes as that initial step to get things moving.

Regarding the purpose of DMP implementation, I feel it varies significantly depending on which department leads the initiative. For example, if Marketing leads it, the goal is often targeting; if IT leads it, the goal is often monitoring. For targeting purposes, if you don't fundamentally change the definition of performance itself—like traditional CPC—there's a risk investment will concentrate on people who respond easily or who would buy the product even without seeing the ad.

Similarly, when monitoring, we sometimes advise setting optimal KPIs for each segment. Ultimately, if the DMP only focuses on harvesting, it may not outperform listing in efficiency, or you might question whether using a DMP is even necessary.

What are the winning patterns for DMP implementation in the current landscape?

Arakawa: I believe many clients struggle with the gap between their high expectations for DMPs and the reality of actual implementation. Given this, what kinds of DMP implementations are currently succeeding?

Kondo: Currently, successful DMP implementations fall into two patterns. One is top-down: overlooking the tendency for means to become ends, and pushing through regardless of cost, resulting in accumulated knowledge and execution capability at the operational level. The other is cases where the implementation objectives are well-defined, focusing spending only on areas with clear potential for performance gains.

Sakai: As mentioned earlier, when a top-down, full-scale rollout isn't feasible, starting small is the golden rule. Since DMPs still involve a lot of idealistic thinking, it's crucial to pare down requirements. For example, before company-wide implementation, try data linkage just between Division A and Division B, or start by linking email addresses to cookies, or attaching social IDs.

It's better to focus on areas that are realistic and likely to deliver business impact. I rarely hear stories of success where they started with an overly ambitious plan.

Kondo: Even with a small start, there's still a lot that needs to be done. Sakai, do you have any examples you consider successful DMP implementations?

Sakai: Regarding targeting, we've seen success using first-party data we own, rather than so-called audience data. It was a B2B campaign with relatively high CPM, where we linked cookie data with a list of corporate decision-makers and above for a one-to-one campaign.

The client adopted a strategy called "Share of Display," continuously serving ads to decision-makers and above via a DSP. They controlled bids through the DSP to ensure the company's ads consistently appeared whenever the target audience was active online. They didn't focus particularly on clicks as a metric; instead, they kept serving the banners within tolerable levels. The result? Targets started noticing, thinking, "I've been seeing this company's ads a lot lately." This led to a threefold increase in the response rate to sales calls made to those target companies.

Arakawa: That's impressive. It's rare to hear of such thorough targeting using first-party data, which tends to have strong attributes but a smaller pool.

Sakai: When dealing with high-value B2B transactions, it makes sense to go that far. By integrating data and pinpointing ad delivery, we managed with just a few million yen in ad spend. But when calls actually connected with people who saw the ads, it led to several orders worth hundreds of millions of yen. Sales reps gave us feedback like, "Wow, that's incredibly efficient!"

This was only possible because we had first-party data that allowed us to pinpoint customers precisely. As this case shows, linking with customer data advertisers already possess is a powerful approach.

Kondo: For consumer goods, where individual product prices aren't that high, I think it's better to use DMPs not primarily for targeting, but for generating user-centric communications or discovering previously unanticipated target demographics. That said, we must also recognize that there are accuracy issues with attribute data inferred from web browsing history. There can be a trade-off between sample size and accuracy. In that sense, it's crucial to constantly balance precision and volume by combining behavioral history with panel data.

Sakai: Exactly. That's precisely why it's wise to leverage DMPs without over-relying solely on third-party cookie data derived from web behavior history, while also linking it to first-party data.

When using a DMP to discover previously unseen users, it requires a more analysis-oriented approach than simply matching CPA in ad delivery. This includes selecting the right DMP and data for finding new targets, and using the DMP to validate whether the target hypothesis was correct. Once you start using the DMP in this more analytical way, the next stage is inevitably hitting the wall of data accuracy and operational challenges.

DMP implementation with an eye toward the data circulation era

Arakawa: I often hear concerns that even when the goals of data marketing become somewhat clear, the resources to drive it internally are lacking. Even if a DMP is implemented, if there are few people in-house capable of analyzing the data and it isn't fully utilized, it's often judged as not worth the cost. I believe fostering a data-driven marketing culture is essential, alongside advancing tool implementation and data collection.

From a field perspective, some argue that if the DMP's sole purpose is monitoring, alternatives like linking site measurement tools with panels could suffice. What are your thoughts on this, Sakai-san?

Sakai: It's true that currently, most data seller-type DMPs in Japan rely heavily on cookie data, so the data accuracy is somewhat questionable. However, looking ahead to the not-too-distant future, even if the data accuracy is questionable now, it's better to start trying it out sooner rather than later. In the future, changes in the legal environment, such as personal information protection laws, could stimulate data circulation. Companies holding data in a non-personally identifiable form, not just cookies, could become active participants in this data flow. This would enable businesses to procure the data they need themselves and utilize data like credit card creditworthiness and payment data, as is already starting in the US.

Indeed, Google's advertising products in the US already feature dropdown menus in their management interfaces allowing targeting of segments like "American Express Platinum Card holders or higher." Assuming the fundamental premise that attribute data does not contribute to personal identification, we are likely entering an era where such data cassettes will be swapped out as needed, much like game console cartridges, to utilize data effectively.

That's precisely why companies that start experimenting now—even if the data precision is still somewhat limited—and challenge themselves with data-driven marketing will gain a head start.

Up to this point, we've discussed practical, field-level insights to ensure DMP implementation doesn't remain a pipe dream. In the latter part, we'll further explore how marketers should engage with DMPs.