The following Q&A with Wei Guo, founder and CEO of FugeTech, is a followup to earlier articles about conflicts of interest in China's programmatic market. It is a response to iPinYou's rebuttal on why being involved in both PDB (programmatic direct buying) and RTB (real-time bidding) is not a "conflict" as well as Hdtmedia's reinterpretation of its DSPAN (DSP and ad network) role as a "stock broker" recommending media buys with upside potential. The interview has been edited for clarity and brevity.
You mentioned you are not 100 percent in agreement with iPinyou's point of 'using only one DSP otherwise resources may be wasted'?
It is a weak argument. Grace [iPinyou CEO] said that is in favour of the advertiser's interests. Let's look back to the advertiser's objective: using a certain budget to get the best advertising performance. How do we achieve that? Using one DSP? It's not going to happen.
Because each of the DSPs has their own strengths and weaknesses; each of them have access to different inventory. Reach and frequency will be very limited if you use just one DSP. Even the BAT publishers don't have access to all the inventory, not to mention one independent DSP.
So to put it simply, it should be a concept of diversification?
Yes. If the client wants to target, say, 25- to 35-year-old females, each different DSP can reach a different set of inventory. And even within the same set of inventory, different DSPs can get different prices depending on the volumes they buy.
When we designed our own trading desk, our basic rule is to use multiple DSPs and DMPs, and deeply integrate the various vendors.
The bottleneck, or administrative pain, for using more than one DSP is efficiency, if the goal is performance-driven. In overseas markets, the meta-DSP model for dynamic budget allocation is common for removing this bottleneck. Of course, other risks for using more than one DSP like the inadequacy of frequency caps and the possibility of "bidding against yourself" must be managed.
We already did a couple of case studies using multiple DSPs and DMPs within one single campaign. You actually want all the vendors to compete with one another to get to your target audience at the lowest price to maximise your performance. You can't just use one DSP.
For one of our clients, Student.com, we used three DMPs: Baidu (data size 3.7 million), TalkingData (0.5 million), AdRatings (1 million) and two DSPs (ZZY and Yosemite) in December 2016 campaign with the goal of generating online enquiries.
Take the illustration of Qunar.com, a travel metasearch engine, that allows you to compare prices for flight tickets. A meta-DSP will also allow you to compare prices for each DSP, then you can simply switch DSPs without administrative pain.
In China there are more than 200 DSPs, and at the same time there are a lot of overseas ad-tech firms and verification vendors wanting to enter China. Clients cannot tell who has the real technology and who can really perform. I can tell you that the old way they chose their DSPs was all based on presentation. The best way to tell which DSP has the best technology is to let them compete. Now, that's in the interest of advertisers.
In your opinion, what are other pressing challenges for China's programmatic scene?
When we ask our clients about the challenges they have, basically, they have two key ones.
Number one is pressure from advertisers to have performance transparency at the per-cost-per-URL level. DSPs offering real-time bidding (RTB) are like a black box. Essentially, when advertisers spend their budgets on RTB, they don't know the actual cost or actual delivery. Even the agencies cannot get actual results either. They can only get spreadsheets or powerpoints from the DSPs. For example, an advertiser may run 10 campaigns over one year and use different programmatic vendors for each. At the end of the year if it wants to know the overall performance, no one has a complete view.
Data transparency, especially on PDB, is the number two challenge. Brands spend a lot of money on DMPs and buy data to improve their reach. But after they pay for the data, the DSPs work directly with DMPs, and no one knows whether the data is used or not.
You seem to be arguing for the meta-DSP model, but how can that be adapted to China?
Despite the name, a meta-DSP is not actually a DSP; it is an advanced trading and management platform that integrates multiple programmatic technologies, or a trading desk, from the agency perspective.
Campaign Asia-Pacific welcomes further input from companies mentioned in the other conflict scenarios in our original piece on this topic. Please write to email@example.com.