Q&A - Pau Quevedov-final.jpg

On this occasion, we had the amazing opportunity to sit down with Pau Quevedo, the talented Lead of Programmatic Trading at Goodgame Studios. For this interview, Pau shared with us tons of insights about programmatic, his incredibly honest opinions about the current state of the market, and where he thinks everything is headed. It is definitely a must-read for everyone interested in marketing, DSPs, and programmatic buying - check it out!

Question 1: Pau, could you tell us how did you get started in mobile marketing, and specifically - programmatic buying?

Answer: I started out in mobile marketing over 6 years ago but I didn’t start with programmatic until 4 years ago when I had the opportunity to buy for our Desktop and Mobile titles at Goodgame Studios.

Question 2: What is the most important and challenging thing in mobile marketing?

Answer: In my honest opinion, User Acquisition, outside of the Google-Facebook duopoly, remains the hardest nut to crack on DSP Mobile. The big channels have reached a considerable level of automation and they are now focused on the creative aspects of it all while in Mobile DSP we are still playing catch up.

Question 3: As DSPs require a lot of tweaking to optimize audience segmentation and personalization of messages, what are your go-to sources to enrich data for DSP modeling besides log-level data and in-game data?

Answer: If we want to use contextual data, we enrich our models and targetings with Publisher IDs from other partners that are successful, although this doesn’t normally work since behavioral data is commonly more valuable. For behavioral data, we reach out to DSPs or Data Providers that can supply us with consumer behavior data, especially purchasing power. Some DSPs own this data and the only way to access it is by using their DSPs, which is something that we are currently testing with the big ISP from the US.

Unfortunately, most of this data is only usable in the US and sometimes, it is not recent enough to be valuable. Some networks/exchanges have important data regarding the users, and you can get deals that already include those targeting - but so far, they are not so focused on purchasing patterns but more interested in playing activity and other retention metrics.

Question 4: What do you do to protect your data? Is there a way to prevent DSPs from using your data to benefit competitors? Are contractual limitations of user reselling enough or are there more preemptive measures you can take?

Answer: This is a tricky question. In a nutshell, every DSP out there shares the learnings from your data with other advertisers, since this is how they provide a competitive service during the exploration of campaigns. Only specific products like Bidders, for example, can isolate your data and not share it, but this is not usually recommended during the prospecting phase unless you have very specific needs with very clear targetings. There are very few products in the market that can provide this service. Big companies with large pools of data prefer to build their own DSP or work with a Bidder instead of plugging their DMP into someone else’s ML.

As far as contracts, we enable them to bind us in terms of controller and processor of data. We can forbid them to directly use our data with our competitors but what we cannot forbid is that the learnings of the algorithms be shared across different advertisers. The mobile identifier itself is not so important, they can still retarget that user as the learnings that ML can extract out of our campaigns.

Question 5: How do you evaluate the quality of inventory that DSPs offer besides taking a hard look at QPS data?

Answer: QPS data doesn’t tell you much about the quality, or even the amount, of bids the DSP actually listens to, and there are very different needs if we are talking about banner or video. In rewarded videos, which is a typical placement for Gaming, you don't need much QPS to actually listen to all the traffic. For this reason, small DSPs that are focused on gaming could potentially listen to all relevant traffic. On the other hand, big DSPs normally don't have direct connections with Rewarded Video networks since they don’t have many clients that buy on that traffic since brands find it too expensive for them.

Some omnichannel DSPs use a 3rd party infrastructure to throttle their QPS for rewarded video. We know from the Ad Monetization side of things that this is not ideal when buying large amounts of traffic since 3rd party infrastructures only get a fraction of the traffic that other direct Networks get. For this reason, omnichannel DSPs are normally not ideal for mobile UA, unless they have some great audiences or premium inventory you otherwise can’t get.

For us to tell if a DSP has relevant traffic for us, we normally ask about the clients they have, we want to see if they are already listening to certain traffic we’re interested in. We’ve had some surprises, normally coming from big DSPs who are not connected, not even via 3rd party infrastructures. On the other hand, smaller DSPs don’t have banner traffic in many apps or GEOs. Because of this, if you want to do a very holistic approach by targeting users in all kinds of traffic, you might not find a DSP that covers all your needs.

Seeing this problem, what some DSPs agree on is to help us by booking some traffic just for us, but that normally takes time and you have to build a strong partnership for this to happen on a regular basis.

Question 6: We all know that fraud is present, but do DSPs really care about fraud? Are they truly committed to implementing anti-fraud measures?

Answer: It all depends on what source we buy our traffic. Mobile UA for gaming is largely focused on Rewarded Video. Although there’s fraud everywhere, we see that the networks already do some cleaning beforehand and we don't see much fraud, especially if you compare it with affiliates or such. Visibility or brand safety are topics that rarely come across in Mobile UA, whereas for Desktop is daily business.

In general, fraud is a topic that is not very discussed but everyone is aware of it. We rely on MMPs or specific Fraud detection partners, and so far, we haven’t seen large amounts of fraud in comparison with other channels. As budgets increase in the years to come, fraud is a topic that will definitely grow.

Attribution remains an important topic in all this, most DSPs are not self-attributing networks and we’ve seen issues with assisting installs, especially to incentivized traffic partners. For this reason, we are testing a SAN DSP to see if we notice any differences in this regard.

Question 7: If any, what would you say are the industry standards for ad inventory?

Answer: For Desktop, we use IAB standards, which they also have for Mobile banners. For gaming, we mostly use video with nice stuff on top. More and more DSP departments are hiring coders just to focus on creating fun and engaging creatives since this plays such a big role.

But when we speak of standardization of Ad Inventory, since each SDK has a video player and this determines the ad behavior, we see a lot of different video players that affect the player experience. This also has an effect on the engagements we get from users. Some video players send clicks in the middle of the video, and unfortunately, it's not even clear what a click is considered since each player might fire them when or how they want. This is really bad from an ML perspective since we are getting very different signals when in reality the behavior behind them is not that different.

We are hoping to see some standardization in this regard in the years to come.

Question 8: If you had a magic wand, what would your ideal DSP look like? And a follow-up question: If you waved your wand and chose for a highly analytics-driven and reporting-oriented platform, what type of modeling and visualization would you like to see?

Answer: The most important element is the way the ML operates. We currently only onboard DSPs that are user-level centered, so this would be the first requirement.

I would like to see a DSP that is very transparent in the way the auction works, enabling us advertisers to easily structure our log-level data and allow us to capitalize on it easily. Ideally, the DSP could provide an ID graph so traders can focus on delivering the right message and at the right time. We are trying to build our own models so the ideal DSP must support BYOA (Bring Your Own Algorithm) and also some accessible language where we can easily generate bidding and targeting strategies.

Having those modeling tools on the DSP itself would help us to focus on outlining strategies and focusing on the big picture.

On the creative side of things, DSPs don’t give you transparency on how your ad will be displayed, in the case of mobile, ideally, we could see the video player of the SDK where the ad will be shown so we can adapt our creatives perfectly. There’s no DSP that provides this and we have to manually do this for each network.

Also, having a tool that allows us to access Metamarkets is very important, as it will help us clearly understand where the supply is so we can define our goals. Many DSPs provide some sort of view, but normally it is limited and it's hard to relate that information to your campaigns. For this, some sort of actionable Metamarkets dashboard would be a dream come true.

Question 9: When thinking specifically about scaling, what is the most important thing to look at in a DSP partner? Algorithms, the pool of publishers, pricing models, and bidding?

Answer: Unfortunately, the most important aspect of a DSP when it comes to performance is the bidding process, which is the battlefield for the different DSPs and it's where they are spending most of their resources to differentiate themselves.

Obviously, you also need traffic sources to buy from but these, unless they are premium, are just a matter of time and priorities until they can get to listen to that traffic.

Pricing models are very important as well. Before, we used to see many DSPs that applied a fee on top of the bid you made, making the whole process very intransparent since many bids take place and this could hinder the amount you wanted to bid. Now, new DSPs are pricing in terms of media spend or even QPS.

Question 10: What do you do to successfully coordinate multiple DSPs? Both managed and self-served?

Answer: Usually, we are used to running several DSPs at the same time, no more than 3 self-served and a few managed tools are recommended at the same time. Since there's not a one-size-fits-all solution, we have to use different tools for different purposes, GEOs, formats or games. We try to automate things as much as possible and the maintenance downtime is generally not very big.

In the past, I tried to use learnings from one DSP into another, but this normally failed since the good DSPs are using user-level data in their optimizations and this data you don’t get, so applying lists of publishers from one DSP to another is most likely not going to work as expected. It can help during prospecting phases though.

So far, we spend most of our time setting up workflows in different DSPs but we would like to spend more time optimizing creative processes, this is the part where DSPs still have a lot of room to improve.

Question 11: With the current state of uncertainty in the world, do you see a shift in the horizon for the way marketers re-engage or attract users? Do you see things like DSPs taking a backseat?

Answer: DSPs are the best channel to capitalize on data sets. As the industry moves forward and the publishers understand the need to segment their own data and enrich it even further, DSPs will greatly benefit from these circumstances. Unfortunately, for mobile gaming, the industry is still shifting from an old contextual bulk-selling selling process to a more complex user-centered approach. At the moment, we have several players within programmatic that are comfortable with their position that they have no incentive to move or shift towards a user-level centered strategy, for instance, in-app bidding. This is slowing things down and it's unclear how this new context will affect this process. Until this shift doesn't happen, DSPs are not in a favorable position regarding the auction, greatly limiting our scaling possibilities.

Again, we would like to thank Pau for his time, commitment, and willingness to share his wisdom with us - not only did we learn heaps and bounds about programmatic and DSPs, we thoroughly enjoyed his way of explaining everything.