Auction trends from BidSwitch

BidSwitch sits in the middle of a lot of the programmatic market, so I think their data is worth looking at.

Interview: Cadi Jones from Index Exchange

Marketecture Vendor interviews are free for a week, then require a subscription. Read more about subscription options at marketecture.tv.

Marketecture was in London at the IAB Tech Lab’s Summit. I got the chance to catch up with my former employee and now GM EMEA for Index Exchange, Cadi Jones.

Marketecture Podcast: Marc Guldimann

Marc from Adelaide gave us a deep dive on everything “attention” — uses, methodologies, adoptions. If you’re interested in metrics you should listen to this one.

The QPS is too damn high

This week I had the pleasure to attend and speak at the BidSwitch Summit in New York. I did my best Zach Galifianakis impression and sat between two actual ferns (plants of some kind) interviewing Joe Zawadzki. Joe asked me if I had a time machine, would I go back and kill baby Hitler, and I said, no, I’d instead put a price tag on ever escalating levels of QPS.

Anyhoo, Barry Adams from BidSwitch gave a keynote where he showed some really interesting stats on the volume of auctions running through the platform and he generously allowed me to re-post here with some commentary. If you don’t know, BidSwitch is sort of a meta-SSP that connects DSPs and exchanges but doesn’t actually act as a principal on either side of the equation. Their position in the market gives them a great deal of visibility into volume trends in the industry.

First, overall QPS continues its inexorable rise, though SSP consolidation means fewer parties are involved on the sell-side. This makes sense as we’ve seen SSPs like Yahoo and EMX exit the market.

The streams crossed! For the first time, app QPS exceeded web, and web was down year-over-year. It is hard to tease out of this data where the big trends are versus blips that are specific to BidSwitch or certain short-term trends. But some things are clearly happening in the market:

• Actual visitation to web-based properties had likely passed its peak. We know that search and social referrals are down, and unlikely to (ever) come back up.

• The app market remains less efficient that web, as “unified auctions” from companies like AppLovin are not as well adopted as header bidding. This allows less efficient paths to buying to continue to proliferate.

• CTV continues growing, though despite the lack of y-axes, Barry verbally indicated that this is a very small volume. Also, in all likelihood, fewer CTV auctions make their way to BidSwitch than go direct in some fashion.

A reseller renaissance? One of the expected outcomes of ads.txt adoption was a reduction in the number of resellers. Instead, what seems to have happened is that all the resellers bullied their publishers into listing them as “direct”, then the buyers took notice and they quietly reclassified themselves as “resellers.” Thus, we have the below graph showing a pincer pattern between direct and reseller. In any case I think we can all be very happy about the low incidence of “unknown” and “unauthorized,” time to pat yourself on the back, ad tech industry.

The next graph is quite disturbing, please make your children leave the room before continuing. The graph shows that buyers are bidding equally for auctions that are resold multiple times versus those that are direct. Notice the eCPM remains roughly unchanged across the number of “hops.” Honestly, I’m having a hard time with this data, given how almost every legit DSP tried to prune inventory that’s resold in this way.

On the more positive side, the adoption of TID, or Transaction ID, is solid, at close to 50%. With wide scale adoption of TID you could much more easily eliminate duplication, should you choose to.

Someone once described Deal IDs to me as the WD-40 of ad tech. They are the universal solvent that just makes any data flow through the pipes. Perhaps Ex-Lax would be a better metaphor. In either case, BidSwitch found that 70% of bid requests had at least one Deal ID, and that some requests had up to 300! It should be no surprise that the prevalence of deals is growing as they are the primary mechanism by which curation — the hot topic du jour — takes place.

It is unclear how representative the BidSwitch data is versus the market as a whole. If you consider the very largest buyers and sellers, those connections are unlikely to move through BidSwitch. So even if you assume their data is biased towards smaller buyers and sellers, you still see some interesting trends that are, as the kids like to say, narrative violations. Reselling is still a viable path, as are multi-hop selling. Deals are becoming a dominant form of data resolution. And in-app continues to grow, while web has maybe peaked, maybe forever.

Thanks again to Barry Adams and the BidSwitch crew for letting me use their slides.

Reading list

• TTD and Magnite both touting CTV (link)

• Culper Research with extensive accusations that Zeta is a scam of some sort (link, pdf)

• Another blog post about Zeta accusing it of shady stuff (link)

• Viant acquires IrisTV (link)

• Forrester SSP survey is a joke (ATG tweet w/ summary and image)

• Perplexity: ads are coming (link)