US sports publisher × Adverge · head-to-head vs. the prior setup
+19.15%
uplift against a stack the whole industry already treats as the default.
A US sports publisher ran Adverge head-to-head against its incumbent platform — one of the largest and best-known monetization partners in the market.
Adverge came out 19.15% ahead.
It did not win by adding demand partners the incumbent did not have. It won on what it knew about each session before the auction opened.
One signal source. Two places it changed the outcome.
The publisher was already collecting first-party signals on its own audience. They were sitting in the property rather than reaching the auction. Adverge put them to work twice.
The input
First-party signals the publisher already owned
Logged-in state, return frequency and on-site behavior — data the publisher had, and which the incumbent setup was not fully putting in front of buyers.
Use one
Sent to demand before the auction
Buyers priced a known reader instead of an anonymous impression. The same partners bid differently once they could see who they were bidding on.
Use two
Refresh decided by those signals
Refresh fired against observed engagement rather than a fixed timer — more real impressions from readers who were still there, none manufactured from readers who were not.
What we actually changed.
Two changes, both drawing on the same underused asset.
01
Better use of first-party signals
The publisher’s own audience data was synced to demand partners ahead of the auction rather than left inside the property. An incumbent stack tends to treat every session the same way because that is the safe default at scale; a sports audience is not average, and the bid should not be either.
02
Refresh optimization driven by that data
Refresh timing was set from first-party engagement signals instead of a clock. On a sports property attention is not evenly distributed — it concentrates hard around fixtures and results. Reading that from the publisher’s own data means an impression is called when the reader is still present, not on a schedule that assumes they are.
Scale is not the same thing as attention. Only one of them can read the session.
A large incumbent brings demand relationships and volume. What it does not bring is a view of this publisher’s individual reader. That view already existed here — it simply was not reaching the auction, and connecting it was worth 19.15%.
Why sports inventory rewards this.
Sports traffic arrives in bursts and leaves just as fast. A fixture, a trade, a result — sessions spike, run deep for a short window, then collapse. A stack tuned on averages prices the spike like the trough and refreshes through both at the same cadence. First-party signals are what tell the auction which of the two it is currently in.
Figure is the publisher’s own reported uplift for Adverge against its incumbent platform in a head-to-head test. Neither the publisher nor the incumbent platform is named here. This account describes the changes Adverge made; the daily test data sits with the publisher.
Run the same test on your traffic.
Adverge tests head-to-head against your incumbent on live inventory, measured daily. If your first-party data is sitting in the property instead of reaching the auction, a month of side-by-side data will show you what it is worth.