Same creative, 28x CTR spread by SSP

SSP quality differs on remnant display because the supply behind each seat is not the same inventory, even when the creative is. On the 2026-09-02 Supply Bench pull, the same creative and format produced 0.111 percent CTR on OPENX and 3.074 percent CTR on SMARTADSERVER. That is a 28x spread. THIRTY_THREE_ACROSS sat at 18.18 percent. OPENX took 44 percent of spend, so the low-CTR path was also the expensive path.

If you are asking why one SSP "performs" and another does not, this flight says the creative is the wrong variable. We held it still. The exchange did not.

What stayed the same

Campaigns 4791 through 4810 flew one creative family and one format across open-exchange supply. Targeting did not pick a short SSP list. The report then broke delivery by inventory source. CTR is billed clicks over impressions inside each source. The numbers above are those rates, not modeled conversions and not page views.

OPENX at 0.111 percent is below the flight average of about 0.28 percent. SMARTADSERVER at 3.074 percent is about eleven times that average. THIRTY_THREE_ACROSS at 18.18 percent is in the same implausible band as people.com at 15.97 percent on the domain table. A source that clicks like that is not a win. It is a reason to stop using CTR as the score. See CTR is a trap.

What we will not claim

We will not rank SSPs as good or bad from CTR. A high rate can be contamination. A low rate can be ignored inventory, a weak slot, or a source that simply does not mint accidental clicks. We had no IVT vendor attached, so we cannot convert any of these rates into a strict invalid-traffic number.

We will not say OPENX is clean because it clicked less. OPENX took 44 percent of spend inside a pull where 1,073 placements had zero clicks and seven domains produced 41.3 percent of all clicks. Low blended CTR on a large source can hide a contaminated subset. Grain still matters. Seven domains produced two fifths of the clicks is that grain.

We will not treat inventory source as a private marketplace. These are open-exchange SSPs. Deals are a different object. This child seat has historically had zero deals. That comparison is a plan, not a result.

What the seat can do with an SSP

inventory_source is an allowlist. Excluding one exchange means restating every other one you still want. An empty allowlist compiles to match nothing and the apply path returns 409. Details sit in Why an empty allowlist serves nothing.

A hard exclude is not the only future lever. Bid modifiers on this seat can scale a bid on inventory_source instead of removing it. That path is not productized yet. See Bid down a bad placement instead of killing it.

Until measurement is attached, the honest sentence is: same creative, 28x CTR spread by SSP, OPENX paid for most of the impressions, and CTR still does not tell us which source a person actually used.