Most placements we bought did nothing
On the 2026-09-02 Supply Bench pull, 93.4 percent of placements had zero clicks and still took 64 percent of observed spend.
Published 2026-09-09 · Search job: why programmatic placements get no clicks
Most programmatic placements get no clicks because remnant display is a long tail of cheap inventory, not a short list of working slots. On the 2026-09-02 Supply Bench pull of inventory_agg and domain_agg for Buzz campaigns 4791 through 4810, 1,073 of 1,149 placements (93.4 percent) recorded zero clicks and still took $202.77, which is 64 percent of observed spend.
That is the search job answered with a dated window. The rest of this page is how we know, what the pull could not see, and why a zero-click placement is not the same thing as a failed creative.
What we pulled
The report covered both Supply Bench waves. The placement view returned 1,149 placements. The domain view returned 1,535 domains. The top ten domains took 31.9 percent of spend, about $0.215 per domain across the rest of the list.
Coverage was partial on purpose, not by accident. This seat ignores report offset. A limit of 50,000 returns HTTP 400. Filters on campaign_id fail. The usable pull is the top roughly 5,000 rows sorted by spend, which captured $316.64 of a combined total near $432. Every share on this page is biased toward larger placements. That ceiling is part of the finding.
Returned rows use label-cased keys. campaign_id arrives as Campaign ID. A parser that looks for the snake_case name you requested will read every metric as zero and look exactly like a buy that did nothing. See Why a dashboard can show zero clicks before you treat a blank column as truth.
What zero clicks means here
Zero clicks on 1,073 placements is not proof that nobody saw the ad. It is proof that those placements produced no billed exchange clicks in the slice we could read. Exchange clicks are not page views. Click tokens are not delivery truth. We had no IVT vendor attached, so invalid_impressions staying at zero means nothing was measuring, not that the traffic was clean.
The placements that did click were not evenly spread. Seven domains produced 41.3 percent of all clicks from 4.7 percent of impressions. That concentration is the other half of the same pull, written up in Seven domains produced two fifths of the clicks. A flight can look ordinary on blended CTR while almost all of the click signal sits on a handful of hosts.
What we will not claim
We will not call the 1,073 placements fraud. We will not call them wasted creative. We will not treat $202.77 as a certified invalid-traffic bill. We will not pretend the missing $115 beyond the $316.64 slice would have looked different. It might have. We could not page it.
We also will not pretend we could have named those placements before the flight. You cannot pick the exclusion list first. That is why Scan then exclude is the operating rule, not a slogan.
What we do with this
Judge remnant display on concentration, click-to-page-view once the join exists, and strict IVT once a vendor add-on is on the creative. Do not judge it on CTR. CTR is a trap is the short version of that sentence.
On this seat, placement and site are direct targeting keys. Domain exclusions still need list objects. The grain you can act on after a scan is narrower than the grain the report shows. Map that limit before you promise a cleanup. The grain map is Open exchange waste, mapped by grain.