CTR is a trap

Click-through rate is the first number everyone reads and the worst number we have used to judge remnant display.

A 300x250 on the open exchange can produce a CTR that looks ordinary, even healthy, while the clicks come from a handful of domains that no real person would click at that rate. The June 2026 learning flight was read on clicks and CTR. That readout later had to be thrown out. The later Supply Bench pull made the same point with grain we did not have in June.

What the later pull actually showed

On 2026-09-02 we pulled inventory_agg and domain_agg for Beeswax campaigns 4791 through 4810. The placement view covered the top spend slice of both Supply Bench waves, about $316.64 of a roughly $432 combined total, because this seat ignores report offset and 50,000-row limits 400. So every share below is biased toward larger placements. That limit is part of the finding, not a footnote.

Inside that slice:

• 1,149 placements and 1,535 domains.
• 1,073 placements, 93.4 percent, had zero clicks and took $202.77, 64 percent of observed spend.
• Seven domains produced 41.3 percent of all clicks from 4.7 percent of impressions.
• people.com showed 144 impressions and 23 clicks, a 15.97 percent CTR.
• msn.com showed 785 impressions and 42 clicks, a 5.35 percent CTR. The July 2026 placement scan had already flagged it at about 3.7 percent against a 0.15 percent flight average.
• Reseller inventory clicked 2.3 times more than Authorized-Direct (0.423 percent vs 0.184 percent).
• SSP CTR on the same creative and format ran from OPENX at 0.111 percent to SMARTADSERVER at 3.074 percent, with THIRTY_THREE_ACROSS at 18.18 percent.

A 0.47 percent flight CTR and a $1.04 CPC, which is what the current fleet showed when we wrote the trafficking playbook, look unremarkable. They tell you nothing about those seven domains.

What we will not claim

We did not have an IVT vendor attached. invalid_impressions and viewable_impressions were zero because nothing was measuring, not because the traffic was clean. The 1 percent and 3 percent CTR thresholds we used to flag contamination are our reading, not a certified fraud verdict.

We also will not treat billed clicks as people. Exchange clicks, click-token redirects, landing visits, and on-screen interaction are four different meters. Collapsing them is how a dashboard shows "engagement" that never reached a page.

What we use instead

Judge remnant display on:

1. Click-to-page-view once the click-to-visit join is wired. Anonymized main-seat brand-name display flights we can cite without naming the advertiser sat at 15.85 percent when the traffic was usable and 3.75 to 6.68 percent when it was not.
2. Strict IVT rate once a vendor add-on is on the creative. The same anonymized flights sat at 0.523 percent when acceptable and 1.544 percent when they triggered intervention.
3. Concentration. Seven domains, 41 percent of clicks. That shape is the story.

Until measurement is attached, say so. Zero invalid impressions is not a quality score.

Why this note exists

CTR is easy to put on a dashboard. It is also the metric leftover inventory is best at inflating. If you only remember one sentence from this publication, remember this one: a normal CTR on the open exchange is not evidence that the buy is working.