The story. Performance CTV is the pitch that connected TV now behaves like search or social: a pixel, an attribution window and a CPA column. An agency executive argues that it does not, and offers one campaign where two methods disagreed by more than tenfold. (AdExchanger, 13 October 2026)
01What happened
- One campaign, two counts. David Nyurenberg of InterMedia Advertising writes that a DSP pixel recorded 109 completed checkouts. Matching impression IPs to the advertiser’s CRM logs found 1,250 conversions.
- The gap is about 11x. That ratio is APH arithmetic on his two figures. The column does not name the advertiser, the DSP or the campaign dates.
- IP mapping is weak for geo targeting. InterMedia and Adstra evaluated residential IPs for ZIP-code targeting, and 23% reached the intended area. He says this is not a CTV success rate, and that device-level CTV identifiers were considerably more stable.
- The household is on an iPhone. For one major direct-to-consumer advertiser, 63% of purchases happen on an iPhone. He ties that to Apple’s iCloud Private Relay and the latest WebKit restrictions on ad tech and identity providers.
- Last touch flatters search. When a viewer sees a CTV ad, searches the brand and buys, search gets the credit. His point is that TV’s job of creating demand goes unpaid.
02What it means inside a GAM network
The 109-versus-1,250 gap cuts against the sell side as much as the buy side. If a campaign’s conversions are undercounted by an order of magnitude, a buyer reading the DSP report will conclude your inventory underperformed. Your floors then get pressure for reasons that sit in the measurement chain, not in the ad.
The chain has more links than the pitch admits. The ad plays on a television, the purchase happens on a phone, and something has to join the two. That join depends on IP signals, device graphs and identity providers, and the column notes that publishers and platforms send inconsistent signals with different timing for conversion data. Apple’s WebKit changes land on the phone end of that join.
A note on scope: this is an opinion column from an agency, and the 109 and 1,250 figures are from one campaign, unnamed. They show that attribution can miss by a lot, not how often or by how much on yours. The CRM match is also its own method with its own error, so 1,250 is better read as a signal that the pixel undercounts than as a truth.
Two things follow for deal design. Buyers who optimize toward a CPA column will starve CTV supply that the pixel cannot see, which is often the household-level inventory that works. And buyers who accept incrementality tests and holdouts will pay for reach and attention you can show.
CTV can sell outcomes, but a pixel, a window and a CPA column measure only the part of the household the pixel can see.
03What publishers should do about it
04The bottom line
CTV can sell outcomes, but a pixel, a window and a CPA column measure only the part of the household the pixel can see. Offer buyers a measurement path that covers the rest, and keep the pitch on the job the television does.