The brand is a mid-market UK DTC retailer selling into a specific consumer category — I have agreed not to name it — with roughly £38m of annual revenue and a marketing team of six. The head of paid media, whose CTV programme this is, took the role in early 2024. She inherited a CTV allocation running at approximately £280,000 per year, negotiated eighteen months earlier by the previous head of paid, against a mix of premium streaming inventory (broadcaster VOD, ad-supported streaming service inventory, connected-TV publisher deals) at CPMs ranging from £32 to £58.
The reported outcomes on the account had been, throughout 2024, "middling but defensible" — her words. Reach was substantial, brand-tracking indicators had marginally improved during the campaign windows, and the direct-response attribution reported by the CTV DSP was in the same neighbourhood as the brand's paid social attribution. Nobody was celebrating; nobody was complaining. The line item persisted.
The audit
She began the audit in early 2025, prompted by a specific piece of finance-team analysis showing that the CTV line's contribution margin per pound of spend was the weakest of any paid line in the account. The audit ran over six weeks and produced a set of findings that were, in her later summary, "worse than I had budgeted for".
The first finding was that the audience-quality argument on which the premium CPM had been negotiated was, on inspection, thin. The publisher-side CTV deals were priced against a claim of "premium audience — high household income, mid-range family, engaged with content" that had been derived, on the vendor's own subsequent admission when pressed, from small survey panels rather than from any direct audience-verification data. The brand had, on inspection, been paying for an audience that the publisher could describe but could not, in any actionable way, verify.
The second finding was more concrete. The DSP-reported delivery had allocated substantially more impressions to daytime and off-peak content than the premium claim had implied. Roughly 42% of the brand's CTV impressions had been delivered during weekday daytime slots, against publisher content that was priced at prime-time-equivalent CPMs. The impression cost per household reached had been substantially higher than the premium tier's marketing had suggested it would be.
The third finding was about attribution. The direct-response attribution the CTV DSP reported turned out, on closer inspection, to be largely proxy-attribution — the DSP was matching impressions to subsequent conversions on the brand's site using cookie or IP-based signals of uncertain reliability. When the brand's MMM partner ran an independent geo-holdout analysis, the actual incremental effect of the CTV spend was approximately a third of the direct-response-attributed effect. The reported ROAS had been meaningfully more flattering than the reality.
"The line was not producing zero value. It was producing perhaps 30-40% of the value the reporting suggested. At the CPMs we were paying, that ratio was upside-down. The category is genuinely valuable at the right price. We were not paying the right price."
What she did
The head of paid presented the audit findings to her CMO in April 2025. The proposal was not to exit CTV entirely — the brand-tracking signal was genuine, and the head of paid believed the category had a legitimate role in the mix — but to renegotiate the entire CTV allocation with substantially lower CPM ceilings, tighter audience targeting, and explicit attribution requirements.
The renegotiation, executed over the following three months, produced three main outcomes. First, the aggregate CPM on the CTV line fell from a blended £42 to a blended £24, primarily by dropping several of the premium publisher deals and shifting spend toward CTV DSP-based buying against explicit audience targeting. Second, the audience-verification data requested from remaining publishers had, in some cases, been better than the publishers had led the brand to believe (there were legitimate audiences behind the marketing claim, once the buyer knew to ask for the underlying data); in other cases, the publishers declined to share the underlying data and were dropped from the roster. Third, the account's monthly attribution reporting was moved to a MMM-driven view rather than the DSP-reported view, with the DSP report retained as a diagnostic rather than as a decision-driving number.
Total CTV spend, at full re-implementation, fell from £280,000 to approximately £165,000. Reported impressions were down proportionally. Reported reach was up slightly (the tightened audience targeting produced more unique reach per impression than the previous allocation had). MMM-attributed contribution margin on the CTV line rose from £0.90 per pound of spend to £2.10 per pound of spend — well over the account average.
What the category still gets wrong
The head of paid's view, and one we broadly share, is that CTV as a category is currently priced above its actual audience-quality delivery in a substantial share of publisher-side inventory. The claim on the CTV audience — that it is more attentive, more affluent, more engaged than legacy display — is, in aggregate, defensible. The claim on the specific premium tier of CTV inventory — that it commands a further premium over the CTV baseline because of specific content quality or audience concentration — is, on the audit sample we have run, much harder to sustain against evidence.
Two specific things are worth being cautious about, if you are currently allocating CTV budget or planning to.
The first is that most CTV audience claims are not, currently, backed by verification data of the sort a mature programmatic buyer would demand from an open-exchange publisher. Ask for the data. If the publisher cannot provide it, treat the audience claim as marketing rather than as an argument for a specific CPM.
The second is that most CTV attribution reporting is more flattering than the underlying incrementality. If your CTV programme is being justified primarily on DSP-reported attribution, run at least one MMM-triangulation or geo-holdout to sanity-check the number. The gap, on the accounts we have audited, is consistently in the direction of the reporting being generous.
CTV is a real category. It is growing. It has real audience value. Buyers who allocate against it with the same rigour they now bring to open-exchange display, on our sample, do well. Buyers who allocate against it on the premium narrative alone, without demanding the verification, will produce accounts that look like this one did — impressive on the dashboard, disappointing on the finance model, and worth substantially less than the CFO has been told they cost.