The conversation was recorded in mid-April at a meeting room above a coffee shop near Old Street. The head of video — she asked to be called Ansel, and to have the specific company remain unnamed — runs an in-house video team of eleven people at a UK financial-services company, and has been in the role for four and a half years. The metric she runs her programme on, and which she was willing to describe in detail for this piece, is one she has never seen used by any of her peers in comparable roles. She thinks this is a mistake on the industry's part.
The metric
Isla Merriweather: Take me back to when you started running this metric. What was the trigger?
Ansel: Something specific in mid-2022. I had been in this role for about two years at that point, and we were reviewing a set of pieces we had shipped over the previous quarter. One of them was a substantial commissioning — a long piece, about eight minutes, that had a specific commercial ask embedded roughly two-thirds of the way through. The piece had performed well on our standard metrics — high view count, decent completion rate, positive engagement — and had been reported to the leadership team as a success.
What bothered me, when I looked at the audience-retention curve, was the specific shape of the curve past the commercial ask. Approximately 65% of viewers who reached the commercial ask kept watching for at least another minute. Approximately 40% watched to the end. The retention past the pitch was, by any reasonable measure, the strongest indicator that the pitch had landed — someone who kept watching past the ask, rather than dropping off at it, was demonstrably still engaged with what the piece was doing.
None of our reporting captured this. We reported view count. We reported completion rate against the full length of the piece. We reported engagement events. We did not report — anywhere, on any dashboard — the specific behaviour of the audience at the moment the piece made its commercial argument. I started asking why not.
The answer, when I worked through it, was that most video reporting is built around the assumption that the piece is either a commercial or an editorial piece, and is measured accordingly. Commercials are measured on their ability to hold attention through a short window — usually 30 or 60 seconds — because that is what commercials are structured to do. Editorial pieces are measured on watch time in aggregate because they do not have a specific commercial ask to test against. Long-form pieces with embedded commercial arguments, which is what our function increasingly commissions, don't fit either model. The industry's default metrics don't measure the specific thing we're doing.
What she built
Isla: What did you build in response?
Ansel: A specific set of retention indicators keyed to the position of the commercial ask within each piece. For every video we ship, we identify the specific timestamp at which the piece makes its primary commercial argument. We then measure the audience retention curve at that timestamp and in the specific 30-, 60-, and 180-second windows after it. We aggregate these numbers across pieces to get a sense of how our programme, as a whole, is performing on the metric that we think actually matters.
The numbers are, in aggregate, more useful for our decision-making than the standard metrics were. A piece with strong retention through the ask, and strong retention in the minute after it, is a piece where the audience has stayed with the argument the piece was making. A piece with a sharp drop at the ask, or in the minute after it, is a piece where the argument did not land. The distinction is much sharper than "did the audience complete the piece" — which mixes together audiences that engaged with the commercial argument, audiences that were watching for other reasons, and audiences who simply forgot to stop the video.
Isla: What have you learned from running this metric for four years?
Ansel: A few things.
The first is that the specific position of the commercial ask within the piece matters enormously. Pieces that place the ask past the two-thirds mark of the total length consistently produce stronger retention past the ask than pieces that place the ask earlier. There is, on our four years of data, a specific structural pattern that produces better outcomes: build the substantive editorial argument first, embed the commercial ask as a natural consequence of the argument, allow the piece to continue past the ask for enough length to signal that the ask was not the point.
The second is that the specific creative form of the ask matters much less than the industry assumes. Whether the ask is a formal call to action, a natural mention within a broader argument, or a subtle visual reference, the retention past the ask is dominated by whether the ask fits the piece's overall editorial voice. If the ask is a stylistic break from the rest of the piece — the standard "and now a word from our sponsor" transition — the audience drops sharply. If the ask is continuous with the piece's voice, the audience mostly stays.
The third is that pieces which fail on this metric are, almost universally, pieces we should not have shipped. On post-mortem review, the pieces with sharp retention drops at the ask are, in almost every case, pieces where our editorial team had internal doubts about the fit between the substantive content and the commercial argument. The metric surfaced problems we had, on inspection, half-noticed at the commissioning stage but not confronted.
"The industry's default video metrics measure attention. My metric measures whether the attention was still with us at the moment the argument was made. Attention without argument is a vanity metric. Argument without attention is a private conversation. The specific combination — attention that persisted through the argument — is what commercial video is trying to produce. It is worth measuring."
What her team does differently
Isla: How does running this metric change what your team ships?
Ansel: The composition of the commissioning schedule has shifted substantially. We ship fewer pieces than we used to, at longer average lengths, with commercial arguments more consistently placed past the two-thirds mark. The individual pieces are more editorially substantive than they were three years ago, because the pieces need to earn their length through the substance of what they're saying rather than through the promise of a punchy conclusion.
The team's briefing document has, over the last three years, evolved to include specific requirements around where the commercial ask will sit within the piece and what the piece will do after the ask. This is now a required element of every commission we run. Directors who are new to working with us occasionally push back on it in early conversations. Directors we have worked with for a while have, on the whole, come to see the constraint as productive — it forces a specific kind of structural thinking about the piece that they think is worth doing.
Isla: Last question. Would you recommend other heads of video build a similar metric?
Ansel: For any team commissioning long-form video with embedded commercial arguments, yes. For teams working exclusively in short-form or in pure editorial, probably not — the metric doesn't fit those formats as cleanly. But the specific category of "long-form commercial video with substantive editorial content" is, in my strong view, an under-measured category by the industry's default metrics. Teams working in it will produce better commissioning decisions if they measure what they are actually trying to produce, rather than what the platform reports on by default. This is, in the end, a fairly obvious point. The industry has been slow to act on it because building your own custom retention metrics requires a specific analytics investment that most teams have not made. That investment is worth it, on our four years of experience.