A creator video does not sell anything by itself. It starts a chain. Someone stops scrolling, watches, hands over an email, buys the product, uploads the receipt. Every link in that chain either carries the effect forward or loses it.
We measured all four links across 157 Meta video ads run by CPG brands driving to retail. Three of the four wash out. One carries, and it is not the one most brands are watching.
Every hook in the dataset was coded by structure first, then joined to verified performance running from impressions through to receipts. Each figure is measured per distinct hook rather than per ad, and compared against that hook's own brand median. Here is the chain, link by link:
Read those in order and a picture emerges. Stopping the scroll and keeping someone watching are not the same mechanism. A video that stops more people does not thereby keep more of them watching. Extended watch time does not guarantee an email capture either, even though the offer appears in the middle of the video, which is exactly where a longer watch should pay off.
And measured straight through from the opening to the email captured, hook rate does not separate cheap-lead videos from expensive ones at all.
The one link that holds is cost per lead to cost per receipt, at +0.701 with a p-value below 0.001. The brands acquiring leads cheaply are the brands acquiring sales cheaply.
That single surviving link is the argument for judging creative on cost per lead, and there are three reasons it holds up.
It is the last stage the video still moves. It reads within days of launch rather than weeks. And it correlates with eventual purchase at the brand level.
One honest caution belongs with that. The cost per lead to cost per receipt relationship is strong at the brand level but weaker within individual brands, at +0.223 and p=0.10. So cost per lead works as a fast directional metric, not as a per-video sales predictor. By the time a purchase actually occurs, brand-level forces have intervened.
Which raises the obvious question. If the video is not driving the purchase, what is?
Quite a lot, and most of it is bigger than the video.
The most useful number here comes from a meta-analysis of 872 advertising elasticity estimates drawn from 56 studies. Sethuraman, Tellis and Briesch found a mean short-term advertising elasticity of 0.12, against a price elasticity of −2.62. Price moves purchase behavior more than twenty times as hard as advertising does. That paper is from 2011 and it is an empirical generalization rather than a news statistic, which is precisely why it is still the reference point.
Current shopper behavior points the same way. Zappi surveyed 2,000 US consumers in March 2026 and found 70% cite price or value as the top influence when shopping snacks and beverages, with brand-name-only purchasing falling from 21% to 10% year over year. Ibotta's survey of more than 5,000 grocery shoppers found 62% of food shoppers require a discount of 25% or more to switch brands.
So between the click and the register sit the offer, the retailer, the price on the shelf, whether the product is in stock in that shopper's store, and how much work the rebate takes to redeem. Every one of those is a real variable, and none of them belongs to the creative.
Holding a creator to receipt rate asks them for something the data says they cannot deliver.
The measurement gap here is well documented, and it runs in a specific direction.
Modern Retail surveyed 125 marketers in July 2026 and asked what they count as creator marketing success. Impressions came first at 28%, conversions at 27%, watch time at 17%, click-through rate at 14%, and commerce or sales last at 13%. The two ends of the funnel are getting the attention and the middle is thin.
Linqia's survey of more than 200 enterprise marketers found 79% struggle to measure ROI, with 48% naming attribution as the biggest gap. And the IAB's State of Data 2026, based on 400-plus senior planning and analytics decision-makers, found that 48% of marketing mix model users say creator and influencer marketing is underrepresented in their models, alongside 75% saying their advanced measurement approaches underperform on rigor and timeliness.
That last figure is worth sitting with. The models brands use to decide whether creator video worked structurally under-capture it, which means the channel gets judged on a metric it is not fairly represented in.
Crystal Duncan at Tinuiti put the practical version of this well in the Modern Retail research: measuring influencer campaigns outside of last-click attribution remains a challenge, and while whitelisted content in paid media makes creative performance easier to read, that only tells part of the story.
None of this is an argument to stop caring about the opening. A video nobody stops for cannot sell anything, and hook rate is the right measure of whether a hook worked as a hook.
The limit is what hook rate can stand in for. On its own it does not predict leads or sales, so a video should not be called a winner because it stopped people.
There is a real gap here too, and it is worth naming. Of everything we coded, only opening framing moved hook rate, by about three points. Opening length does not move it at all, sitting at −0.036 with p=0.76, even though opening length moves hold rate and cost per lead further down the chain. So beyond the framing rule, this dataset cannot yet name the structural choices that lift hook rate. That is a gap to close in the next batch rather than one to paper over.
The chain cannot be climbed like a ladder. Improving hook rate does not walk an effect down to the sale, because the rungs between are missing.
What works instead is straightforward:
Keeping those two apart is what makes a brief fair to the creator and useful to the brand. A creator who writes a fifteen-word opening in their own voice has done their job when leads come in cheap. Whether those leads convert to receipts is a question about your shelf price and your rebate, and it should be answered by the people who control those.
The full analysis, including the correlation table for all four links, the within-brand versus brand-level breakdown and the scope conditions on where these findings apply, is available here.
.svg){kind=icon}
.svg){kind=icon}
.svg){kind=icon}
.svg){kind=icon}
.svg)
.svg)
.svg)
