Hunch solves a problem that is genuinely hard by hand and genuinely boring to describe: turning structured data into thousands of ad variations that are each correct.
The two use cases it is built for
A catalogue advertiser with 8,000 SKUs cannot produce meaningful creative per product manually. Dynamic product ads solve the mechanics and usually look like what they are, which is a feed rendered into a template. Hunch’s enhanced DPAs apply background removal and colour matching so the whole catalogue looks deliberate rather than scraped, and catalog product video extends that into animation.
The second is multi-location, and the numbers make the case on their own: more than 2,000 localized ads in under 30 minutes. A national brand with hundreds of branches cannot hand-build creative per location, so most run generic national ads and accept the loss. Automating it changes what is possible rather than what is faster.
Smart Product Sets is the quieter good idea. Grouping products by GA4 and conversion behaviour rather than by catalogue taxonomy means the sets reflect what actually sells together, which is rarely how a merchandising team structured the categories.
The phrase worth noticing
Among the published results, one reads differently from the others: Academy Sports at 2.3x incremental ROAS.
Incremental implies a holdout. Attributed ROAS, which is what almost every vendor across this entire directory quotes, counts revenue that touched the campaign including purchases that would have occurred regardless. Incremental compares against a control group that saw nothing, and answers whether the spend caused anything.
Using that word is a signal of measurement discipline that most competitors do not display, and it is worth crediting. It also throws the other two figures into relief: 242 percent ROI lift at Carwow and 77 percent lower CPA at Kiwi.com are not described as incremental, and should be read as the softer kind of claim.
Ask which of the three were measured against a holdout. A vendor comfortable using the word once should be able to say.
Where the value depends on you
Everything here multiplies structured input. Without a product feed or location data there is nothing to multiply, and a general creative tool will suit you better.
That dependency cuts both ways. Feed-driven automation propagates feed errors exactly as efficiently as it propagates everything else, so a wrong price or a missing image becomes wrong across two thousand ads in half an hour. Audit the feed before scaling the automation.
The other cost is convergence. Template-driven output across many advertisers using the same platform starts to look similar, and at catalogue scale nobody is reviewing every asset. Review output as a set periodically, and keep your hero creative outside the machine.
No pricing is published, which sits awkwardly next to the site’s emphasis on fair pricing and no hidden fees. Establish whether cost scales with spend, feed size or campaign volume before comparing it to anything.
