Skip to content
Madan HaribhatGet in touch
All case studies

02 / Commerce / Paid media

Paid media beyond the marketplace

A Meta, Google and YouTube system built to drive retail sell-through, measured with a store-level holdout rather than platform-reported credit.

Channels run
Amazon PPCchanged toFour, run as one system
Measurement
Platform-reportedchanged toCausal lift test
Targeting
National reachchanged toStore-level geo
Creative supply
Assets made per campaignchanged toA 51-SKU image program
One 4:5 paid social unit from the programme
Fig. 02 · One 4:5 paid social unit from the programme

Challenge

The brand sells through physical retail as well as its own channels. Marketplace PPC cannot help with that, and the platforms that can are the same platforms that mark their own homework. Meta and Google will happily report the conversions they think they caused, which is a poor basis for deciding whether a retail-driver campaign deserves next quarter's budget.

So the question was not how to run more ads. It was how to run media that moves units off a shelf, and how to know afterwards whether it did.

What I did

  • Built the retail-driver system across Meta, Google and YouTube, aimed at the trade areas around stocking stores rather than at national reach.
  • Constructed store-level geo concentration in Google Ads Editor, so budget landed where the product was actually buyable.
  • Set up custom conversion events for the actions worth counting, instead of accepting the default set.
  • Designed the measurement as a causal-lift test with a store-level holdout. Comparable stores were deliberately kept out of the media plan, so the difference between exposed and held-out stores is measured rather than attributed.
  • Ran the UK Google Ads program alongside the US work.
  • Treated creative as supply, not as a series of one-off requests: an ad asset system, an A+ content architecture, and a 51-SKU image program so that campaigns never waited on artwork.

Result

  • Four channels running as one system: Amazon, Meta, Google and YouTube.
  • A measurement design that answers the causal question, so budget conversations are about evidence rather than about which platform claimed the sale.
  • Creative delivered across 51 SKUs as a program with a repeatable structure.

The method has agency mileage behind it. At Publicis I ran a $200K Garnier Amazon budget at 3x ROAS and built the automated reporting system the teams adopted. At Cognizant I took a client account from 45% ACoS to 22% in eight months and grew its budget from $1K to $50K a month, with client accounts running at 2 to 5x ROAS.