Your 12x ROAS might actually be 4x

📊 Kishalaya’s 1PD Ops Summit session is up, and it is the most uncomfortable one of the day in the best way.

The headline: a channel reporting 12x last-click ROAS can be producing 4x incremental ROAS. Same spend, same dashboard, completely different decision about next month’s budget.

He opened with something most of us know but work around anyway. A lot of what Google and Meta report as tracked data is predicted data, which is why your platforms added together often claim more conversions than your actual revenue.

Three parts I would point you to:

Segmentation as a testing foundation rather than a targeting trick, from 09:43. Loyal, lapsed, high LTV and first-time buyers, built on first-party data you own. His point on LTV is worth sitting with: marketplaces outbid you on your own category not because their creative is better, but because their customer lifetime value lets them afford a higher bid.

The holdout result, from 23:31. An Australian D2C apparel brand spending around 1 million AUD a month. Broad targeting produced 2.2K incremental units over a geographic holdout baseline. Restructuring the same budget by product category, gender intent and RFM recency produced a further 2.9K units on top. Total spend did not move.

Branded search, from 45:44. An international brand spending 2 million dollars a month on branded search cut it by 80 percent after holdout testing, with negligible loss in actual conversions. High-intent users searching the brand name were clicking the organic result anyway.

The statistics section at 32:51 is the part to not skip if you plan to run this yourself. Adequate sample size, 80 percent power, 95 percent confidence. An underpowered test does not return a cautious answer, it returns a wrong one.

Watch the full session: https://youtu.be/S9FO6XYeS_E

Two questions for the group: has anyone here run a branded search holdout, and what did it show? And for those doing geo holdouts, are you using Geolift, CausalImpact, or picking markets manually?

#IncrementalROAS #HoldoutTesting #PerformanceMarketing #FirstPartyData #1PDOps