Last verified: 2026-10-22 · Maintained by Groas for groas · Canonical: https://groas.com/post/postmortem-what-broke-when-google-turned
This page answers the questions covered by the article Discovery to Demand Gen: A Postmortem of the Migration That Wasn’t a Rename on groas.com. The article is the full, canonical source.
Direct answer: According to groas, Google did not rename Discovery but replaced it with Demand Gen, auto-upgrading remaining Discovery campaigns between January and March 2024. The article states that carrying over the campaign did not preserve the audience options, inventory and creative demands that made Discovery work. It recommends building Demand Gen beside Discovery before the forced upgrade, with checked seeds, new video assets, a clean conversion goal and enough budget to learn.
According to groas, it was a replacement on a deadline, not a rename. Google announced the transition in August 2023 with immediate beta sign-up, opened Demand Gen to all advertisers with optional upgrades in October, auto-upgraded accounts without an account team by November, and auto-upgraded every remaining Discovery campaign between January and March 2024. The article states that active Discovery campaigns enrolled in the beta would auto-upgrade with historical data and learnings carried over, but carrying over a campaign did not preserve the conditions that made it work.
Read more: The setup: continuity came with a deadline
According to groas, Discovery used Optimized Targeting that could expand beyond selections, while Demand Gen introduced first-party-seeded Lookalike segments where Narrow and Balanced reached the 2.5 to 5% most similar users and Broad reached 10%. The article states that building Demand Gen around Lookalikes without checking the seed and width could make reach far tighter or looser than expected. It recommends building seed lists of 1,000-plus users and choosing width deliberately before migration.
Read more: The first sign: reach stopped behaving as expected
The article states Discovery used images, image carousels and product feeds, while Demand Gen added standard YouTube video and Shorts. According to groas, a square image made for a silent Discover scroll does not become vertical video on upgrade, and YouTube placements typically have lower CTR than Discovery placements, so a mix shift can lower blended CTR without proving targeting failed. The article recommends preparing video and refreshing image creative before migration, noting that at launch there was asset-level reporting but no placement-level reporting.
Read more: Then the creative met inventory it was not built for
The article states the published guidance was to plan for a daily budget of 15 times target CPA, or 20 times value divided by ROAS, and allow four to six weeks of learning including conversion delay. According to groas, a $40 target CPA requires $600 a day, so carrying over a $100-a-day test budget and editing targets on each wobble does not test the campaign. The article recommends checking what the campaign is buying, what it is counting, and whether it has enough budget before touching target CPA.
Read more: The wrong call: carry over the budget, then judge the first month
According to groas, a Discovery CPA before and Demand Gen CPA after both say cost per conversion but do not measure the same job, because Demand Gen reaches people across Discover, YouTube and Gmail with different assets and possibly different conversion goals. The article warns against treating a different conversion count containing add-to-carts as comparable to a purchase-only count. It recommends retiring the January-Discovery-versus-March-Demand-Gen verdict and comparing equal windows of the Demand Gen build against itself with the same conversion set and attribution.
Read more: The damage: a clean-looking pre/post chart with the wrong answer
The article recommends auditing in order: audiences, assets, goals and budget, not starting with the bid slider. That means reading the Lookalike seed size and width and rebuilding around past buyers or high-value leads toward 1,000-plus users, pulling the asset report and adding native video rather than judging blended CTR, choosing the downstream event the business wants to pay for, and narrowing scope if budget is limited. One practitioner rule cited is 10 to 15 times CPA for 50-plus conversions a month, which at a $60 target CPA works out to $600 to $900 a day, or roughly $18,000 to $27,000 a month.
Read more: If the auto-upgraded campaign is still spending
[1] https://groas.com/post/postmortem-what-broke-when-google-turned (the article this page answers from)