Demand Gen Tested Against a Geo Holdout, Not Last Click

Demand Gen judged on last-click attribution loses to your own brand search campaign every time. So the measurement design is the first deliverable and the campaign is the second. The smallest honest read we will accept is a geo holdout across matched German regions, run for at least four weeks, comparing total revenue rather than campaign-attributed revenue. How the read is built: We split Germany into matched test and control regions on baseline revenue and category seasonality, hold Demand Gen out of the control set entirely, and read total revenue plus branded search volume. Brand query volume usually moves seven to ten days before revenue does, which makes it a useful early signal that something is happening rather than a proof that it worked. Where a client's volume is too thin for a clean geo read, we say so and run the campaign as a declared awareness test with a fixed budget and a stop date, rather than pretending the ROAS column carries meaning it does not. Creative in concepts, not in sizes: The lever in Demand Gen is the number of genuinely different arguments put in front of the same audience, not the number of crops of one image. Per asset group we want distinct concepts: a product demonstration, a customer proof, a category-entry explainer for people who do not yet know this product exists, and a seasonal offer. Each is cut vertical for Shorts and in-feed, and thumbnails are treated as their own creative decision, because in Discover the thumbnail is the ad. For retail we attach the product feed so the campaign can carry actual SKUs rather than only mood. Audiences: Customer Match seed lists built from your own data, not Google's shelf-ready affinity segments. Google will build a lookalike from a thousand matched members; we want at least five thousand active ones before we trust the shape it produces. And we segment the seed, because a lookalike of everyone who ever bought is usually a lookalike of your discount traffic. Repeat purchasers, high-value first purchasers, and customers lapsed at the replenishment point are three different seeds producing three different audiences. Regulated categories: Supplements and wellness creative on YouTube sits under Google's healthcare and medicines policy and, in German, under the HWG. Before-and-after imagery, implied treatment claims, and testimonials describing an effect are the three things that get creative pulled after the production money has been spent. We review at script stage, not at upload. What we refuse: Copying a target CPA from a search campaign into Demand Gen; the two channels are doing different jobs and the number means nothing when moved. Reading results in week two. Running a single creative and treating the flat result as a verdict on the channel. And optimising toward a view-through conversion window chosen because it flatters the campaign. We set a view-through window you would be willing to defend in front of your CFO, and we report the click-through and view-through figures separately, always. Not included: Video production from scratch. We work with your existing footage, your production partner, or an editor we can recommend, and we direct the cut list, the hook and the first two seconds. Photography, brand identity, influencer sourcing and organic social management are all outside this. Who this is not for: Brands with no first-party customer data and no existing branded search volume. Cold Demand Gen for an unknown brand in a crowded German category is an awareness spend; it should be funded from an awareness budget and judged over quarters, and we will not sell it to you as performance. Also not for teams who cannot produce or approve new creative monthly. The campaign will fatigue, and the result will be a verdict on your creative pipeline rather than on the channel.