YouTube Shorts Demand Gen Built on Downstream Conversions

The threshold that decides this: Demand Gen is where Shorts inventory is bought now, and the campaign is only as good as the event it optimises toward. Our floor is fifty conversions per campaign per week. Below that the system never leaves learning and you are paying to train it on noise. When a client sits under the line we consolidate campaigns, or we optimise toward a proxy event we can show correlates with the outcome you care about, and we name the correlation and the sample it came from rather than asserting it in a slide. Optimising toward money, not forms: For hiring accounts we import ATS stages back into Google Ads against the click identifier, so the campaign chases interview attended and started work. For commerce we send margin-weighted values rather than order counts, which usually changes which products the system decides it likes. Enhanced conversions for leads, with hashed email and phone, covers the gap where the click identifier is lost in the redirect chain. This plumbing takes two to three weeks and it is the reason the account behaves like a different account in month two. Creative, and how we retire it: A Shorts ad is decided in the first second and a half. Letterboxed sixteen-by-nine cutdowns lose to phone-shot vertical every time we have run them against each other, so we do not run them at all. Text is legible with sound off and written for sound on. Each cut gets a fixed budget and a fixed window. If it has not cleared the account's median cost per outcome by the end of that window, it comes down without a meeting. Sunk production cost is not an argument we accept, and we say that at the start so it is not a fight later. What we report: Cost per downstream outcome, incremental lift where the spend supports a holdout, and the creative log with every decision dated. Views, view-through rate and average watch time are diagnostics for the editor, not results for your board. If the reporting line above you requires impressions and VTR as headline numbers, this engagement will irritate both of us and you should say so now. Brand lift: Google's lift studies need a spend floor before they return anything readable. Under roughly forty thousand euro in a single market across four weeks we will not sell one, because the confidence interval comes back wide enough to support whichever conclusion you walked in holding. Above the floor they are worth running and we will design the cell structure. Not included: No organic channel management, no thumbnails, no long-form production, no creator integrations; that is a separate negotiation and we deliberately keep it separate. We do not touch Search or Performance Max inside the same account without a written boundary on budget and audience overlap agreed first, because otherwise nobody can tell which campaign bought the conversion and the argument is unresolvable. Who should not buy this: Advertisers whose target event fires under about twenty times a week in a market and cannot be consolidated with anything else. App advertisers measuring on install alone. And anyone who wants Shorts run as an awareness line with no downstream event defined at all. There are agencies who will take that brief and bill it cheerfully for a year. How it runs: Weeks one to three are measurement plumbing and the first creative batch. From week four the account runs on a fixed test cadence, with one written note a week naming what was killed and why.