In-Feed Video Bought as Intent, Not as Reach

In-feed video is bought against a choice. Nothing is charged until somebody looks at a thumbnail and decides to open it, which makes the placement closer to a search ad than to a film buy and means it should be judged the same way: cost per session that reaches a page you care about, not cost per view. In most accounts we inherit, in-feed sits inside the same campaign as in-stream, pulling the blended cost per view down by roughly a third and concealing that it produced nothing. Separating the two is the first change we make, usually inside week one. Three surfaces, three intents: YouTube search, watch-next and the home feed are sold together and behave nothing alike. A watch-next placement inherits the intent of the video that just finished. A search placement inherits a typed query. The home feed inherits nothing at all and has to manufacture interest out of a thumbnail. We split them into separate campaigns with separate bids and separate creative, and we pay more on search placements without apology, because one person who typed how to compare energy tariffs is worth a great many who were browsing. The thumbnail is the advertisement: The film is the landing page. We design thumbnails as artwork rather than lifting a frame from the edit, and we test them in a fixed grid, five thumbnails against three headlines, with an impression floor before anything is retired so nothing dies on one bad afternoon. Every candidate is checked at rendered size on a phone, because text that reads on a monitor disappears at the size the feed actually uses. Faces beat product shots in education work by a wide margin and lose in energy, where the bill itself is the strongest image available. What we count: Earned views, earned subscribers and playlist adds from the platform reporting, then the part that decides the budget: sessions from the watch page onward with a defined engagement threshold in GA4, and the category action inside a seven-day window. A prospectus download. An open-day registration. A switch started. A survey booked. Chosen views are reported separately from paid views throughout, because blending them flatters the buy. Where in-feed does not work: Brands with no reason for a stranger to choose their video over the one beside it, which is most brands most of the time. Products dull enough that the thumbnail has to overpromise, where you win the click and lose it in the first fifteen seconds and the watch page says so plainly. Local businesses with a footprint smaller than a city, where the ranking system has too little behaviour to work with and delivery collapses to a handful of impressions a day. Scope boundaries: We do not manage your channel, publish organic content or moderate comments. We will write the response policy for comments on paid placements and hand it to your team to run. We do not shoot, edit or subtitle; we specify what the thumbnail and the opening fifteen seconds have to do, then review what comes back. We do not buy Meta or TikTok feed placements. Different medium, different practice, and we will refer you rather than pretend. How the engagement runs: Two weeks of separation and instrumentation before any new creative, because the existing account almost always already contains the answer to what in-feed is really costing you. Then a thumbnail and headline cycle every fourteen days for the length of the flight, with a written call on each retirement and the reason for it. You receive the raw placement and search-term exports, not a screenshot of a dashboard.