Spotify Ad Manager Campaign Build With Geo Holdout

Spotify is bought here as a frequency instrument rather than a reach instrument. A campaign goes live with one audience definition, three to five script variants in rotation, a cap of three impressions per listener per week, and one metro deliberately excluded from targeting so the result has something to be read against. Everything below follows from that first choice. What gets built: Campaign architecture in Spotify Ad Manager with a separate line item per creative variant, so variant is a readable dimension in the report rather than a blend nobody can decompose afterwards. Audience assembled from Spotify's own signals, including age, platform and listening context such as workout, commute and focus, with podcast placements added where Streaming Ad Insertion carries the format. Companion banner and clickable overlay configured on every placement that supports them. A click on Spotify comes from a banner tapped by somebody whose phone is in their pocket, and the rate lands under half a percent on nearly every campaign we have run. It is a delivery check. It does not go in a results deck. The failure mode nobody plans for: A Spotify audio ad on the ad-supported tier cannot be skipped. That reads like an advantage and is not. The failure is not a skip, it is the app going to the background and the phone going face down, and no platform report will ever show you it happened. It is also why South African ad-supported listening, much of which runs on prepaid mobile data, behaves differently from the same demographic in London: data has a visible price per megabyte here and the listener knows it. So the offer sits inside the first four seconds and the ad earns the remaining twenty-six. How the result gets argued: The Spotify Pixel goes on the site and it will under-report, because the listen and the visit usually happen in different sessions and often on different devices. The number we defend instead is the holdout: a comparable excluded metro, read week over week against branded search volume in Search Console, direct sessions, and a spoken URL a person can still recall after one hearing. Above roughly forty thousand dollars of committed spend we design that as a matched-market test with a pre-period and a stated minimum detectable effect. Below it we put in writing that the read is directional, and we do not dress it up. Not included: Script writing and voice recording. That is separate work with its own booth time and talent rights, priced separately, and we would rather you bought it from us than had a media plan carry a weak read. Also excluded: Spotify video takeover as a standalone campaign. We will place video inside an audio buy where the plan calls for it, but a Spotify campaign made only of video is a display buy wearing headphones. And we do not buy flat-fee playlist sponsorships that carry no impression floor. Who should not buy this: Anyone who needs this channel to return a last-click ROAS figure inside thirty days. It will not, and any agency saying otherwise is reporting people who were going to convert anyway. Anyone whose total audio budget is under roughly five thousand dollars a month, because the frequency needed to move branded search cannot be bought out of that, and a thin Spotify campaign is worse than none: it teaches a company the channel does not work when what failed was the budget. And brands whose domain cannot be spelled correctly by somebody who has only heard it.