Podcast Attribution Rebuilt Around an Actual Holdout

Chartable stopped serving third-party attribution in 2024 after Spotify bought it, and Podsights became Spotify Ad Analytics, which puts your measurement inside a company you also buy inventory from. Rebuilding on independent instruments takes about three weeks. In the first month after we do it, reported podcast revenue usually falls between 30 and 60 percent. That fall is the deliverable. The earlier number was counting people who were going to buy anyway. What we install: A Podscribe pixel across the funnel rather than only the confirmation page, so we can separate a listener who lands from a listener who buys. An OP3 or Podtrac prefix on anything you publish yourself. Vanity URLs on your own domain with server-side redirects readable in your logs, never a shortener we cannot audit. One promo code per show per flight, never a single code shared across three shows because somebody wanted a tidy landing page. And a post-purchase survey field written into the order record, because self-report is the only instrument that sees the listener who heard the read while driving and searched your brand two days later. The holdout: We do not sign an attribution report that has no control group. On dynamically inserted inventory we run a PSA holdout: same slot, same shows, a public-service creative to a 10 to 15 percent slice, then compare converted households across the two groups. On baked-in host reads there is nothing to hold out inside the show, so we run a matched-market design instead, paired provinces in Türkiye or paired countries elsewhere, four weeks of pre-period, spend held flat everywhere else. Both designs are agreed with you before a euro moves, because a holdout designed after the results are in is not a holdout. KVKK, and the Turkish coverage problem: Household IP matching is how Podscribe and Spotify Ad Analytics work, and matching an IP address to a customer record is personal data processing under Law 6698. We write the line that belongs in your aydinlatma metni and the retention rule beside it, and we put it in front of your counsel rather than assuming. Separately, we tell you where the instrument does not reach. Several Turkish networks host on platforms that will not accept a third-party impression pixel at all. On those shows the honest measurement is a dedicated code plus the survey field, and they go in their own column rather than being blended into one attributed-revenue figure that implies precision nobody has. What this does not include: No media mix model. Below roughly two million euros of annual audio spend an MMM tells you less than a clean holdout and costs more; above it we say so and name who should build it. No BI development. We deliver into your warehouse or your existing reporting and we do not become your data team. No creative, no buying, no rate negotiation inside this engagement, because mixing them makes us the marker of our own homework. Who should not hire us for this: Anyone who needs the podcast number to go up. It goes down first, and if the internal politics cannot survive that, the engagement fails in week five instead of week one. Teams who cannot add a survey field or change a checkout inside the engagement window, because the build stalls and you pay for our waiting. And anyone whose reporting is locked to last-click in a system they do not control, since we will produce a second number that contradicts the first one every month and somebody has to be allowed to act on it.