Dynamic Insertion Buying on Megaphone, Acast and Spotify

Dynamic insertion is where the inventory is and where the credibility is not. An announcer-read spot inserted dynamically on Turkish supply clears roughly 8 to 14 euros CPM against 25 to 45 for a host-read on a comparable show, and against a control group it converts at a fraction of the host-read rate on every campaign we have measured. We buy it deliberately, around a host-read core, for reach and for frequency. Never as the campaign itself. Where we buy and how: Megaphone's targeted marketplace and targeted drops, Acast direct and programmatic, the Spotify Audience Network and Spotify's ad exchange, AdsWizz and Triton where supply is radio-adjacent. Audio impressions are confirmed server-side when the ad bytes are delivered, which is why an impression count in a platform report is not a count of people who heard anything. We reconcile each platform's delivery against Podscribe reach before a number enters a client report, and we have sent invoices back over the gap. Frequency, and the thing nobody in audio can do: There is no cross-platform frequency cap in podcast advertising. Anyone selling you one is capping inside their own supply and calling it a campaign cap. We cap at three exposures per week per platform, state plainly that overlap exists, and measure true combined frequency from the attribution side. When average four-week frequency passes about eight we cut spend rather than rotate in new creative. Rotating creative is what gets recommended because it bills better, and it does not fix the problem. Creative specification, because most delivery failures are file failures: Fifteen, thirty and sixty second cuts, delivered as 48 kHz 24-bit WAV and trafficked as 192 kbps MP3. Integrated loudness at -16 LUFS stereo and -19 LUFS mono, true peak no higher than -1 dBTP, so your spot does not arrive noticeably louder than the show and get skipped on reflex. Turkish and English cuts voiced by separate native speakers rather than one voice attempting both. Code said twice, URL once, no hyphens in either. Brand safety, and why the standard blocklist breaks foundation campaigns: Transcript keyword blocking is sold as protection and works as a filter against exactly the content a foundation is trying to reach. A campaign on refugee resettlement blocked from every show that discusses migration has bought nothing. We build exclusions show by show, having listened, and take the extra work rather than hand a keyword list to a platform and call it governance. What we will not buy: Music streaming inventory bundled into a podcast line. Audience extension where the show list is not disclosed before signature, because if a seller cannot name the shows we cannot tell you where the money went and eventually you will ask. Companion banners priced as though anyone sees them. And any platform refusing a third-party pixel, unless you accept up front that the line is measured on code redemption alone and we label it that way in every report. Not for: Buyers who want scale first. Dynamic insertion multiplies whatever your host-read core is already producing, including nothing. If you have not run a host-read flight against a control group, this is the wrong engagement and we will say so on the first call rather than take the budget. It is also wrong for anyone whose reporting cannot tolerate two different numbers, since platform confirmed impressions and measured reach will never reconcile exactly, and that gap is a fact about audio rather than a fault to escalate.