Your overseas marketing team is weighing two podcast sponsorships for a product made with a Chinese supply partner. Both proposals describe their downloads as “deduplicated over 24 hours.” One offers a better estimated CPM. That phrase alone does not establish that the denominators are comparable.
Before using downloads to rank the proposals, ask when the clock starts and what resets it. ONCE’s recommendation is to attach a short measurement buying card to the scope of work, then compare prices. Audience relevance and creative fit still need their own assessment.
A specific clarification buyers can use
IAB Tech Lab’s final Podcast Technical Measurement Guidelines v2.3 page was updated September 29, 2026. Its scope includes audio and video podcasts delivered through open RSS or applications relying on server-side file delivery; it is not a universal definition of video-platform views.
Section 5.5.1 of the technical text distinguishes a daily reset, a window anchored to the first qualifying request, and a window extended by later requests. It describes these alternatives rather than selecting one for every provider.
Provider disclosures illustrate the distinction. RSS.com’s methodology specifies a fixed window beginning at 00:00 UTC. Flightcast’s public methodology specifies last-touch rolling deduplication, with counted downloads subsequently placed in UTC daily buckets. Flightcast also says it is not formally IAB-certified. These are statements about the providers’ published methods, not an ONCE audit or evidence that the market has adopted one implementation.
Four requests can produce three different totals
This is an original hypothetical timing exercise, not campaign data. Assume four requests for one episode, using the same identification combination, pass all other necessary checks. Every time below is UTC.
| Request | Daily reset at midnight | First-touch 24 hours | Last-touch extension |
|---|---|---|---|
| Monday 23:00 | +1 | +1 | +1 |
| Tuesday 01:00 | +1 | No addition | No addition; extend |
| Tuesday 23:30 | No addition | +1 | No addition; extend |
| Wednesday 00:30 | +1 | No addition | No addition; extend |
| Total | 3 | 2 | 1 |
Crossing midnight matters to the first column. Passing the original Monday window matters to the second. In the third, every adjacent request arrives within 24 hours, extending the window again.
Do not turn 3:2:1 into a correction factor. Those totals belong only to this constructed sequence. A real show’s difference depends on its request timing. An aggregate download figure cannot tell you how to reconstruct another method’s total.
ONCE’s measurement buying card
The following is our proposed purchasing method, not an IAB contract requirement. Ask the seller to complete it before you approve a download-based estimate.
| Field | Required answer | Buying consequence |
|---|---|---|
| Unit purchased | Downloads, ads delivered or a fixed episode sponsorship; included shows and markets | Keep different units out of the same CPM ranking |
| Deduplication clock | Daily, first-touch or last-touch; reset timezone if fixed; counted object | “24 hours” alone leaves the method unresolved |
| Observation period | Matching episode age and date range; daily display versus daily deduplication | Align the period before comparing |
| Settlement version | Accepted report, freeze date and treatment of methodology changes | Agree these before negotiating delivery guarantees |
The distinction between display and deduplication is practical. Flightcast describes rolling deduplication even though its report groups the resulting counts by UTC date. A daily x-axis does not establish a midnight reset.
Recalculate a sample only when the underlying records support it
When two tools disagree about the same show, ask the data holder to process an authorized sample with matched filters, dates and window logic. Request a reproducible aggregate explanation; your brand does not need raw IP addresses or identifiable audience records for this exercise.
If only totals are available, keep the original provider labels and state that normalization was not possible. Do not manufacture a multiplier. Across different shows, aligned clocks still leave differences in audience fit, geography, placement and integration quality.
A pricing model with an explicit denominator
Let P be the sponsorship price and D the qualified downloads under the agreed method. The planning download CPM is 1,000 × P ÷ D. Change the method behind D while holding P constant and the apparent CPM changes, without proving that production or advertising became more effective.
For a fixed sponsorship, itemize the media commitment, production or adaptation work, and usage rights. Treat downloads as a reporting observation unless the agreement actually makes them the payment unit. An estimated CPM should not silently become a guaranteed delivery commitment.
What to change in the next brief
For the brand lead, add four questions to the request for proposal: window type, timezone, observation period and settlement source. An unresolved method can justify a limited, clearly scoped fixed-fee pilot; it does not justify scaling spend because one headline CPM looks cheaper.
For the production partner, store the measurement-method version separately from the episode’s creative version. If the method changes, mark the break in the reporting series before recommending a new opening, another shoot or a renewal. This is particularly useful when a China-based production team hands ongoing reporting to an overseas marketing team.
For ONCE’s service design, an annual podcast engagement should preserve both content assets and the assumptions needed to compare reports over time. Whether this improves margins or renewals is a future project question, not an outcome claimed here. Explore related work in our video marketing articles when planning the broader content program.
Sources and limits
Checked October 31, 2026: IAB Tech Lab, Podcast Technical Measurement Guidelines v2.3, final-page update September 29, 2026, and section 5.5.1; RSS.com, Description of Methodology (DOM), whose displayed modification date is June 27 without a year; Flightcast, Podcast Measurement Methodology, no publication date displayed.
The definitions and provider disclosures are sourced facts. The timing exercise, buying card and purchasing recommendations are ONCE analysis. We did not inspect live campaign logs or estimate a typical market-wide discrepancy.
