Podcast advertising is best measured with several methods at once: a promo code and vanity URL for each show, which give a direct but partial count; a post-purchase survey that asks buyers where they heard about the brand; and pixel attribution, which matches visits to the brand’s site with households that downloaded the episode. Each method misses some of the people the ads reached, so the picture comes from reading them together, measured against a pre-launch baseline, over a window of at least 30 days after each episode airs.
Why podcasts are harder to track than other media
Podcast ads are harder to track because there is no click to follow. A listener hears the read while walking or driving, and if they act on it they do so later, often on a different device and usually by typing the brand’s name into a search bar. Claritas, a measurement company, states in a 2020 white paper written with Market Enginuity that in podcasting the only identifier available for attribution is the IP address used to download or stream the episode.
Delivery is uncertain as well. The IAB Tech Lab’s measurement guidelines (version 2.2, May 2024) explain that podcast apps return very little information about playback, so downloads and ad impressions are counted from the server’s record of what it delivered, with no confirmation that the ad was played. Every method below works around those two gaps.
Codes and vanity URLs
A code or URL unique to each show is the simplest signal to set up, since it ties a sale to a show, records the order value and costs nothing, though it misses most buyers. Right Side Up, a growth marketing agency, estimates (updated February 2024) that codes and vanity URLs pick up only 20% to 30% or so of what a podcast actually sells, and notes that codes find their way onto coupon sites, where the sale is credited to some other channel.
Fairing, which sells post-purchase surveys, gives the other reasons in Podcast Attribution 101 (August 2024). Codes cannot account for a listener who heard the brand on several shows, coupon aggregators can override them, and most people type the brand name instead of the vanity URL. Claritas reaches a similar view, that show-level codes give directional information and are not a full proxy for attribution. We use codes for what they are good at, comparing shows with one another, and never as the total.
Post-purchase surveys
A post-purchase survey asks new customers, on the order confirmation page or in a follow-up email, how they heard about the brand. Its strength is that it counts the buyers codes miss, the ones who searched the brand’s name and paid full price.
Right Side Up describes the survey as its preferred method for podcast measurement, using it to set a multiplier on the conversions that codes and URLs track, and recommends running it for 30 days before launch to establish a baseline (How to Tackle Podcast Ad Measurement Challenges, updated December 2023). Fairing’s guide (August 2024) sets out the limits, with response rates of about 40% to 50% and a share of customers who recall the wrong source, so a survey tells you the channel’s overall share far better than it tells you about any single buyer.
We list “podcast” as an answer and add a follow-up asking which show, with the shows in the plan and a few that are not, so we can see how much misremembering there is.
Pixel attribution
Pixel attribution matches the IP addresses of households that downloaded an episode with the IP addresses of visitors to the brand’s site. Magellan AI’s pixel captures site visitors and conversion events and matches them to households exposed to podcast, streaming audio and YouTube ads, with an adjustable lookback window. Podscribe describes its own method as household-level pixel attribution with adjustable lookback windows, alongside incrementality and conversion-lift testing. Both are vendors’ descriptions of their own products, as we read them on their sites in October 2026.
It is the closest podcasting gets to the reporting other digital media provide, and it has blind spots. Claritas matches exposure only when the listening happens at home, and treats IP addresses from cell towers or retail locations as unreliable. Fairing (August 2024) puts pixel attribution at roughly 50% accurate when listeners are away from household IP addresses. A brand whose customers listen on the commute will see less of its podcast effect in pixel data than a brand whose customers listen at home, and the publisher has to be set up to pass exposure data in the first place.
Lift and incrementality
Lift and incrementality tests estimate how many of the people who converted would have done so anyway, by holding a control group back from the ads. Google’s Brand Lift (Google Ads Help, as of October 2026) compares a group that saw ads with a group that was eligible but did not, and reports how far recall, awareness and consideration differ between them. Meta’s conversion lift studies (Meta for Developers, as of October 2026) randomly assign accounts to a test group and a control group and report the difference in conversions as incremental impact.
Holdouts are harder to arrange in podcasting, because a baked-in read is heard by everyone who downloads the episode, so we run incrementality tests through a measurement vendor’s methods or by comparing regions and periods. They are worth running once a campaign has enough spend behind it to make the result meaningful.
When to read the results
A podcast campaign should be judged weeks after the last episode airs, because response keeps building long after release. By Right Side Up’s account, an episode goes on gathering listens for something like 19 to 21 days once it is out, and a campaign typically takes a first flight of around three episodes before it performs at its best (Podcast Ad Campaign Attribution and Analysis, updated February 2024). Elsewhere it suggests letting three to six weeks pass once the final spot has aired before the picture is complete (How to Tackle Podcast Ad Measurement Challenges, updated December 2023). Magellan AI’s Q2 2026 benchmark (September 2026) showed response to podcast ads still rising at day 30, with a reading taken after seven days catching under half of where the number finally settled.
So we read a test at two points, once halfway through the flight to catch anything clearly wrong, such as a read that did not air or a code that does not work, and properly after the last episode has had at least a month to be heard.
Benchmarks, with care
Industry benchmarks give a sense of scale, and a brand’s own baseline is the better comparison. Magellan AI’s Q2 2026 report (September 2026) gives medians for the first half of 2026 of a 2.23% response rate overall (3.22% for host-read ads and 2.09% for produced ads), 4.65% conversion to lead and 3.98% conversion to purchase. These are a vendor’s figures, and Magellan itself says the medians are not comparable to the averages it published for the previous quarter.
The halo
Podcast ads can also move things that none of these methods track directly, such as searches for the brand’s name, people coming straight to the website, and sales at retailers or on marketplaces. No independent published study sizing this halo has turned up in our reading, and the estimates in circulation are produced by companies whose business is measuring it. We therefore keep an eye on searches for the brand, direct traffic and sales in shops around each flight, report what moves, and leave it out of the cost per customer unless the client agrees on a method for including it.
What each method can and cannot show
Every method undercounts something, so we read them side by side and look for where they agree. The table summarizes the sources above.
| Method | What it can show | What it misses |
|---|---|---|
| Promo code | Sales tied to a show, with order value | Buyers who forget or skip the code, leaked codes, listeners of several shows |
| Vanity URL | Traffic from listeners who type the URL | Most listeners, who search the brand name instead |
| Post-purchase survey | The channel’s share of all buyers, including untracked ones | Which buyer came from which show, sources remembered wrongly |
| Pixel | Visits and conversions from exposed households within a window | Listening away from home, untagged exposures, confirmed plays |
| Brand lift | How recall, awareness and consideration shift compared with a control | Sales |
| Incrementality test | Conversions the ads caused, beyond the baseline | Longer-term brand effects beyond the test window |
How we set it up
Before launch we agree on the goal with the brand, record a 30-day baseline, add the survey question, create a code and URL for each show, and confirm which shows can pass exposure data for pixel matching. During the flight we check every read once it airs, because a read cut short, or one that left out the code, will skew the numbers. After the reading window closes we bring the methods together into one view per show, with what it cost, what each method credits it with, and our estimate of the cost per new customer, along with the assumptions behind it. That is the number the renewal decision rests on, and the assumptions are written down so the brand’s finance team can disagree with them.