How to measure B2B podcast ROI

Akshay Datt

host of Founder Thesis. Over 400 founder interviews and more than 1,000 hours recorded, which is long enough to have watched the download number stop meaning what everyone assumed it meant.

To measure podcast ROI, stop treating downloads as the outcome and instrument four things you can observe: self-reported attribution captured on your own forms, branded search volume against a pre-launch baseline, guest relationships tracked into your CRM, and platform-side consumption figures tracked per platform over time. Podcast listening happens off your property, in apps that send no data back.

What is podcast ROI?

Podcast ROI is a set of instruments that connects a show’s cost to business outcomes it can be observed to produce, because the podcast download is a distribution metric that cannot show whether anyone listened.

The break-even calculation is the primary target. Take the show’s fully loaded annual cost, call it X. Take average contract value, call it Y. The show breaks even at X divided by Y closed deals per year. When the annual cost of the show is about the size of one contract, that is one deal a year.

Everything else is a secondary indicator, tracked per platform, with the measurement window written next to it.

Podcast ROI metrics that work

Seven instruments produce observable signal: platform-side consumption metrics, self-reported attribution, branded search lift, guest relationships into pipeline, sales team usage, content output derived from the show, and audience composition.

None of them require listener-level tracking, because that does not exist.

Podcast consumption metrics from Apple and Spotify

Apple and Spotify both publish consumption data inside their creator dashboards.

Apple Podcasts Connect: Plays are “the total number of times people pressed play on your episode”; Listeners are “the total number of people that listened to or watched your show”; Engaged Listeners are “total number of people who listened to or watched at least 20 minutes or 40% of an episode”; Average Consumption is “a percent average indicating how much of the episode users played per device.”

Spotify for Creators: Plays are “the number of times any episode of your show was watched or listened to for at least 30 seconds on Spotify during the selected time period”, and Completion rate is “the percentage of your audience that made it at least 95% of the way through an episode on Spotify during its first 7 days after publish.” It also splits new audience from returning audience.

For a B2B show, Apple’s Engaged Listeners and Spotify’s completion rate are far better health indicators than any download count. They describe people who sat through the content.

Never add these together. Apple’s threshold is 20 minutes or 40%. Spotify’s is 30 seconds. They measure different behaviours and neither is a cross-platform person identifier. Track each platform’s series against itself, quarter over quarter.

Podcast attribution: ask on your own forms

Self-reported attribution is the only podcast attribution that works, because there is no pixel to fire. So ask.

Put an open-text “How did you hear about us?” field on demo request and contact forms, and ask the same question verbatim on discovery calls with the answer written into a CRM field. Use open text, not a dropdown. A dropdown collapses “I heard your CEO on [show]” into “social media” and the signal disappears.

Branded search lift from a podcast

If the show is working, more people go looking for your company by name. Google Search Console’s Performance report gives you Clicks (“the number of times a user clicked your site from Google Search results”), Impressions, CTR and Average position, grouped by the queries dimension so you can isolate brand terms.

Before launch, record a baseline for your brand terms over a full quarter. After launch, compare matched periods of equal length, quarter against quarter, not week against week. The report defaults to the past three months, so make baseline capture a launch task.

Branded search moves for many reasons, including funding announcements, paid campaigns and press. Direction over several quarters is the signal. A single good month is not.

Podcast guest relationships into pipeline

Podcast guest relationships are a business asset independent of audience size, if the show interviews people who resemble your buyers or who sell to them. Instrument the guest list like any relationship-driven channel: a CRM field or campaign for podcast guests, every guest and every person on the guest’s side who touched the recording logged, and normal pipeline reporting running over the top.

Sales team usage of episodes

Sales team usage is the cheapest signal available and almost nobody counts it. If reps are sending episodes into live deals, the show has product value. Use a tracked link per episode reserved for sales outreach, or a standing question in the pipeline review: which episodes went out this month, to which accounts, at what stage. If the answer after two quarters is none, that is data.

Content output derived from the podcast

Content output is what an hour of recorded conversation with a credible person turns into. Count what came out of it: clips, articles, newsletter sections, sales one-pagers, webinar segments. If the show produced twelve articles and forty clips this quarter, that is output the business would otherwise have had to commission separately, and it belongs in the budget conversation.

Podcast audience composition, not audience size

Podcast audience composition matters more than how many people listen. Senior executives at organisations with 500 or more employees make up 4% of all adult monthly podcast listeners, and 83% of them listened in the past week versus 66% of other monthly listeners.

What is a podcast download?

A podcast download is a server log entry recording that enough of an audio file left your host to be theoretically playable. It is a distribution metric, not a consumption metric.

The formal definition in the IAB Tech Lab Podcast Measurement Technical Guidelines: “Download: a unique file request that was downloaded. This includes complete file downloads as well as partial downloads in accordance with the rules described in section 5, which outlines the filtering process for measurement.”

The question its own metric answers: “Downloads: How many times is my network/show/episode downloaded and potentially listened to, at least in part?” The standard governing the industry’s headline metric will only claim potentially, and only in part.

The mechanics behind that hedge:

• The one-minute rule. “To count as a valid download, the header information plus enough of the podcast content to play for 1 minute should have been downloaded.” The reasoning: “downloads below a certain size are unlikely to result in human consumption because too little of the file was received to listen to any content.” One minute of file transferred, not one minute heard.
• Deduplication by IP and user agent. “A combination of IP Address + UA is used to identify unique listeners and downloads. For example, if the same file is downloaded 10 times by 6 user agents behind one IP address within a 24 hour window, that would count as 6 listeners and 6 downloads.”
• Bot filtering. “In podcasting, downloads generated by bots or other systems that will not be heard are considered invalid traffic.”
• Its own listener-count caveat. Mobile listeners change IP addresses constantly, which the spec calls IP-hopping, and this “can result in double counting listeners, but IP addresses can also be recycled, resulting in undercounting listeners. These factors can be difficult to account for.”

The structural reason none of this can see actual listening: “Podcast aggregator apps, like Apple Podcasts and Spotify, typically do not send playback data to publishers.” iOS and Apple Podcasts account for 59% of podcast players, which “prevents any client-side tracking or even the ability to count a ‘play.’”

What is a normal number of podcast downloads?

The median episode gets 27 downloads in its first 7 days. The top 10% gets 407 over the same window.

Two rules govern every download figure: always attach the measurement window, and never compare two figures measured over different windows.

Measured over the first 7 days after release:

Over the first 30 days, more than 130 downloads beats half of all shows, more than 1,100 beats 80%, more than 3,300 beats 90%, more than 7,700 beats 95%, and more than 35,000 puts a show in the top 1%.

Why “average downloads per episode” is a useless number. Over 7 days, the top-1% threshold (4,526) is 167.6 times the median (27). In a distribution that skewed, the arithmetic mean sits far above the median and describes almost nobody. If someone tells you the average podcast episode gets several hundred downloads, they are describing a number most shows will never reach. Ask for the median and the percentiles instead.

Why are podcast downloads falling?

Podcast downloads across the whole industry fell 41.3% from July 2023 to July 2026, driven by an iOS 17 change to how Apple Podcasts handles automatic downloads. Listening over the same period hit record highs.

Over that period active podcasts fell 9.4% and new episodes 15.1%, while downloads per new episode dropped 30.9%. The median 7-day figure went from 31 to 27. That is not shows getting worse.

iOS 17 shipped on 18 September 2023. Apple’s description: “episodes that are older than seven days will not be considered ‘new’ and automatically downloaded”, and “Automatic Downloads are paused when a listener hasn’t played a show they follow (latest five episodes for more than 15 days) to preserve device storage.”

Daily shows with habitual audiences saw declines of roughly 10%, shows on other cadences averaged 30% declines in monthly downloads, and most large shows dropped an average of 15%. The change cost Acast $7.2m after a 12% listening slump, and Audioboom an estimated $9m.

Actual listening over the same stretch: weekly listening is 130 million people, up from 115 million a year earlier.

Downloads down 41%. Listening at an all-time high. The iOS 17 mechanism connects them: it removed phantom downloads that were never heard without removing a single listener. The practical consequence for a budget review: if your download chart is down over the last three years, a meaningful part of that has nothing to do with your show. If nobody in the room knows that, your show gets killed for a software update Apple shipped in 2023.

Will podcast plays replace downloads?

A proposed replacement unit exists, adoption is voluntary, and plays are expected to become the default trading currency in the second half of 2027.

On 11 June 2026 Spotify redefined a play: “Play counts will now reflect the number of times an episode was watched or listened to for at least thirty seconds.” Its stated reasoning is that “thirty seconds is a reliable indicator of intentional consumption, filtering out accidental starts, and other low-intent activity.”

On 23 July 2026 the Alliance for Measurement in Podcasting, an industry body convened in July 2025 by media agency Oxford Road, published the AMP Accords. AMP has 12 voting members including Spotify, SiriusXM Media, Libsyn, Podscribe, DraftKings, BetterHelp, UTA, FlightStory and Shopify. Its three stated workstreams are standardising impression metrics, developing cross-platform performance measurement, and defining what counts as a podcast at all.

The proposed replacement unit is a Podcast Play, defined as “a user consumes at least 30 consecutive seconds of a podcast, whether audio or video”, with Podcast Audience defined as “the number of unique users who played a podcast at least once.” AMP’s definition of a podcast is content that “works with your eyes closed”, audio or video, RSS-distributed or not.

AMP describes these as “companion metrics, intended to be phased in during a transition period, so that nobody’s numbers change overnight”. The framework has not yet been through IAB technical review; AMP says it will “deliver its outputs to groups like IAB for consideration and deeper technical review.” Several major players are absent: iHeartMedia, NPR, Audacy and Westwood One.

Dan Granger of Oxford Road, on what was hard about it: “The hard part was never writing the framework. The real accomplishment was getting companies with competing interests to agree on anything at all.”

When is a small podcast audience fine?

A small podcast audience is fine when the break-even deal count is low and you can name specific listeners inside your target segment. It is not fine as an assertion with nothing behind it.

The arithmetic: a show whose fully loaded annual cost is about the size of one contract needs one deal. Where ACV is four times the annual cost of the show, break-even is a quarter of a deal, which in practice means one deal every four years is break-even and anything more is upside.

Then ask a harder question than “how many listeners”. Ask how many people exist in the entire world who could sign that contract. For most enterprise B2B categories the answer is a few thousand, sometimes a few hundred. Against a total addressable audience of 800 people, 27 downloads in the first week is 3% of everyone who matters, in a week, for the length of a full conversation.

Two guardrails so this does not become an excuse:

• You must be able to name the people. If you cannot point to specific listeners in your target segment, from Apple’s Engaged Listeners trend, guest relationships, replies or CRM notes, you are not making the small-audience argument. You are avoiding measurement.
• Set the ratio against benchmarks, not vibes. Median CAC payback is 16 months, top quartile 6 months or fewer, bottom quartile 24 months or more. A channel that takes four years to return its cost may still be worth funding for other reasons, but it should be argued for on those reasons, not smuggled through as ordinary marketing spend.

How to set a podcast ROI target

Set the podcast ROI target by writing down the break-even deal count, the review horizon, and the pre-launch baseline before episode one publishes. All three cost nothing to capture and cannot be recreated later.

The failure mode is predictable. A show launches with no defined success criteria. Eighteen months later somebody in a budget review asks how it is performing, a download number gets produced, and it is judged against a figure invented in that meeting by someone comparing it to a media property with full-time staff. The show dies for missing a target that never existed.

Do this instead, before episode one:

1. Write down the break-even deal count. Fully loaded annual cost divided by ACV. One number, one sentence, agreed with whoever owns the budget. This is the primary target and it is a business outcome, not a download figure.
2. Set the review horizon in the same sentence. Four quarters is a reasonable floor for a B2B show given sales cycles. Agree it in writing so it cannot be shortened later.
3. Capture the pre-launch baseline now. Branded search for the last full quarter from Search Console. Current inbound demo volume. Current count of self-reported attribution mentions. Screenshot them. In six months you cannot recreate a baseline you did not take.
4. Instrument attribution before the first episode publishes. Open-text “how did you hear about us” on the forms, a CRM field for podcast guests, a tracked link convention for sales usage. Retrofitting this is how shows end up unable to prove anything.
5. Name secondary indicators, with their windows. Apple Engaged Listeners quarter over quarter. Spotify completion rate. Median 7-day downloads against the 7-day percentiles. Write the window next to each metric so nobody compares across them later.
6. Write down explicitly that downloads are not the primary metric, and why. Include the iOS 17 note. This one paragraph, agreed in advance, prevents the most common way a working show gets cancelled.
7. Define the stop condition.

A target agreed before launch is one negotiated when nobody was under pressure. A target invented during a budget review is one negotiated while losing.

When should you stop a podcast?

Stop the podcast when, after four full quarters of consistent publishing and properly instrumented measurement, all of the following are true:

• No self-reported attribution mentions of the show have appeared in the CRM.
• Branded search shows no measurable movement against the pre-launch baseline.
• No sales rep has used an episode in a live deal.
• No guest relationship has entered pipeline.
• Platform-side engagement, Apple Engaged Listeners and Spotify completion rate, is flat or falling on its own trend line.

When every measurable and every reasonable proxy is flat simultaneously, the remaining case for the show is unfalsifiable. “It builds brand” with no observable trace in any of five independent instruments is not a strategy, it is a hope. Stop, and put the budget somewhere with a feedback loop.

Sunk production cost is not a reason to continue. The equipment is bought and the format is built, which lowers the cost of continuing but does not change whether continuing is worth it.

One thing that is a legitimate reason to continue with weak numbers: if the show is your primary mechanism for getting time with people you otherwise cannot reach, and you can name those people, then it is a business development instrument that happens to publish. Judge it as one. Count the meetings, not the downloads.

Frequently asked questions