Podcast Analytics: The Metrics Worth Editing For

Four numbers matter: audience retention, average consumption percentage, follower growth, and conversion. Everything else is context.

Retention is the shape of the graph showing where people leave. Consumption is Apple's version, a single percentage of the episode the average listener finished. Follower growth tells you whether the people who arrive decide to stay. Conversion is whatever the show is for: a booked call, a newsletter signup, a demo request.

Downloads are the number everyone quotes and the number I trust least. A download fires when an app fetches the file. It fires when a phone auto-downloads overnight for someone who has not opened your show since March. An episode can add 900 downloads and 40 minutes of total real listening.

If your show sells something, the honest metric is cost per booked conversation. One of the shows we cut gets fewer downloads this year than last and twice the inbound, because the host changed who the episodes were for. The graph got shorter and the pipeline got longer. That is a win.

Pick two numbers per quarter. Retention plus one business outcome is enough. Tracking twelve metrics means acting on none of them, and I have watched hosts spend more time in dashboards than in the recording chair.

Read a retention graph in three zones: the first 60 seconds, the mid-episode slope, and the ending cliff. Each zone has a different cause and a different fix.

The first 60 seconds is where the steepest drop always lives. Losing 15 to 25 percent there is normal. Losing 40 percent means your opening is doing housekeeping instead of promising something. A cold open of the sharpest 20 seconds from the middle of the conversation fixes more first-minute drops than any other single change I make. That is a pure editing decision, which is why I bring it up in almost every kickoff on how it works.

The mid-episode slope should be close to flat. A gentle decline is healthy attrition. Visible steps down at 14 minutes, 31 minutes and 48 minutes usually map to something concrete: a long tangent, an unedited technical hiccup, a guest anecdote that ran four minutes past its point.

The ending matters less than people think. Everyone drops at the outro. Judge the last five minutes before the outro instead.

On video, YouTube gives you the same graph with the ability to scrub to any dip and watch what happened. That is a gift. Spend twenty minutes there per episode and you will learn more about your own show than a year of download screenshots.

For an interview show, aim for 60 percent or better average consumption on audio and 30 to 40 percent average view duration on YouTube. Those are different scales measuring different behaviours, so never compare them to each other.

Audio listeners are captive. They are driving, walking, doing dishes. They let episodes run. Apple's consumption figures sit high because of that. Video viewers have a sidebar full of alternatives and a thumb. A 45-minute video podcast holding 35 percent means roughly 16 minutes of attention per view, which is enormous by video standards and unremarkable by audio standards.

Episode length changes the benchmark too. A 25-minute episode will beat a 75-minute episode on percentage almost every time while delivering less total attention. This is why I ask hosts what they want optimised before we set an edit style. Watch time and completion rate pull in opposite directions.

Benchmark against yourself. Your last ten episodes are the only fair comparison set, because they share your host, your topic and your audience. Industry averages come from mixed pools of daily news shows and 3-hour rambles, and they will tell you nothing useful about whether Tuesday's episode worked.

Ignore raw downloads, chart positions, and social vanity counts on clips. They move for reasons unrelated to whether your show is working.

Chart positions are velocity-weighted and gameable. A launch week of friends subscribing can put you at number 4 in a category, and it means nothing about week twelve. I have never seen a chart screenshot close a deal.

Clip view counts are the trickiest trap. A vertical clip can hit 200,000 views because the first frame was funny, and send eleven people to the full episode. I still think clips are worth cutting, and we make them at $25 each, but judge them by profile visits and follows, not by the big number. There is more on this in the piece on clips.

Also ignore your own listen-through of the raw recording as a quality signal. Hosts consistently rate episodes they enjoyed recording higher than the ones that perform. The graph disagrees with your memory, and the graph is right.

One more: unique listener counts across platforms cannot be added together. Apple, Spotify and your host will each count the same human differently. Adding them produces a bigger number and a worse decision.

Editing moves retention more than topic selection does, because most drop-off happens at moments of friction rather than moments of boredom. Friction is fixable in the timeline.

Here is what we do on the shows we cut, and what it tends to affect. Cold opens move the first-minute cliff. Tightening dead air and false starts flattens the mid slope. Cutting a rambling answer from four minutes to ninety seconds removes an entire step down. Chapters give people a reason to jump instead of leave, which is why show notes and chapters at $35 are the add-on I recommend most for interview shows. On video, cutting to the second camera on a reaction resets attention in a way that a static frame never will, which is the whole argument for multi-cam editing.

This is also where automated tools hit a ceiling. Silence detection finds gaps. It does not know that the guest's best line came after an eight second pause while they thought. Human editors watch the whole episode here, and nothing gets auto-cut by silence detection, because the pause before the good answer is often the reason the good answer lands.

Consultants and specialists tend to feel this fastest. I go deeper on that in the notes on podcast editing for consultants.

Yes, and mixing them is the most common analytics mistake I see. Discovery is how people arrive. Retention is what happens after. They need different fixes and different review cadences.

Discovery lives in your titles, thumbnails, search presence and clips. If impressions are flat, no amount of editing helps, because nobody is showing up to be retained. If impressions are healthy and click-through is under 3 percent on YouTube, the problem is packaging, not content. That is a thumbnail and title conversation, and it is why our sister studio ClickTheory does thumbnails at 8 for $250 rather than us bolting it onto the edit.

Search presence is its own lane again. Titles, descriptions and transcripts do the work there, and I wrote the practical version of that in the podcast SEO guide.

Review discovery monthly and retention per episode. Discovery numbers are noisy week to week, so a monthly view stops you from rewriting your whole title strategy because one episode landed on a public holiday. Retention is stable enough to read straight after publish, usually about seven days in once the bulk of listens have landed.

Thirty minutes, once a month, four questions. That is the whole system, and it beats a daily dashboard habit every time.

Question one: which two episodes had the best retention, and what did they have in common? Usually it is a tighter guest, a narrower topic, or a cold open that promised something specific.

Question two: where is the shared dip? Open three retention graphs side by side. If all three sag at minute 12, you have a format problem at minute 12, probably a segment that stopped earning its slot.

Question three: what did the show produce? Calls, replies, signups, inbound. Write the actual number, not a feeling.

Question four: what one change goes into next month's edits? One. Not five. Then tell your editor, because they can only act on what they know. If you are working with an agency for video podcast editing, this note should take you ninety seconds to send and it changes every episode that follows.

Hosts tell us the review is the part they skip when they get busy, and it is the part that compounds. Put it on the calendar the day after your monthly invoice lands so it is tied to something you already do.