I got two files on the same Tuesday. One was a 74 minute interview with almost no structure and a host who talked for eleven minutes before the guest said a word. The other was a 22 minute scripted piece with tape, music beds, and a written cold open. Both hosts told me the same thing on the intake call: "we want it to feel like a story."
Only one of them had recorded a story. The other had recorded a conversation and hoped I could turn it into a story in post. I can do a lot in post. I cannot do that.
The format is a decision you make before you press record, not a flavour you add afterwards. Almost every show I get that feels flat is a show that never made the decision. It is half narrative, half hangout, and the two halves keep interrupting each other.
What each format actually is
A narrative show is written. There is a script, or at least a beat sheet. The host is not discovering things in real time, the host already knows the ending and is walking you toward it. Tape gets used as evidence. A guest is not a guest, a guest is a source who provides one or two lines that prove a point the script is making. The unit of work is the sentence.
A conversation show is not written. The value is that nobody knows where it goes. The host is a good listener with a rough map. The guest gets to wander, contradict themselves, land on something they did not plan to say. The unit of work is the exchange.
Those are different products. They get consumed differently, they get shared differently, and they get edited with completely different instincts.
The hybrid trap
The most common thing I see is a conversation show wearing a narrative costume. Scripted cold open, music sting, a promise that this episode will reveal something. Then 55 minutes of two people catching up. The intro wrote a cheque the body cannot cash.
It does not fail loudly. It fails quietly, in the drop off numbers. People stay for four minutes because the open was good and leave at nine because the show stopped being the thing the open promised.
The reverse happens too, and it is rarer but worse. A genuinely narrative idea, a real investigation with documents and stakes, recorded as a loose chat between two friends. All the good material is in there. It is just buried under twenty minutes of throat clearing that has to be surgically removed, and when you remove it the piece has no connective tissue left, because nobody wrote any.
If you cannot describe what the episode proves in one sentence, you did not record a narrative. You recorded a conversation, and that is fine, but edit it like one.
How I tell which one I am holding
When a new show comes in I do a fast pass before touching a timeline. Three questions.
- Does the host ask questions they already know the answer to? If yes, it is narrative. That is not a criticism, that is craft.
- Is the best moment planned or accidental? Narrative shows peak where the script says. Conversation shows peak somewhere nobody expected, usually about 40 minutes in when the guest forgets they are being recorded.
- Would the episode survive being reordered? Narrative dies if you move blocks around. Conversation usually survives, and often improves.
That third one is the tell. If I can move a ten minute chunk from the back to the front and nothing breaks, there was no argument being built. There was just talking. Again, fine. But then the edit job is compression and rhythm, not architecture.
What changes in the edit
On a conversation show my job is mostly subtraction and pacing. Cut the setups, cut the second time the guest explains the same idea, tighten the gaps between a question and the answer so the exchange feels alive. I am protecting energy. The most common failure is guests who take three sentences to start saying anything, which I wrote about in why your guest answers are too long. Fixing that is half of what makes a chat show feel sharp.
On a narrative show my job is construction. I am moving blocks, building beds under transitions, deciding where a piece of tape lands so it hits after the setup rather than before it. Silence is a tool, not a defect. A two second gap before a line can be the best edit in the episode. On a conversation show that same gap is dead air.
This is exactly why I will not run a show through silence detection and call it done. Our editors watch the whole episode, every time. A machine looking for gaps cannot tell you that the pause at 31:40 is the emotional centre of the piece and the pause at 12:05 is a guy looking for his water bottle. I have opinions about how much of an edit can be automated, and most of them ended up in what I learned cutting a thousand episodes.
Length follows format, not the other way around
Narrative episodes want to be shorter. There is a fixed amount of story and once it is told, the episode is over. Twenty to thirty five minutes is normal. Padding a narrative piece to hit an hour is how you turn a good story into a boring one.
Conversation episodes have no natural length. That is the danger. They end when someone gets tired, which is not a structural decision. Hosts tell us they feel pressure to keep going because the guest travelled, or because a short episode feels like poor value. It is not. Nobody ever complained that a show respected their commute. I went deeper on this in how long should your episode really be, and the short version is that the correct length is the length where nothing is boring.
There is a production cost angle here too. Narrative work is front loaded, most of the labour happens in scripting and structure. Conversation work is back loaded, most of it happens in the cut. That is worth knowing before you build a schedule you cannot keep, which is the argument I keep making against a weekly release cadence you have not costed out.
Clips behave differently too
This is the part people underestimate. Conversation shows are clip machines. A good exchange is self contained, it has a question and a punchline, and it works with no context on a phone screen. I can usually pull three or four strong verticals from an hour of real conversation without straining.
Narrative shows resist clipping. The best 45 seconds of a scripted piece usually depends on the four minutes before it. Cut it loose and it is just a person saying words in a serious voice over music. When a narrative show wants social cuts, we generally have to build them, which means writing a tiny standalone version rather than lifting a moment.
So if short form distribution is your main growth channel, that is a real argument for conversation. And if you are pushing episodes to YouTube, the packaging matters as much as the cut, which is why our sister studio ClickTheory exists at all. A narrative episode with a lazy thumbnail is an episode nobody clicks.
The show notes tell on you
Here is a cheap test. Write the show notes before you record. If you can write them, you have a narrative. If you genuinely cannot, because you do not know what the guest will say, you have a conversation.
Most people write the notes afterwards and they end up as a wall of timestamps nobody opens. I have a whole rant about that in show notes nobody reads, but the format point stands on its own: notes are easy for narrative shows because the structure already exists, and hard for conversation shows because you have to find the structure after the fact. That is work. It is why chapters and notes are an add on for us rather than something we pretend is free.
Picking, and then actually committing
My honest bias is that most independent hosts should run a conversation show and stop apologising for it. Narrative audio is expensive. It needs a writer, a producer who can think in structure, tape you had to go out and get, and a tolerance for spending three weeks on twenty five minutes. If you have that, it is the most rewarding format in audio. If you do not, you will produce a thin imitation of it and burn out by episode nine.
What I would push back on is the idea that conversation means unprepared. The best chat shows I cut are heavily prepared and lightly scripted. The host has done the reading, has five sharp entry points, and is willing to abandon all of them the moment something better appears. That is a craft, and it is a different craft from writing.
If you are setting up a show now, decide the format first, then decide the room, then decide the cameras. In that order. A two camera setup on a conversation show earns its keep because reaction shots carry the exchange, which is the whole logic behind multi-cam editing. A narrative show often needs no camera at all.
And if you are handing episodes to an editor, tell them which show they are cutting. It sounds obvious. It almost never happens. We ask on intake for exactly this reason, and it is baked into how we run remote podcast editing so the first cut comes back shaped the way you intended rather than shaped the way a stranger guessed.
Pick one. Then let the edit serve that choice instead of hedging between two. If you want a second opinion on which one your show already is, send me an episode and I will tell you what I actually hear in it.