I sat with a stopwatch on a Tuesday, going through an episode a host had sent us, and I timed every guest answer. The shortest was 22 seconds. The longest was 6 minutes and 40 seconds. The average was a hair over 3 minutes. That episode ran 74 minutes and the host spoke for about 11 of them.
Here is the part that matters. Of that 6 minute answer, maybe 50 seconds were good. The rest was the guest circling the runway, re-explaining a point they had already made, and then landing on the same conclusion twice because nobody told them they had already landed.
Long answers are almost never a sign of a generous guest. They are a sign that nobody in the room set the length. That is the host's job, and most hosts avoid it because they think editing will clean it up later. Editing can clean it up. It costs you something every time.
The four minute answer is not a depth problem
The story hosts tell themselves is that a long answer means a thoughtful guest. Sometimes that is true. Usually it is not.
When I listen back to long answers, they break into a handful of shapes, and only one of them is depth:
- The warm-up. The guest spends 40 seconds telling you what they are about to tell you. Pure preamble.
- The résumé detour. A question about a decision turns into a career timeline nobody asked for.
- The safety loop. The guest is not sure they answered it, so they answer it again from a slightly different angle. Then a third time.
- The list with no exit. They start counting reasons and cannot find a way to stop counting.
- Actual depth. A single idea, developed, with a beginning and an end.
Only that last one earns four minutes. The other four shapes are anxiety wearing a suit. Nobody rambles because they are confident. They ramble because they are not sure whether they have given you enough, and silence from the host reads as "keep going".
Silence is an instruction, and hosts give it constantly
Podcast advice from about 2015 onward told hosts to shut up. Let the guest breathe. Do not step on them. All of that was a fix for a real problem, which was hosts who could not stop talking about themselves.
The fix overcorrected. Now I hear hosts who nod for six minutes straight, and the guest, reading nothing back, keeps talking because the room has given them no reason to stop.
A guest cannot tell the difference between "you are doing great, continue" and "I have checked out and am waiting for you to finish". They both look like a nod.
So the guest picks the safer read, which is to keep going. Then we get the file, and there are eleven of those in an episode, and someone has to sit through all of it. That someone is a human editor at our end, watching the whole episode. We do not cut by silence detection. That means every one of those six minute answers gets watched in full by a person deciding which 50 seconds survive.
What long answers actually cost you
The cost is not editing hours. It is more specific than that, and it shows up in four places.
You lose the clip. A tight 90 second answer with a clean start and a clean end is a vertical clip on its own. A six minute answer has a good 50 seconds buried at minute four, with no clean entry point, because the setup for that thought happened three minutes earlier. We can still build something from it, but it takes stitching, and stitched clips always feel slightly built. Hosts who want a real pipeline for short vertical clips from every episode get far more out of a guest who lands answers in two minutes than one who takes eight.
You lose the trailer. Trailers live on moments where a sentence starts and finishes with force. When every answer meanders, the trailer cut for the episode becomes a hunt for fragments rather than a pick of the three best beats.
You lose the middle of the episode. The listener drop point is rarely the intro. It is minute 18, on answer number four, when they realise the guest has been on the same thought for a while and the host has not stepped in. That is where people leave.
You lose your own questions. If the guest takes four minutes an answer, a 70 minute recording gets you maybe fifteen questions. You had thirty. The best ones were at the bottom of the list, and you never reached them.
Say the length out loud
The single change that fixes most of this takes eleven seconds and happens before you press record.
Tell the guest the shape of the show. Something close to: "Most answers here land around a minute or two. If I want more I will ask for more, and if I jump in it is because I want to go deeper, not because you did something wrong."
That last clause matters more than the first. You are not asking them to be brief. You are telling them that interruption is a normal part of this conversation and not a rejection. Once a guest knows that, they stop over-delivering out of fear, and they stop treating your first interjection as a correction. If interrupting still feels rude to you, I wrote about how to cut in without breaking the rhythm, because the technique is mostly about where in their sentence you enter, not how loud you are.
Do this on the call, not in the pre-interview email. Emails are skimmed. A sentence said out loud two minutes before recording is remembered.
Ask questions that have an end
Half of the long answer problem is the question. Broad questions get broad answers, and a broad answer has no natural stopping point.
Compare these two:
- "Tell me about your career."
- "What was the first job where you felt out of your depth?"
The first has no exit. The second has a built in ending, because once the guest describes that job, the answer is complete and they know it. Questions with a concrete anchor, a moment, a number, a decision, a specific year, come back at a workable length on their own. You do not have to police anything.
The other trick is to ask for the count. "Give me two reasons, not five." Guests love this. It removes the guesswork about how much you want, and it is the only time I have seen someone actually stop at two.
Remote makes it worse, and you should plan for that
In a room, a guest reads your posture, your eyebrows, the half breath you take when you want to come in. On a call they get a slightly delayed rectangle of your face, and if you are looking at your notes rather than the lens, they read that as disinterest and, again, keep talking to fill it.
Latency also makes gentle interruption physically harder. You start, they do not hear you for 200 milliseconds, you both stop, you both restart. Two of those and the host quietly decides not to try again for the rest of the episode. That is how a remote recording turns into a monologue, and it is a big part of why remote guests feel harder to steer than in person ones.
If you record remote, do the length conversation before you hit record and then use verbal signals rather than visual ones. A short "love that, and on the second part" works where a nod does not.
What we can and cannot fix in the edit
Being honest about this is more useful than selling you a rescue.
What a human editor can do with a long answer:
- Remove the warm-up. Almost every long answer has a 30 to 45 second front end that adds nothing. This is free.
- Collapse the safety loop. If they made the same point three ways, we keep the best one. Also mostly invisible.
- Reorder within an answer, carefully, so the conclusion is not buried under two minutes of setup.
- Cover a cut with a second angle where there is one, which is why multi-cam editing buys you so much more room to tighten. On a single camera the same cut needs to be gentler or hidden.
What we cannot do:
- Invent an answer the guest never gave.
- Make a wandering answer feel like a decisive one. Trimmed rambling is still rambling, just shorter.
- Manufacture a clip start where the guest never opened cleanly.
So an episode with long answers does get better in post. It does not get to where a well steered recording starts. And if you are early in the run, still building the reflexes, that gap is the thing that compounds. It is one of the quieter parts of the grind of the first hundred episodes: the recordings get shorter and better before the audience notices anything at all.
The number I would give you
If I had to hand a host one target, it is this. Aim for your guest's median answer to sit between 90 seconds and two and a half minutes, with two or three deliberate exceptions per episode.
The exceptions are the point. You want a couple of long answers, and you want to have chosen them. When you ask a guest to go deep on a specific thing and they take five minutes, that is a decision. When every answer is five minutes, that is a default, and defaults are what make episodes feel the same.
Try it on one recording. Say the length out loud, ask two questions with a hard anchor, interrupt once in the first ten minutes so the pattern is established early. Then listen back and time the answers. My guess is the median drops by a third and the episode gets shorter without losing a single idea.
If you want the edit side handled while you work on the room side, our first cut lands in 48 hours and a person watches the whole thing before anything gets cut. But most of this one is on you, in the first eleven seconds, before the record button.