Guide
How to start an interview podcast
Most advice about starting a podcast is written for a show that doesn't exist yet in any particular form. This one assumes you've already decided: one host, one guest, an hour or more, unscripted. A long-form interview show.
That's a specific format with specific failure modes. Interview shows almost never die from bad audio. They die at episode 14, when the host walks into a conversation underprepared and the guest gives the same answers they've given nine times that year. Everything below is ordered by how much it actually decides whether your show survives its first year.
Settle the format before you buy anything
The format question that matters isn't interview versus solo. You've answered that. It's length, cadence, and whether you have a co-host.
Length should be whatever the conversation is worth. If it's done at 42 minutes, end it at 42 minutes. Padding a good conversation to hit a round number is the most common way a promising early episode gets ruined.
Cadence is the decision people get wrong. Weekly is the default answer, and weekly is brutal when every episode requires booking a guest, doing real prep, recording, and editing. Every other week is a legitimate schedule. Pick the cadence you can hold during a bad month, not the one you can hold during the excited first month.
On co-hosts: a second host halves the burden of carrying a conversation and doubles your scheduling friction. If your co-host won't reliably be there, they aren't a co-host, they're a recurring guest. Naming it correctly saves you from a show that feels broken every time one person is missing.
Equipment, honestly
A decent USB microphone beats a bad XLR setup, and most first setups are bad XLR setups. An interface with the gain set wrong, a condenser mic pointed at an untreated room, and a cable that picks up hum will sound worse than a $100 USB dynamic used correctly.
The things that genuinely change how you sound, roughly in order:
- Get close to the mic. Four inches, not eighteen. Distance is the difference between "podcast" and "conference call" and it costs nothing.
- Use a dynamic mic, not a condenser. Dynamics reject the room, and your room is untreated. This is why studios you've seen on YouTube use them.
- Fix the room before you upgrade the mic. Soft surfaces, avoid bare parallel walls, don't record in a kitchen or a bathroom. A rug and a bookshelf beat a mic upgrade.
- Headphones for everyone, host and guest. Speakers put the other person's voice back into your mic, and no editor can undo that.
- Check levels before you start, and record a backup. Discovering a dead channel at minute 40 of a conversation you can't recreate is the worst afternoon in podcasting.
The upgrade to XLR and an interface makes sense when you have a permanent second mic position in a room you control. Before that, it's spending money to add failure points.
Remote or in the room
In-person is better for long unscripted conversation, and it isn't close. Overlap, interruption, and the timing of laughter all depend on hearing someone in real time. Remote adds a couple hundred milliseconds, and two people quickly train themselves to wait for silence before speaking. That's why so many remote interviews feel like alternating statements instead of a conversation.
Record remotely anyway if that's what gets you guests. A guest you can actually book beats a better recording of a guest you can't. Just record local tracks on both ends, either with a platform that does it or by having the guest record themselves, and treat the live stream audio as a backup rather than the master.
Send guests a two-sentence setup note: headphones, a quiet room, wired internet if they have it. Not a PDF. The longer your instructions, the more likely they get skimmed.
The first ten episodes
Your first ten episodes exist to turn you into someone who can host a show. They're reps, not a launch. Treating them as reps takes the pressure off and, counterintuitively, makes them better.
Two problems show up immediately. Interesting guests don't want to come on a show with no episodes, and you don't yet know what your show sounds like.
The first one you solve by starting with people who'll say yes for reasons other than your audience size: people you already know, people whose work you've written about publicly, people early enough in their own visibility that a thoughtful hour is worth their time. Book six of those before you record any of them, because your yes rate will be lower than you expect and the gaps are what kill momentum.
The second one only gets solved by doing it, but you can speed it up. Relisten to your own episode within two days of recording. It's unpleasant and it's the fastest feedback loop available. Write down every moment you talked over an answer, asked a two-part question, or moved on right when the guest was about to say the real thing.
Publish three episodes on day one so a new listener has something to stay for. Then keep going. Don't relaunch, don't redesign, and don't delete episode one until you're at episode 30 and nobody's looking.
Prep is the actual differentiator
Everything above is table stakes. Prep is where interview shows separate from each other.
A guest with a book has done forty interviews about that book. If your questions come from their bio and the press kit, you'll get the same answers everyone else got, delivered slightly more tiredly. The listener has no reason to pick your version.
Real prep for an interview show looks like this. Listen to their last two or three long-form interviews, not to learn their story but to find the question they're bored of and the thing they said once that nobody followed up on. Find one place where their public position has an edge to it, something unresolved rather than a gotcha. Come in with three questions nobody has asked them.
And know your own archive. By episode 30 you'll have covered things on your own show you no longer remember covering. "A guest last spring argued the opposite of that" is the most valuable move a long-running interview show has, and it depends entirely on recall you won't reliably have while you're also listening, watching the clock, and thinking about the next question.
That's the gap prep documents can't close, because you can't read a document while you're talking. It's the problem Tally works on: live intelligence, meaning the right information appearing on screen while you're still talking, pulled from your prep and your own back catalog. Useful later, once you have an archive worth searching. For episode three, a good set of notes is fine.
Distribution is simpler than it looks
As far as the platforms are concerned, your podcast is an RSS feed. A podcast host stores your audio files and generates that feed. Apple Podcasts, Spotify, Overcast, and the rest just read it. Nobody hosts your episodes but your host.
Which means the only lock-in worth worrying about is whether your host lets you export your feed and set up a redirect to a new one. If it does, every other decision is reversible. Submit the feed once to Apple and once to Spotify. Most of the smaller directories pull from Apple's index, so you don't need to chase them individually.
Two details people get wrong. Cover art has to be legible at thumbnail size, so make it at 3000 by 3000 and then look at it at 60 pixels wide before you commit. And put the guest's name and the actual subject in every episode title. "Episode 14" is invisible to anyone searching inside a podcast app, which is where a surprising share of your early listeners will find you.
What to skip until episode 20
A custom website. A video studio. Sponsor outreach. A Patreon. A commissioned intro track. A logo you pay for. A newsletter. Editing out every breath. A second show.
All of these feel like progress and none of them compound before you have a catalog. Twenty episodes is roughly where you can tell whether the show is good, whether you still want to make it, and what it's actually about, which is usually not what you thought it was about at episode one.
The one administrative thing worth starting on day one is a list of every guest you've contacted and what they said: yes, no, or not right now. By episode 30 that list is your booking pipeline, and rebuilding it from memory and old email is miserable.
When prep stops fitting on a page
Tally is in an invite-only beta, going out in waves. Leave your email and we'll get you in as capacity opens up.