How to organize podcast research
Research for a podcast isn't hard to gather. It's hard to retrieve. You save the article, and three weeks later when you're actually recording, you can't remember whether you saved it, where you put it, or what it said.
Almost every system fails the same way, and it isn't a discipline problem. It's that the moment you save something and the moment you need it are separated by weeks, and the version of you that filed it made assumptions the version of you that's searching doesn't share.
The shape of the failure
You find something good on a Tuesday. Maybe it goes in a browser tab that stays open for a month. Maybe a Slack message to yourself. Maybe a bookmarks folder, a Notes file, a Notion database with three properties you filled in for the first four items and then stopped.
On recording day you know you saw a chart about the thing. You don't remember which of those five places it went, so you check two, don't find it, and move on without it. The research existed. It just wasn't retrievable in the ninety seconds you had.
Three specific things break:
You file by source, you retrieve by topic. Saving is organized around where it came from. Retrieving is organized around what you need. A bookmarks folder called "articles" is the wrong index for "the study about how many hours founders sleep."
The tax is paid at the wrong end. Good filing takes effort at save time, which is when you're distracted and it isn't due yet. Retrieval is when you're under pressure. Any system that front-loads effort onto the distracted moment stops being used by week six.
Video and images don't get filed at all. You can paste an article's key paragraph into a note. Nobody transcribes a forty-minute YouTube video into their research doc, so it stays a bare link, and a bare link is unsearchable. The best material in a lot of research is a chart, a clip, or a screenshot, and those are exactly the things text-based systems drop.
Two systems that work
Only two survive a busy month, and they fail in opposite directions.
Save nothing, research on demand. No inbox, no backlog. When an episode is scheduled, you do the research for that episode in one sitting and use all of it within days.
This is better than it sounds. Nothing goes stale, there's no maintenance, and you never search for something you might not have saved. The cost is that you lose the serendipitous find. The thing you stumbled on in March that would've been perfect for the September guest is gone, because you never wrote it down.
Works well for shows on a predictable cadence with booked guests. Works badly for shows where the subject is whatever's interesting that week.
Save everything, organize nothing, search well. One place, no folders, no tags, no properties. Everything goes in with zero decisions at save time, and the entire burden moves to retrieval.
This only works if search is genuinely good, which is the catch. It has to search inside the things you saved, not just their titles. Otherwise you've built a pile.
The middle path is where systems die. A folder structure you maintain sometimes, tags you apply for a while, a database with fields you fill in when you're not busy. Half-maintained systems are worse than either extreme, because you can't trust the result. When search comes back empty you can't tell whether the thing isn't there or whether you just filed it wrong that week, so you go check the other places anyway.
Making the pile searchable
If you're taking the save-everything route, the thing that makes it work is search that reaches the content rather than the label.
For articles that means the body text, so you can find the paragraph, not just the headline. For video it means the spoken words with timestamps, so "the bit where he talks about margins" lands on the minute rather than the video. For images and charts, it means something has described what's actually in them, because a file called Screenshot 2026-03-14.png is invisible forever.
That last one is the biggest gap in most setups. The chart you screenshotted is often the single most useful artifact you have, and in a normal research folder it's less findable than a tweet.
Two habits cost almost nothing and pay off later. Save the source URL alongside anything you clip, because you'll want to check the claim before you say it out loud. And when you save something, add one line about why, in your own words. Not tags. A sentence. "Useful counterargument to the productivity thing." Future you searches in that language, not in the language of the article's headline.
Bring it to the session, not to a document
Research that lives in a doc you have to alt-tab to is research you won't use while recording. If you're on video, looking away is visible. Even on audio, the pause while you find the tab is the pause everybody hears.
So the last step is getting it into view without leaving the conversation. A printed page for the two or three things you're sure about works. A second screen with your notes works. What doesn't work is planning to search for it mid-interview, because you won't, and if you try, everyone will hear it.
This is where our own product sits. Tally's library takes articles, clips, screenshots, and charts, reads the text and watches the video, and makes the contents searchable rather than the filenames. That means you can search for the chart about lithium prices three months later and land on the exact second it's on screen in a video you saved and forgot.
The part that matters during a recording is live intelligence, which means the right information appearing on screen while you're still talking. Your producer can push a saved clip to the studio display, or the app surfaces it on its own when the conversation reaches it. It makes no sound and stays out of your recording chain.
It runs on macOS today and it's invite only while we onboard shows one at a time. Your published archive is searchable the same way, which is a related but separate problem.
Tally is invite only right now
We're onboarding shows in small waves so we can set each one up properly. Leave your email and we'll reach out when the next wave opens.