Five lessons, just over an hour. "Can you look into this for me" is the most common request in this whole job, and the least defined. This module is how you turn a vague ask into a research question with an edge to it, know where to look and when to stop, verify what you find instead of trusting the first result, and hand it back as something Adaeze can act on in five lines.
Adaeze messages: "Can you find me some podcasts I could be a guest on? Trying to get the practice's name out there a bit more." That is a real request and a completely unresearchable one, in this form. "Some" is not a number. "Get the name out there" is not a criterion anything can be checked against. Handed straight to a search box, this produces an hour of open-ended browsing that ends only when you get tired, not when the question is answered.
A research task without a stopping point doesn't end when it's finished. It ends when you give up.
Almost every request that arrives shaped like Adaeze's is missing the same two things, and your job before you open a single tab is to supply them:
Write the brief first, as a short exchange with Adaeze if you can, or as your own working assumption if you can't reach her quickly. Five minutes here saves an hour of research aimed at the wrong target.
Original ask: "Find me some podcasts I could be a guest on."
Research question: Which active, guest-accepting podcasts have an audience likely to include people seeking a therapist — mental health, wellness, women's health, or Nigerian-diaspora-focused shows — that Adaeze could realistically pitch herself to as a guest?
Criteria to count: Takes outside guests (not solo format). At least one episode in the last two months. English-language. Audience plausibly includes people who'd consider therapy — inferred from topic and past guest list, not just the show's stated category.
Stopping point: Thirty qualifying shows, or four hours of research, whichever comes first — with a note on which limit was hit.
Notice the stopping point names both limits and says which one to report against. That single addition is what stops "I ran out of time" and "I found everything there is" from looking identical in the final handover — they are very different situations, and Adaeze needs to know which one she's getting.
Sometimes there's no time to clarify — the request arrives ten minutes before Adaeze goes into a session and you have to start now. In that case, write your own assumed criteria and stopping point at the top of your notes before searching a single thing, and flag them in the handover: "I assumed you meant shows with a wellness or mental-health-adjacent audience, and stopped at thirty — let me know if I should widen it." An assumption stated plainly is a fast, defensible substitute for a conversation you didn't have time for. A silent one is a guess dressed up as a fact.
Every other lesson in this module — where to look, how to verify, how to present — only works once the target is this specific. Skipping this step and going straight to search is the single most common way research tasks in this job go wrong: not from bad searching, but from searching hard for the wrong thing.
Take a vague request you've received recently — from anyone, about anything. Write the research-question version: one sentence, plus criteria, plus a stopping point. If you can't write a stopping point, that's usually a sign the criteria aren't specific enough yet either.
With the brief from 6.1 written, the temptation is to open a search engine and start typing "mental health podcasts." That will work, eventually, but it's the slowest possible route — one search engine, one ranking algorithm, one kind of result. Good research fans out across a small number of genuinely different sources, because each one surfaces things the others miss.
For almost any research task in this role, three different kinds of source will find things a single search never will:
Lesson 6.1 gave three possible stopping points — a number, a time box, or saturation. Saturation is the one that takes judgement, so it's worth spelling out: keep a running log of every show you find, and once new searches start returning shows you've already logged, more often than turning up new ones, you have found most of what's genuinely findable through this method. Continuing past that point burns time for very little new material — better to note that you've hit saturation and stop there, even short of a target number, than to keep searching out of a feeling that you haven't tried hard enough.
Keep a running note of every search term you tried and every source you checked, even the ones that turned up nothing useful. This does two things: it stops you accidentally repeating the same search under a different phrasing, and it lets you tell Adaeze exactly where you looked — which matters more than it sounds like it should, because "I checked Apple Podcasts, Listen Notes, and the guest lists of four comparable therapists" is a very different, and far more trustworthy, sentence than "I googled around."
Not every source deserves equal weight. A directory that's actively maintained and lets you filter by real signals (recency, episode count) is worth more than a "top 50 podcasts" listicle from a marketing blog with no visible update date — those lists are frequently years stale and copied from each other, which is exactly the kind of thing lesson 6.3's verification step exists to catch. Treat listicles as a starting point for names to check, never as a finished answer.
Pick any topic and try all three source types from this lesson — a purpose-built directory, one adjacent example worked backwards, and the network around your first few finds. Notice which one surfaced things the others didn't.
A directory listing says a podcast has "50,000 monthly listeners." The listing is two years old. The show's last episode was eight months ago. Do you include it in the thirty? This is the question verification exists to answer — and the honest answer is often not yes or no, but a label: include it, clearly marked as uncertain.
The failure mode in research is not usually including something false. It's presenting something uncertain as if it were confirmed.
Rather than a binary in-or-out decision, sort every finding into one of three confidence levels. This is the single habit that makes research trustworthy rather than merely thorough:
For the podcast brief, verifying a single row means checking three things directly, not trusting the directory's word for any of them: does the show visibly take outside guests (an episode with a named guest, not just the host talking), is it still active (something published within the stopping point's recency criterion — two months, per the 6.1 brief), and is there a real way to pitch it (a contact email, a submission form, a booking platform listed) rather than just a general "contact us" with no clear route to guest pitches.
Directory says: "Wellness Weekly Naija — mental health podcast, guest interviews, large following."
Checked directly: Most recent episode was 11 months ago — fails the two-month activity criterion. The last five episodes before that were all solo commentary, no guests. Listed "guest interviews" appears to describe a format the show used two years ago, not its current one.
Verdict: Excluded, with a one-line note in the source log — "inactive and appears to have dropped the guest format; found via listicle, not confirmed independently." Worth keeping the note in case the show restarts later.
Notice that excluding it is itself a finding worth recording, not just a row deleted silently. If Adaeze later asks "what about Wellness Weekly, I've heard of that one," the log already has the answer.
Sometimes a show looks exactly right but you genuinely can't confirm activity or a guest format — the website is thin, there's no recent update anywhere. Don't quietly drop it, and don't quietly upgrade it to "likely" because you want it to be true. Include it labelled unverified, with the one specific thing that would confirm it: "unverified — promising audience fit, but no episode found in the last six months; worth a direct email to check if it's still active."
A strict rule — only include fully verified rows — sounds safer, but it silently throws away real leads that just happened to be harder to confirm in the time available. Labelling confidence instead keeps everything visible and lets Adaeze, who knows her own risk tolerance, decide how to use the "unverified" rows rather than never seeing them at all.
Take three claims from anywhere online — a stat, a fact, a listing — and actually check each one against a second, independent source. Label each verified, likely, or unverified based on what you find. Notice how often the honest label isn't "verified."
Thirty rows of research, properly verified and labelled, still fail if they land in Adaeze's inbox as a wall of text or a spreadsheet with fourteen unlabelled columns she has to decode before she can use any of it. The spreadsheet is not a storage format for what you found — it's the interface Adaeze will actually act through, so it needs to be designed for her, not for you.
A research spreadsheet succeeds when the client can sort, filter, and act on it without asking you what any column means.
More columns is not more useful — it's more to read before doing anything. For the podcast brief, six columns carry the whole job:
Set the sheet up sorted by confidence first, then by audience fit — so the thirty best bets are at the top, not scattered through the sheet in the order you happened to find them. Add basic filters on the confidence and audience-fit columns so Adaeze can narrow the list herself in ten seconds, rather than scrolling thirty rows every time she wants a subset.
Above the data, one short block: what the sheet covers, the criteria used to include or exclude a show, and the stopping point that was hit — the same three things from the brief in 6.1, now reported back. This is what turns a data dump into an accountable piece of work: anyone opening the sheet cold, months later, can see exactly what was searched for and why, without needing to ask you.
It's tempting to add a column for every fact you happened to collect — follower counts, episode length, host's other credentials. Unless a column changes what Adaeze does next, it's noise competing for attention with the columns that matter. If a fact is genuinely interesting but not decision-relevant, it belongs in the notes column, in one line, not its own column down the whole sheet.
Open any spreadsheet you've built or received recently. Count the columns. For each one, ask whether it would change what someone does next. Cross out the ones that wouldn't — that's usually two or three fewer columns than you started with.
The spreadsheet is done. Now comes the part most people skip: the message that goes with it. Send the sheet alone, with no summary, and Adaeze has to open thirty rows cold to figure out what she's looking at and whether it was worth the four hours. A five-line summary is what turns a finished spreadsheet into something she can act on in the ninety seconds she has between clients.
If you can't summarise a research task in five lines, you haven't finished thinking about what you found — you've only finished finding it.
Found 27 active, guest-accepting podcasts with an audience likely to include potential clients — mostly wellness, mental health, and Nigerian-diaspora-focused shows.
Checked three podcast directories, the guest lists of four comparable therapists, and the shows connected to those; counted a show if it takes outside guests and has posted an episode in the last two months.
I'd start with Mind & Motherhood and The Naija Wellness Hour — both actively feature therapists as guests and have a clear pitch route listed.
Eight of the 27 are labelled "likely" rather than confirmed — the shows look right but I couldn't verify current guest policy directly; worth a quick check before pitching those.
Sheet's attached, sorted by confidence. Happy to draft the first few pitch emails if useful.
The number isn't sacred — it's a forcing function. Aiming for five lines is what stops the summary sprawling into a second document that restates the whole sheet. If a genuinely important sixth thing needs saying, say it — but notice that the temptation is almost always to add detail that belongs in the sheet's notes column, not new information Adaeze actually needs to act.
The five lines belong directly in the email or chat message, visible without opening anything. The sheet is the reference; the summary is what gets read. Burying the summary inside the spreadsheet's own header defeats the purpose — it should be the first thing Adaeze sees, before she's decided whether the attachment is worth opening at all.
Take any piece of research or reading you've done recently — for this course, for anything. Write the five-line summary now, cold, without looking back at your notes. Whatever you can't remember without checking is a sign of what wasn't actually the headline.
You receive the exact request that opened this module: "Find me thirty podcasts I should be a guest on." Nothing else — no criteria, no deadline, no context beyond what you already know about Adaeze's practice from the last five modules.
Whether you can take an underspecified request and impose enough structure on it — a real question, real limits, real verification — that the result is something a client can trust and act on without redoing your work herself. Every module's drill has tested judgement under a different kind of mess. This one is the mess of not knowing, which is the hardest kind to manage honestly.
You should now have: a completed research brief, a real thirty-row sheet with honest confidence labels, and a five-line summary you'd actually be willing to send. Module 7 turns to working with AI tools as a junior on your team — what to hand them, what must never leave a client's account, and how to check work you didn't do yourself before it goes out under your name.