Finding opportunities: interviews & JTBD

Maya has an outcome. After the work in module one, the trio’s tree has a real root: increase the share of workspaces that send a follow-up within 24 hours of a meeting. What it doesn’t have is anything underneath. The opportunity space — the layer of customer needs, pains, and desires that connects the metric to actual humans — is three placeholder sticky notes and a lot of white wall.

So Maya books her first interview: a solo consultant, six weeks into her trial, moderately active. Maya opens with “Would you use a feature that reminded you to follow up after meetings?” The consultant says “Oh, definitely, that sounds useful.” Maya asks whether she’d want the reminder as an email or in-app. “Either, honestly. Both?” Twenty-five minutes of this. Maya leaves with a page of notes, and every line of it is the customer being nice to her.

Three days later Ravi forwards the cancellation notice. Same consultant. The polite yeses evaporated the moment they had to survive contact with her actual week. Nothing in Maya’s notes predicted it, because nothing in Maya’s notes was about the consultant’s actual week — it was all about Maya’s imagined feature, reflected back at her with good manners.

This module is about the discipline that fixes that interview, the artifact you make from each one, and how a pile of interviews becomes a structured opportunity space. (Course landing page: Continuous Discovery, Hands-On.)

Why “would you use X?” always gets a yes

The question fails for a structural reason, not because customers lie. You’re asking someone to predict their future behavior, and humans are terrible at that — and socially motivated to be optimistic about it in front of the person who built the thing. “Would you use it?” costs the customer nothing to affirm and tells you nothing you can build on. Rob Fitzpatrick’s Mom Test framing: even your mom would say yes to that question, which is why it’s worthless.

The fix is to change tense. Stop asking about hypothetical futures and start asking about specific pasts. Not “do you usually follow up after meetings?” — usually invites a self-flattering average — but “tell me about the last meeting you had with a client. Walk me through what happened after you hung up.” A specific recent instance can’t be polite. It happened one particular way, with real friction in real places, and your job is to walk through it slowly enough to see where.

Teresa Torres calls these story-based interviews, and they’re the weekly heartbeat of continuous discovery. One good story, excavated properly, beats ten opinions about your roadmap.

Anatomy of a story-based interview

Maya’s second interview goes differently. The subject is a sales rep at a nine-person agency — one of the “half my follow-ups slip because I write them on sticky notes after calls” people Ravi flagged from the support inbox. Here’s an excerpt, with commentary on what Maya is doing right this time:

Maya: Tell me about the last client call you had. Not a typical one — literally the most recent.

Anchors in a specific instance immediately. “Typical” would get a rehearsed summary; “most recent” gets what actually happened.

Rep: Tuesday, a discovery call with a logistics company. Went well, actually.

Maya: Walk me through the ten minutes right after it ended.

Doesn’t ask “what did you do next?” in the abstract — narrows time down to where she suspects the pain lives, and asks for a walkthrough, not a verdict.

Rep: Honestly? I had another call at the top of the hour. So I grabbed a sticky note, wrote “logistics — send pricing + intro to Sam,” stuck it on the monitor, and jumped on the next one.

Maya: What happened to that sticky note?

Follows the artifact, not the intention. The sticky note is where the story gets honest.

Rep: [laughs] It’s probably still there. I sent the pricing Thursday, I think. Never did the intro.

Maya: You said the call went well — how did it feel, realizing Thursday that it had sat for two days?

Mirrors his own words back and asks about the feeling. Emotion marks intensity, and intensity is data she’ll need later when sizing opportunities.

Rep: Bad. That’s the game, right? They talk to three agencies, and the one who follows up same-day looks like the one who wants it.

Maya: Has a deal ever actually gone sideways because of a slow follow-up?

Asks for consequence, again as a past instance. “Ever actually” quietly filters out theory.

Rep: One I know of for sure, this spring. Champion told me straight up they went with the firm that “had a proposal in our inbox before dinner.”

Notice what Maya never does: mention Donut CRM’s features, float an idea, or ask him to imagine anything. The entire interview lives in his world, in the past tense. And notice what she got — a struggling moment (post-call crunch, back-to-back meetings), a coping mechanism (sticky notes), a failure mode (the intro that never happened), an emotional charge (that lost deal still stings). Four things a “would you use X?” interview can never produce.

One craft note before you try it: the hardest part isn’t asking the first question, it’s shutting up after it. Silence after “walk me through it” is where the real story arrives. Fill the pause and you’ll get the summary instead.

Your turn: rewrite your questions

Pull up notes from your last three customer conversations and find three questions you asked that were hypothetical, general, or leading — anything shaped like “would you,” “do you usually,” or “how important is.” Rewrite each as a story question anchored to a specific recent instance. The template: “Tell me about the last time you [did the activity]. Walk me through it.” Fifteen minutes. Keep the rewrites — you’ll use them in module three when the interviews start happening weekly.

Your turn: pick the better question

Instinct matters more than theory here, and instinct is trainable. For each pair below, pick the question that would get you a real story instead of a polite opinion.

Interview snapshots: one page, or it didn’t happen

An interview that lives only in your notebook helps exactly one person, briefly. Torres’s answer is the interview snapshot: a one-page artifact you produce within a day of each interview, while the story is still warm. It has four parts — quick facts about who this was, the one quote you’d put on the wall, a rough sketch of their experience as a sequence of moments, and the opportunities you heard, in their words.

Here’s the snapshot Maya writes for the interview above:

# Interview snapshot — 2026-08-19, "Agency rep, sticky notes"
 
## Quick facts
- Sales rep, 9-person creative agency, ~10 client calls/week
- Donut CRM workspace: active 3 months, uses timeline + tasks,
  has never opened meeting recaps
- Recruited via Ravi (support inbox: asked for "call notes export")
 
## Memorable quote
> "The one who follows up same-day looks like the one who wants it."
 
## Experience map (post-call, back-to-back day)
call ends → 3-min gap → sticky note on monitor → next call →
[repeat × 4] → 6pm: tired, triages stickies → some become emails,
some become "Thursday, I think," some become nothing
 
## Opportunities heard
- "I can't capture follow-ups properly when calls are back-to-back."
- "Commitments I make on calls disappear if I don't write them
  down immediately."
- "I want the client to feel like the follow-up came fast."
- (weak signal) Doesn't trust recaps he's never read — didn't
  know the feature existed. Onboarding gap, park for now.

Three rules make snapshots work. Opportunities go down in the customer’s words, not yours — “I can’t capture follow-ups when calls are back-to-back,” not “needs automated task capture,” which has already smuggled a solution into the evidence. One page, enforced — the constraint is what makes you decide what mattered, and what makes teammates actually read it. And every snapshot is shareable by default: Tom and Priya read this one in ninety seconds, which is the whole point of the trio doing discovery together instead of the PM narrating it to them secondhand.

The experience map deserves a special mention. Sketching the customer’s moments as a sequence — even this crudely — does something a quote list can’t: it shows you where in time the opportunities cluster. Maya’s map makes it obvious that everything interesting happens in that three-minute gap and at the exhausted 6pm triage. That’s two different struggling moments, and they’ll become two different branches of the tree.

Your turn: snapshot your last conversation

Build a snapshot template with those four sections in whatever tool your team already reads (doc, wiki page, Notion — it doesn’t matter). Then fill it in from your most recent customer conversation, even if that conversation was weeks ago and the memory is fuzzy — a degraded snapshot of a real interview still beats a blank template. Twenty minutes. Note where you can’t fill a section: that’s the shape of what your current interviewing style fails to collect.

The JTBD lens: what keeps opportunities honest

Somewhere around the fifth snapshot, a failure mode creeps in: opportunities start drifting toward the product. “Needs better recap notifications.” “Wants the task composer on mobile.” These aren’t needs, they’re feature requests wearing a trench coat, and a tree built from them just laundered your backlog through some interviews.

Jobs to Be Done is the lens that keeps the opportunity space solution-agnostic. Not as a rival process — you don’t choose between JTBD and continuous discovery, and anyone selling you that fight is selling workshop tickets. JTBD is a complementary framing: it gives you sharper language for the struggling moment a customer is in, and for the forces that govern whether they’ll actually change their behavior to escape it. Continuous discovery gives you the cadence and the tree; JTBD sharpens what goes on the sticky notes.

The most useful JTBD instrument for Maya right now is the four forces of progress, and the best interview to apply it to is a switch interview — someone who changed behavior, in either direction. Donut CRM has a painfully good candidate: the churned user who went back to a spreadsheet plus calendar reminders and told Ravi, “your tool asked me for more than it gave back.” Maya gets forty-five minutes with him and analyzes it as a switch:

• Push (what shoved him away from the old way): he’d lost a renewal because a promised follow-up never happened; the spreadsheet had no memory of commitments.
• Pull (what attracted him to Donut): the timeline — “everything about a person in one place” was the demo moment that sold him.
• Anxiety (what made the new way feel risky): every meeting demanded logging, tagging, categorizing before it gave anything back. “I was doing data entry for a robot that never returned the favor.” The anxiety wasn’t about switching in — it accumulated after, until the tool felt like a second job.
• Habit (the gravity of the old way): fifteen years of spreadsheet muscle memory, and calendar reminders that — crucially — fired on their own. Zero-effort nagging beat rich-context silence.

Read those four forces together and the finding is uncomfortable and precise: Donut lost this user not because it lacked features but because its cost-before-value curve was inverted — it demanded work up front and paid off later, while the crude old system paid off immediately. That’s an opportunity (“I put effort in before I get anything out”), it’s stated in the customer’s world, and no feature request in the backlog would have surfaced it.

That’s the test for whether JTBD is helping you: it should make your opportunities less like your product, not more. If a jobs workshop ends with the same feature list you walked in with, it was a séance, not a lens.

Structuring the opportunity space

Eight interviews in, Maya has a wall of opportunity sticky notes and the new problem: a flat pile of forty quotes is barely more usable than no quotes. The opportunity space in an opportunity solution tree isn’t a list — it’s a hierarchy. Parent opportunities are broad needs; child opportunities are specific, addressable versions of them. Two rules govern the shape:

1. A child must be a genuine instance of its parent. If solving the child wouldn’t chip at the parent, it’s filed wrong.
2. Siblings must be distinct. If two children under the same parent would be solved by the same solution, they’re one opportunity written twice. Merge them.

The sizing question — which branches are big — isn’t answered by vote or vibe but by frequency and intensity of evidence. Frequency: how many distinct interviews produced this opportunity? Intensity: when it came up, how charged was it — a shrug, or a lost deal someone still resents? A pain heard once with fury usually deserves less weight than one heard seven times with mild irritation, but both dimensions go on the tag.

In module one the trio drafted a v0 opportunity space from what they already believed. Now, with eight snapshots behind it, it gets restructured on evidence. Here’s the two-level space, with evidence counts (interviews out of eight) on every node:

OUTCOME: Increase the share of workspaces that send a
follow-up within 24 hours of a meeting
 
├── I can't capture what I committed to during the meeting  [6/8]
│   ├── Follow-ups slip when calls are back-to-back —
│   │   sticky notes, memory, nothing  [5/8]
│   └── I can't remember what we talked about last time,
│       so I can't tell what I still owe  [4/8]
│
├── Writing the follow-up itself takes too long, so I
│   defer it (and deferred = dead)  [4/8]
│   ├── Reconstructing the meeting to draft the email is
│   │   the slow part  [3/8]
│   └── I don't trust the CRM's data enough to send
│       anything it wrote for me  [2/8, high intensity]
│
├── The tool asks for effort before it gives value back  [2/8,
│   high intensity — includes the churn switch interview]
│
└── I show up to meetings unprepared, so the meeting itself
    creates vaguer commitments  ("honestly, I winged it")  [3/8]
    └── (children TBD — thin evidence, needs targeted interviews)

A few things to notice about how this differs from a brainstorm. Every node is in customer language, in the customer’s world — even “I don’t trust the CRM’s data,” which is one interview away from being written as “improve AI accuracy” and thereby ruined. The capture branch and the writing branch stayed siblings because they’re genuinely distinct: one is about remembering the commitment, the other about the cost of honoring it, and no single solution kills both. The trust node carries a low count but a high-intensity tag, so it stays visible instead of being averaged away. And the prep branch is honest about being thin — a marked gap in the tree is a recruiting instruction for module three, not an embarrassment.

Solo vs. as a trio. Alone, mapping the space is an affinity sort: every opportunity from every snapshot on its own card, cluster the cards by the need underneath them, name each cluster in customer words, then count evidence per cluster. The discipline that replaces having colleagues is a forced second pass: for every card, ask “is this really an instance of this parent, or did I file it here because I wanted this branch to look big?” As a trio, run it as a silent sort first, discussion second: everyone groups cards without talking for ten minutes, then you argue about where the clusters disagree — the disagreements are the interesting part, and talking first lets the loudest early opinion anchor the whole wall. For bigger groups, 1-2-4-All is the pattern that keeps eight people from becoming one HiPPO with an audience; the full Liberating Structures toolkit gets its treatment in module four.

Your turn: draft your opportunity space

Take every customer-stated pain, need, and desire you currently have evidence for — snapshots if you’ve started them, support tickets and call notes if you haven’t — and draft a two-level opportunity space under the outcome you wrote in module one. Tag each node with an evidence count, even if the honest count is 1, and mark the thin branches explicitly. Then add the top-level opportunities to the tree you started building in module one, so the artifact grows with the course. Thirty minutes. Resist adding any node you merely believe in: a small evidenced tree is worth more than a big imagined one.

What you should have now

There’s a dependency hiding in that list: everything above assumes a steady supply of interviews, and eight conversations scraped together from Ravi’s inbox is a stockpile, not a supply. Making interviews happen every single week — recruiting, outreach that gets replies, and what to do when nobody shows up — is module three.

Further reading

• Continuous Discovery Habits — Teresa Torres — chapters on interviewing, snapshots, and mapping the opportunity space are the backbone of this module.
• The Mom Test — Rob Fitzpatrick — the short, blunt book on why “would you use X?” fails and what to ask instead.
• Demand-Side Sales 101 — Bob Moesta — the four forces and switch interviews, from the person who runs them best.