The Workshop: Jobs to be Done

Customers don’t buy products – they hire them to do a job. JTBD is the interview technique that surfaces the actual job and the alternatives they’d defect to. Switch interviews – structured interviews with people who recently switched products or services, asking what triggered the move – are the core mechanic. Why Subscribers Actually Stay is the worked example; this post is the playbook.

Jobs to be Done

Jobs to be Done runs switch interviews with recent customers, pulls the four forces (push from the old situation, pull of the new option, anxiety about switching, habit holding the customer in place) out of what they say, and shapes 3–5 candidate job statements in the form “When [situation], I want to [motivation], so I can [outcome]” – in a single half-day session, so the team has a shared and specific view of what customers are actually hiring the product to do.

Sometimes called JTBD, job mapping, or outcome-driven innovation, though outcome-driven innovation is a distinct quantitative framework (Ulwick) that layers over the qualitative interviewing. Sometimes confused with user personas – personas describe who a user is; jobs describe what they’re trying to get done, which is a different axis and a more useful one for product decisions. The switch-interview technique itself comes from Bob Moesta and the Re-Wired Group; the four-forces framing comes from Moesta and Chris Spiek; the broader theory from Clayton Christensen.

What’s It For

A team builds a feature someone asked for. The feature lands, the telemetry looks fine for a week, and then the customer who asked for it cancels. Nobody connects the cancellation to the feature – it was a quarter ago, a different person, a different conversation – but the pattern repeats quietly over the year. The backlog fills with requests. The product changes shape. Churn doesn’t move.

The problem is that asking a customer what they want produces a list of features. The list is honest and useless. Customers describe solutions they can imagine because describing causes they’re half-aware of is hard. The Jobs to be Done school of thinking – Moesta, Christensen, Ulwick – reframes the interview: don’t ask what they want. Ask what happened the day they switched. What prompted it. What they were trying to get done. What they’d been doing before. What would have made them stay with the old thing.

Switch interviews replace the feature wishlist with a story about a decision. The story contains the job. The job is usually not the one the team expected.

This workshop exists to collapse that reframing into a single session: three or four interviews, silent discovery, a clustering round, and a set of candidate job statements the whole team watched emerge. The statements are the artefact. The shared view is the point.

Reach for it when:

Churn is stable but not improving and you’ve exhausted surface-level theories
A new feature area is being considered and the team can’t agree on the problem it solves
You have access to recent switchers – people who started or stopped using the product in the last ninety days
Personas are in use and clearly not driving decisions
Several teams are prioritising against different implicit jobs and colliding

What It’s Not For

Skip it when:

You don’t have switchers to talk to. The technique is switch interviewing; discovery without interviews is just speculation in a conference room.
The decision you’re trying to make is tactical. JTBD is a framing exercise, not a backlog refinement tool.
The team believes they already know the job and you’re being asked to validate it. Confirmation-seeking kills the interviews.
You can’t get 2–3 hours of focus out of the product lead. The discovery cannot be delegated.

Stop a session that’s already started if:

The interviewees can’t remember why they switched – they’re not recent enough
The sticky-note wall is mostly empty after thirty minutes – the interviews didn’t land
The room is arguing about whether switch interviews are valid – you have a trust problem, not a method problem

Stopping after the first interview to regroup on technique is not failure. Running three mediocre interviews and producing confident statements from them is.

Definitions & Background

Three dimensions of a job. Every job has three layers, and the richest material lives in the second and third:

Functional – the practical thing being done. “Plan the week’s meals.”
Social – how the customer wants to be seen while doing it. “Be the parent who feeds the family well.”
Emotional – how they want to feel, or stop feeling. “Stop having to think about dinner on Sunday.”

Most teams capture only the functional layer and miss why the job actually matters. Push every candidate job statement to expose all three.

The switch timeline. Moesta’s interview structure has five anchor moments along the customer’s path to switching. Knowing the names lets the interviewer ask for each one explicitly:

First Thought – the moment the customer first considered making any change. Usually weeks or months before the switch. Push lives here.
Passive Looking – low-effort browsing, not actively shopping yet.
Active Looking – shortlisting, comparing, asking around.
Deciding – choosing between candidates. Anxiety spikes here.
Consuming – using the new thing for the first time. Habit sets in or it doesn’t.

The four forces (push from the old situation, pull of the new, anxiety about switching, habit holding the customer in place) map onto these moments rather than being asked about abstractly.

Inputs

Three to six recent switchers scheduled for 45-minute calls (ideally two recent customers who started, one recent canceller, or the equivalent for your switch).
A rough interview guide – we’ll give one below – but not a script.
The team to have read one or two switch interview transcripts before the session so they recognise the shape.
Recording setup (with permission) and a shared document for verbatim notes.
Sticky notes and a wall for the discovery and clustering phases.

If you don’t yet know which customers to talk to or what switch you’re trying to understand, run Event Storming a Domain first to map the customer landscape, or pull a churn list from your CS team to seed the recruit.

Outputs

What lands at the end:

3–5 candidate job statements in the form “When [situation], I want to [motivation], so I can [outcome].” These are the headline artefact.
A wall of verbatim quotes, clustered, with each cluster named.
Interview transcripts filed somewhere the team can read them for months. Redact names and any personal detail not relevant to the job.
Tags against the four forces – which quotes show push, which show pull, which show anxiety, which show habit. The tags are the evidence behind each job statement.
A “not our job” list, sometimes – the requests you can now deliberately decline.

These outputs feed straight into:

Impact Mapping – Impact Mapping tells you what behaviour to change; JTBD tells you what job the customer is hiring you for. Run JTBD first when you don’t know the job yet; run Impact Mapping first when the job is clear but the behaviour change isn’t.
User Story Mapping – once the jobs are named, User Story Mapping lays out the journey through them and slices the backlog against each job.
Example Mapping – Example Mapping turns a story into concrete rules; JTBD turns a customer conversation into a story worth writing. They compose at opposite ends of the refinement pipeline.

Who’s Needed

Four to six team members for roughly 3h 45min (with two interviews) to 4h 30min (with three):

Facilitator. Runs the session, conducts or co-conducts the interviews, keeps the discovery honest. Someone who has done switch interviews before is a large advantage. If nobody in the room has, schedule a practice interview with a friendly customer the week before.
Interviewers. Two or three, rotating. One person asks, another listens and notes, they swap between interviews. The rotation matters – it stops any single interviewer’s theory hardening into the session’s finding.
Note-taker. Often the facilitator doubles here, but if the interviews are back-to-back, split the role. The note-taker captures verbatim quotes, not paraphrases. Paraphrase is where the team’s existing theory sneaks in.
Product lead. Mandatory. The job statements will reshape the roadmap, and the product lead needs to have been in the room when they came out. If they arrive only for the readout, the statements will land as someone else’s conclusions, and they will be argued rather than used.
Optional ops / CS observer. Someone who talks to customers every day. Their job is to contradict the neat story that emerges from three interviews with the people who picked up the phone. They know the customers who didn’t, and that context keeps the discovery honest.

Group size is 4–6 team members (interviewees are not counted). Below four and the clustering lacks the friction it needs; above six and the silent discovery phase becomes committee writing.

Who to leave out:

Large groups of stakeholders. This is not a readout session. Discovery with more than six voices collapses into consensus-seeking.
People who can’t let go of existing features. If someone is going to defend the current roadmap sentence-by-sentence during clustering, they will prevent the session from doing its job. Invite them to the readout afterwards.
Anyone who won’t suspend their theory for three hours. JTBD interviews are deliberately theory-free. Bring a theory into the listening and you’ll hear confirmation.

How To Run It

Phase	Duration	Materials	Key question
Brief and prep	15 min	Interview guide, recording setup	“What are we listening for?”
Switch interview 1	45 min	Phone / call, notes	“Tell me about the day you switched.”
Switch interview 2	45 min	Phone / call, notes	“Tell me about the day you switched.”
Switch interview 3 (optional)	45 min	Phone / call, notes	“Tell me about the day you switched.”
Silent discovery	30 min	Sticky notes, quotes	“What did we actually hear?”
Cluster into candidate jobs	30 min	Clustered quotes	“What story do these clusters tell?”
Name 3–5 candidate jobs	20 min	Job statement cards	“When… I want to… so I can…”
Wrap-up	10 min	—	“Who owns what next?”
Total	~3h 45min with two interviews, ~4h 30min with three

The interviews drive the day. Everything else is in service of extracting the job from what the switchers said. If the interviews don’t happen – schedule slips, no-shows, technical failures – postpone the discovery. Don’t fake it with remembered quotes.

Listening and discovery

Two distinct modes, and the discipline of keeping them separate is most of the technique:

Listening mode. During the interviews. Open questions, long silences, “and then what happened?” nudges. No theorising, no reframing the question, no rescuing the interviewee when they stall. The pauses are where the good material comes out.
Discovery mode. After the interviews. Quotes on sticky notes, clustered by pattern, named at the end. Silent individual work first; discussion only after the clusters are visible.

The four forces come out in the second mode, not the first. Don’t ask interviewees about push, pull, anxiety, and habit – listen for them.

Push of the current situation. What was annoying, broken, or insufficient about what they were doing before.
Pull of the new solution. What drew them toward the new thing. What it promised.
Anxiety about the change. What made them hesitate. What they were afraid would go wrong.
Habit of the present. What made it easier to keep doing what they were already doing, even when it wasn’t working.

A real switch story contains all four. The four forces are a lens for reading the transcript, not a question to ask.

“Pulling them out” is the mechanic of that lens. After the interview, go back through the transcript and tag the lines that show each force. To make this concrete, here’s how it might land in a switch interview about a meal-box subscription: “The supermarket veg kept going off before we’d eaten it” is push, “My neighbour’s box looked amazing on Instagram” is pull, “What if we get things we don’t know how to cook?” is anxiety, and “We’d done the same Saturday shop for years” is habit. The interviewee never labelled any of them – they told a story, and the team tagged it afterwards. Those tags are the evidence behind the job statements you write later: when the situation clause reads “When the weekly shop has stopped working…” you can point at the push quote it came from.

Forces you can’t tag matter too. Strong push and weak anxiety is a switcher who was already on the way out. Strong habit and weak pull is a switcher who needs a bigger nudge than the product is currently offering. The tagging is what turns three interview stories into a map of the decision shape, not just three transcripts in a folder.

Phase 1 – Brief and prep (15 min)

Gather the team. Walk through three things, briefly:

“We’re going to run switch interviews. The shape is: tell me about the day you switched, walk me backwards to when you first started thinking about it, and tell me what else you considered. We’re listening for what was going on in their life when they made the change – not for feature feedback. We’re not going to ask them what they want. We’re going to ask them what happened.”

Second: the four forces, one line each. Tell the team to keep the four forces in the back of their heads, not the front. The interview is not a forces-extraction machine.

Third: the roles for the first interview. Who asks, who takes notes, who observes in silence. Set the expectation that roles rotate.

What to watch for:

Pre-loading theories. “I bet they’re going to say it’s about convenience.” Name it and park it: “Let’s see what they say.”
Prepared questions. Interviewers who’ve written a list of fifteen things they want to ask will interrupt the story. The guide below is three prompts, not fifteen.
Recording permission missed. If you’re recording, confirm permission explicitly at the top of the call. If you can’t record, double the note-taking.

The interview guide:

“Take me back to the day you decided to switch. What was happening that day?”
“When did you first start thinking about it? What else were you considering?”
“Was there anything that almost stopped you? What made you go ahead anyway?”

Everything else is follow-up prompts: “tell me more about that,” “what happened next,” “who else was involved,” “how did that feel.”

Phase 2 – Switch interviews (45 min each)

Run the interview by phone or video. Camera on if the interviewee’s comfortable, off if not. The note-taker captures verbatim quotes in a shared document, with timestamps if the call is recorded.

Ask the first question and then wait. The interviewee will start. Don’t fill silences. If they stop after thirty seconds, prompt with “and then what?”

Push for the concrete scene. “What day of the week was it? Where were you when you first thought about it? Who did you talk to?” Abstractions hide jobs; specifics reveal them.

Walk them backwards along the timeline. When they’ve finished the story of the day itself, walk back: when they first thought about it, what they were doing before, what triggered the first thought. The “first thought” is often weeks or months before the switch, and that’s where the push usually lives.

When they’re done with the timeline, ask about alternatives. “What else did you consider? Why did you pick this one? What would have made you stay with what you had before?”

What to watch for:

Generic answers. “It was just more convenient.” Push: “Convenient how? Give me the specific thing that annoyed you last time you did it the old way.” A generic answer is an unearned abstraction; the story is always underneath.
Rationalised stories. The interviewee has told themselves a tidy narrative about why they switched. You’ll hear marketing language in their mouth. Rewind: “Before you decided that, what were you actually doing?” Walk to the concrete scene.
Interviewers filling silences. Note-takers should kick the interviewer under the table. Thirty seconds of silence almost always produces the best quote of the interview.
The team diagnosing during the call. “Oh – they want a pause feature.” No. Listening only. Diagnosis is the next phase.

End the interview cleanly. Thank them. Don’t summarise back to them – summaries bias the memory of what they said.

Phase 3 – Silent discovery (30 min)

Print or project the transcripts. Each team member works alone. The instruction is one sentence:

“Write every verbatim quote that feels telling onto a sticky note. One quote per note. No interpretation.”

Telling means: reveals a push, a pull, an anxiety, a habit, a moment of decision, a named alternative, an outcome they were trying to achieve. If the quote is about a feature they wanted, it’s probably not telling – that’s a solution, not a job.

Silent, individual, no discussion. Set a timer. When the timer ends, everyone posts their notes on the wall without comment.

What to watch for:

Paraphrase creep. Someone writes “customer wants convenience” on a note. That’s paraphrase. Push back: “What did they actually say? Use their words.”
Feature requests mistaken for jobs. “They said they want a weekly summary.” That’s a solution. The question underneath is what the weekly summary is being hired to do.
One team member producing twice as many notes as anyone else. Good. Don’t suppress it. The clustering will balance.

Phase 4 – Cluster into candidate jobs (30 min)

Look at the wall. Ask the room to cluster notes that belong together. No talking for the first five minutes – the affinity-map convention – silently group sticky notes by similarity, then name the clusters – people move notes silently, and if two people keep moving the same note back and forth, it’s flagged for discussion.

After five silent minutes, open the conversation. For each cluster, ask two questions:

“What’s the pattern here? What are these quotes all saying?”

“Is this a situation, a motivation, or an outcome?”

Those three words – situation, motivation, outcome – are the JTBD shape. A cluster might be all situations (things that were going on in customers’ lives), all motivations (what they were trying to get done), or all outcomes (what they wanted to be true afterwards). Often a single cluster contains one of each and is the seed of a job statement.

Expect 4–7 clusters from three interviews. Fewer than four and you’ve over-abstracted; more than seven and you haven’t clustered enough.

What to watch for:

Everything-is-one-cluster. The room collapses the wall into two huge piles. Push for distinctions: “What’s different about these two quotes?”
The wishlist cluster. A cluster forms around feature requests. Re-frame: “If we built all of these, what would customers be able to do that they can’t do now?” The answer is usually the job.
Forgotten negative space. Quotes about anxiety and habit rarely cluster on their own unless you prompt for them. “Which of these clusters contains ‘what almost stopped them’? Is that cluster complete?”

Phase 5 – Name 3–5 candidate jobs (20 min)

Pick the 3–5 strongest clusters and turn each into a job statement. The form is strict:

“When [situation], I want to [motivation], so I can [outcome].”

This is Klement’s job story form (Alan Klement, who refined the JTBD interview practice into the modern job-story shape), which bakes the situation in. Christensen’s classical job statement form is shorter – “Help me [verb] [object] [modifier]” – and useful for headline framing. We use Klement’s form here because the situation is what the four-forces evidence directly supports.

Each slot is a concrete phrase, not a category.

Situation: the context the customer is in when the job arises. Time, place, people, constraints.
Motivation: the action they want to take. A verb and an object.
Outcome: the state of the world they want to be true as a result. What they get to do next, how they want to feel, what they no longer have to worry about.

Write each statement on a card the whole room can see. Read it aloud. Challenge it against the quotes – does any sentence in the transcripts contradict the statement? If yes, adjust.

What to watch for:

Abstract outcomes. “So I can be happy.” Push for what happy looks like. “So I can stop having to think about dinner on Sunday.” That’s a specific outcome.
Product names in the statement. “So I can use our app.” No. That’s a solution. “So I can plan the week without a grocery trip.” That’s the job.
Two jobs in one statement. “When it’s busy, I want to plan the week and also try new recipes, so I can feed the family without stress.” Split into two statements. One job per card.
Committee wording. The room rewrites the same statement four times. Park it. Accept the rough version and move on; polish later.

Phase 6 – Wrap-up (10 min)

Pin the 3–5 job statement cards on the wall. Photograph them. Read each aloud one more time with the team. Then name the owners:

“Product lead – these are yours from here. Ops observer – you’re running the sanity check against what you hear on calls next week. Engineering lead – I’ll walk these past you Monday.”

End on commitments, not summaries.

Worked example

See Jobs to be Done: Why Subscribers Actually Stay for a fictional team’s first switch-interview session – including the moment three interviewees independently describe the same Sunday-night job the team had never heard named. The product in that story is a meal-box subscription, but the shape of the session is the same in any domain.

What Can Go Wrong

The feature wishlist. The interview turns into a list of features the customer wants. Recovery: Interrupt gently: “Let me take a step back – what were you doing before you switched? Walk me through that week.” Pull them back to the timeline. Stop if: The same interviewee keeps returning to features despite three rewinds. Thank them, end the call, and try a different interviewee.

The generic answer. Everything is “convenience” or “quality” or “the vibe.” Recovery: “When you say convenience, what did the morning of your Monday look like before, versus after?” Specifics always. Stop if: They genuinely can’t recall. They’re probably not a recent switcher – check when they actually switched.

The rationalised story. The interviewee has a clean narrative that sounds like your own marketing. Recovery: Walk to the concrete scene. “Before you decided that, what were you actually doing on a Tuesday night?” Stop if: They resist the concrete scene. Rationalisation is often protective – don’t force it.

Jobs conflated with solutions. During discovery, someone keeps writing job statements that include the product. Recovery: Delete the product name and see if the statement still holds. If it doesn’t, it’s not a job – it’s a feature brief. Stop if: The whole wall is solution-shaped. The interviews didn’t produce enough material; schedule more.

The wording committee. Four people argue about the wording of a single statement for twenty minutes. Recovery: Force a rough version. “Worst acceptable version. We’ll polish next week.” Stop if: The argument is actually about whether the job is real. Go back to the quotes and check.

Confirmation bias. The room is finding what it already believed. Recovery: Ask the ops / CS observer to challenge every statement against the customers they talk to. “Would anyone you speak to on the phone recognise themselves in this?” Stop if: Two observers independently say the statements don’t match what they hear. The interviews may be unrepresentative – schedule different interviewees.

Other failure modes worth naming so you can spot them early:

The team treats three interviews as definitive and skips follow-up
The statements get written and shelved; the roadmap continues as before
Interviewers slide into persuasion mode and start explaining the product to the interviewee
Discovery collapses into consensus around the theory the product lead walked in with
Quotes get paraphrased into notes and the verbatim material is lost

The session has costs as well as benefits, and naming them helps the team commit honestly: 4–5 hours of session time plus 3–4 hours of interview scheduling and coordination, the emotional cost of hearing customers describe problems you haven’t solved, and the fact that the candidate job statements are candidates – they want validation with more interviews before they drive anything irreversible. Interviewer skill compounds; early sessions produce rougher material.

Next Steps

The session ends; the work begins.

Same day, the facilitator:

Photographs of the wall: the full cluster layout, each candidate job card in close-up, the verbatim quote notes against their clusters.
Files the interview transcripts somewhere the team can read them for months. Redact names and any personal detail not relevant to the job.
Writes a short summary: the 3–5 candidate jobs, one sentence each about the strongest quote behind each, and the four forces where they showed up.

This week, the product lead:

This is where the pattern earns its cost, and the work is mostly the product lead’s.

Walk the candidate jobs past the ops / CS team. People who talk to customers every day will either nod or wince. Both reactions are useful. The wince is more useful.
Schedule three more interviews to validate the strongest candidate. Three interviews is not enough to commit; three more either strengthen the statement or reveal the hole. Treat the first session’s output as a hypothesis.
Map the current roadmap against the jobs. Which features serve a named job? Which don’t? A feature that doesn’t serve any job is either a job you haven’t articulated yet or work that shouldn’t be in the quarter.
Refuse the next feature request that doesn’t match a job. Politely. With a reason. This is the hardest week-after task and the one that makes JTBD earn its keep.

Ongoing:

Re-run JTBD when the product shape changes materially – a new segment, a new pricing tier, a new acquisition channel. The jobs change when the customers change.
Keep the job statements visible. Pin them in the team’s main room. When someone proposes a feature, they should be able to point at the job it serves.
Track the feature requests that don’t match any job. If the list grows, you’re either missing a job or missing the discipline to say no.

The benefits compound: job statements concrete enough to make product decisions (features that serve a named job get built; features that don’t get parked), a shared framing across product, engineering, and operations that reduces backlog churn, interview transcripts that earn their keep for months as future hires read them and onboard faster, a “not our job” list that is just as valuable as the jobs themselves, and the push / pull / anxiety / habit lens available for every future product conversation.

Variants

The default switch-interview shape captures one direction: people who chose the product. Two adjacent shapes capture the jobs you’re missing – customers who left, and prospects who never started.

Switch-out interviews (churn). Run the same playbook with people who cancelled in the last ninety days. The prompts adapt:

“Take me back to the day you decided to cancel. What was happening that week?”
“When did you first start thinking about it? What pushed you over the edge?”
“What are you doing now instead? Did you switch to something else, or go back to what you had before?”

The four forces re-orient. Push is what your product was doing wrong – the feature that broke, the support reply that landed badly, the price increase that finally tipped them over. Pull is the destination, which is often nothing: the cancelled customer went back to the way they did it before, not to a competitor. That’s a stronger signal than competitive churn – it means the job you thought you were doing wasn’t being done well enough to displace the old way at all. Anxiety is what made cancelling hard – the workflow they’d built around the product, the data or history they’d lose, the loyalty discount they’d give up. Habit is the inertia that kept them paying past the point of value: how many months did the bill go out after they’d stopped really using it?

A churn interview where the canceller went back to the way they did it before is the most useful kind you can run. It tells you the job you wrote down isn’t real, or isn’t being delivered. The team won’t want to hear it; the temptation will be to dismiss the canceller as not the target. Resist.

Non-adoption interviews. People who looked at the product and didn’t sign up, or fit the audience and never engaged. Harder to recruit – you don’t have their email – but the most valuable shape when growth has stalled and churn doesn’t explain the shortfall.

The prompts shift, because there’s no “day they switched”:

“Tell me about the last time you thought about a product like ours. What was happening?”
“What did you end up doing instead?”
“What stopped you from trying it?”

The forces re-orient again, and the missing forces are the finding. Push is what’s not working about whatever they’re using today; usually it’s weak, because the existing alternative is an adequate solution for adequate people. Pull is what your product promised them; usually it’s weak too, because if pull had been strong they’d have signed up. Anxiety is what stopped them – what if it doesn’t fit how they actually work, what if they can’t get value out of it, what if it locks them in. Habit is the strongest force in this set: most non-adopters are well served by what they’ve used for years, and the real question is whether anything could ever move them.

Recruit non-adopters by:

Asking churned customers to introduce contacts who also considered the product but didn’t sign up
Running a short paid screener through a research panel
Offering a small incentive through channels where the audience you serve already gathers
Using mutual connections, carefully, and never as a sales channel

Three non-adopter interviews are harder to schedule than ten switch interviews, but the missing jobs they reveal are the ones the rest of the playbook can’t see.

When to run which:

Switch-in only. A new team learning the technique, or a product where adoption and retention are both healthy.
Switch-in plus switch-out. The default for a team that wants the full picture of who they keep and who they lose. Run a session of each in the same fortnight; make sense of each separately, then compare the job statements.
All three. When growth has plateaued and churn data alone doesn’t explain it. The non-adopter shape is the one that finds the job you haven’t named yet.