Jobs to Be Done Framework: A Demand-First Guide (2026)
JTBD only works if a job is observable demand, not a persona guess. Here's how to find, write, and prioritize real jobs.
Jobs to Be Done gets taught like a persona exercise with better vocabulary. It isn't. A job is only real if you can point to someone already spending money, time, or attention to get it done today, not a job you imagine they'd want solved.
What the Jobs to Be Done Framework Actually Is (and the Milkshake Myth)
The core idea: people hire products to make progress
The premise behind the jobs to be done framework is simple: people don't buy products because of what the products are, they hire them to make progress in a specific situation. A commuter doesn't want a milkshake, they want something thick enough to last a boring drive and one-handed enough not to spill. The product is a means to an end, and the end is the job.
Why 'job' means observable behavior, not a wish
The part teams skip is the discipline. A job isn't "what would be nice," it's something you can watch people already doing, paying for, or working around with a clumsy substitute. If nobody is currently spending effort on it, in a spreadsheet, a stack of sticky notes, a competitor's product, or an awkward manual process, it isn't a job yet. It's a guess wearing job language.
The two schools: Ulwick's ODI vs. Christensen/Klement's switch
Two lineages dominate how teams apply this. Tony Ulwick's Outcome-Driven Innovation, detailed through Strategyn, treats the job as a stable process with measurable desired outcomes you can survey at scale. Clayton Christensen's version, popularized in his HBR article on knowing your customers' jobs and extended by Alan Klement's job story format documented at jtbd.info, focuses on the moment someone switches from one solution to another. Both reject personas. They disagree on method.
| Dimension | Ulwick's Outcome-Driven Innovation | Christensen/Klement's switch approach |
|---|---|---|
| Unit of analysis | Job steps and desired outcomes | The switching moment and circumstance |
| Primary method | Large-sample outcome surveys | Small-sample depth interviews |
| Best for | Mature markets, feature prioritization | New markets, positioning, messaging |
| Output | Ranked outcome statements | Job stories and forces diagram |
Job Stories vs. User Stories vs. Personas: Why JTBD Replaces the Demographic Lens
The job story format: 'When [situation], I want to [motivation], so I can [outcome]'
Intercom popularized a format built for this exact gap, described in its jobs to be done guide: "When [situation], I want to [motivation], so I can [expected outcome]." Notice what's missing: no name, no age, no job title. The situation does the work a persona used to do, and it does it better because it's tied to context, not demographics.
Why 'Marketing Mary, 34' predicts nothing
A persona describes who buys. It rarely predicts who's about to switch. Two 34-year-old marketing managers can have wildly different jobs depending on whether they just lost a teammate, just got a new budget, or just got yelled at in a board meeting. Nielsen Norman Group's comparison of JTBD and personas makes the same point: situation beats segment for predicting behavior.
Functional, emotional, and social dimensions of a single job
Every real job has three layers stacked on top of each other. The functional layer is the task itself (file the report). The emotional layer is how the person wants to feel doing it (not stupid, not exposed). The social layer is how they want to be perceived by others (competent, decisive). Miss the emotional or social layer and you can nail the functional job and still lose the sale.
| Lens | Describes | Predicts switching | Example output |
|---|---|---|---|
| Persona | Demographics, role, firmographics | Weakly | "VP of Sales, 40-55, enterprise SaaS" |
| User story | A feature from the user's seat | Not designed to | "As a user, I want to export data" |
| Job story | Situation, motivation, outcome | Directly | "When a deal stalls, I want to see why, so I can unstick it" |
How to Uncover Real Jobs: The Switch Interview
Interviewing the moment of purchase, not the opinion
The switch interview, a technique refined by Bob Moesta and the team behind the Rewired Group's jobs-to-be-done work, doesn't ask people what they want. It reconstructs the actual sequence of events around the last time they switched from one way of solving a problem to a new one. Opinions are cheap and often wrong; a reconstructed purchase decision is a fact pattern.
The timeline: first thought to active looking to decision to first use
A good switch interview walks four stages: the first passive thought that something wasn't working, the moment that thought turned into active looking, the decision point where one option won, and the first use that confirmed (or didn't) the choice. Each stage has its own forces, and skipping straight to "why did you buy it" collapses all four into a tidy story that hides the real trigger.
Four forces: push, pull, anxiety, habit
At the switching moment, four forces are in play: the push of dissatisfaction with the current approach, the pull of the new solution's promise, the anxiety about whether the new thing will actually work, and the habit of the old way that keeps people stuck even when they're unhappy. If you want the operational script for running this kind of interview without leading the witness, our guide on how to run customer discovery interviews lays out a ten-question version built for exactly this.
Finding the Jobs You Can't Interview For: Mining Unfiltered Demand
Where people describe jobs in their own words
You can only run so many switch interviews, realistically ten to fifteen before patterns repeat. But people describe their jobs, unprompted and in their own words, constantly: in support tickets, app store reviews, competitor review threads, and community forums. That's unfiltered demand, and it scales past what any interview schedule can cover.
Reading feature requests as job statements in disguise
A feature request is almost never the job itself, it's someone's best guess at a solution to a job they can feel but haven't articulated. "Can you add a dark mode toggle" might really be "when I work late, I want my eyes to not hurt, so I can keep working without a headache." The request is the artifact; the job is underneath it.
Separating a feature ask from the job underneath it
The move is to ask "what were they doing right before they asked for this" and "what would they do instead if we said no." Our breakdown of where to find real feature requests for product ideas covers six sources of this kind of raw, unsolicited demand, which is the material you mine to infer jobs at a scale interviews can't reach.
Put the Framework to Work
Turn observed jobs into tracked, votable demand
A jobs map that lives in a slide deck goes stale the week you write it. The jobs that matter are the ones with ongoing, visible demand behind them, and that demand needs a place to accumulate so you can see which jobs are heating up and which have gone quiet.
Stop guessing which job matters most
That's the gap requestproduct.com is built to close: instead of maintaining a static JTBD document, you collect and weight real feature requests as they come in, so the observable demand behind each job stays current and votable rather than frozen at the moment someone ran a round of interviews.
Writing Job Statements That Survive Validation
The anatomy of a testable job statement
A job statement earns its place in a roadmap discussion only if it names a situation, a motivation, and an outcome, and if each of those three pieces points to something you could actually go observe. "When I'm closing the books at month end, I want to catch errors before my manager sees them, so I can avoid looking careless" is testable. "Users want better reporting" is not.
Red flag: jobs with no observable proxy
If you can't name a single behavior, workaround, or spend that proves the job exists today, you have a hypothesis, not a job. That's not disqualifying, hypotheses are where real jobs often start, but it means the next step is validation, not a sprint.
Turning a job hypothesis into a validation test
Every job statement should convert into a falsifiable test: a landing page, a concierge offer, a waitlist, a fake-door feature. Our seven-step demand-first playbook for validating a product idea walks through exactly how to design that test before any code gets written, which keeps the job statement honest.
Using Jobs to Prioritize the Roadmap
Importance times satisfaction: the opportunity score
Ulwick's Outcome-Driven Innovation contributes a simple math trick worth borrowing even if you don't run full ODI surveys: score each job on how important it is and how well current options satisfy it, then look for jobs that are highly important but poorly satisfied. That gap is the opportunity.
| Job | Importance (1-10) | Satisfaction (1-10) | Opportunity gap |
|---|---|---|---|
| Catch billing errors before month-end close | 9 | 3 | High |
| Export data to a spreadsheet | 6 | 7 | Low |
| Get alerted when a deal stalls | 8 | 4 | High |
| Customize dashboard colors | 3 | 6 | Low |
Weighting jobs by how many people demonstrably have them
Importance and satisfaction scores from a handful of interviews are directional, not statistical. Weight them against how many people you can actually show have this job, through request volume, support ticket frequency, or workaround prevalence, before you let a high opportunity score drive a roadmap decision on its own.
From job map to roadmap sequence
Once jobs are scored and weighted, the sequencing question becomes mechanical rather than political. Our guides on how to prioritize feature requests with a demand-weighted system and how to build a demand-first product roadmap pick up exactly here: jobs answer why something matters, these two translate that why into an actual build order.
Jobs, Competitors, and Pricing: What People Hire Instead of You
Your real competition is any product that does the job, including a spreadsheet
Once you define the job instead of the category, the competitive set gets a lot bigger and a lot more honest. A project management tool doesn't just compete with other project management tools, it competes with a shared spreadsheet, a whiteboard photo texted to a group chat, and doing nothing and hoping it works out. Christensen's milkshake research, described in the HBR piece on marketing malpractice, found the milkshake's real competitor wasn't other milkshakes, it was a banana and a bagel.
Finding the demand gap competitors leave open
Mapping jobs against existing alternatives usually turns up at least one job that's important and poorly served by everyone, including you. That's the gap worth building for. Our demand-gap method for competitor analysis is built to surface exactly that kind of underserved job rather than just listing feature parity.
Pricing to the value of the job, not the feature list
If the job is worth real money or real time to the person hiring your product, price toward that value rather than toward a feature checklist. Our guide to pricing a SaaS product covers how to anchor willingness to pay to the outcome a job delivers instead of the number of toggles in settings.
Measuring Whether You Nailed the Job
Signals that the job is getting done (retention, not satisfaction surveys)
Satisfaction surveys ask people how they feel about your product. Retention, usage frequency at the moment the job recurs, and whether people stop maintaining their old workaround all show you whether the job is actually getting done. The second kind of signal is harder to game and harder to ignore.
Watching for jobs you solved too narrowly
A common failure mode is solving a thin slice of the job well enough that people adopt you for that slice and then go find something else for the rest. That shows up as usage that plateaus below what the job's full scope would predict, which is worth watching for even when early retention looks fine.
When a 'done' job reopens as the market shifts
Jobs aren't permanently solved. New alternatives, new habits, or new regulations can reopen a job you thought was closed. Our breakdown of five demand signals that beat the 40 percent survey for measuring product-market fit is the lagging proof that you matched a real job, and it's also the early warning system for when that match starts to slip.
Frequently Asked Questions
What is the Jobs to Be Done framework in simple terms? It's a way of understanding why people buy something by focusing on the progress they're trying to make in a specific situation, rather than their demographics or stated preferences.
What is the difference between a job story and a user story? A user story describes a feature from a user's seat ("as a user, I want X"). A job story describes the situation, motivation, and outcome behind a switch ("when Y happens, I want to Z, so I can achieve W"), without assuming a specific feature or user type.
Who created the Jobs to Be Done framework? The framework has two major lineages: Tony Ulwick's Outcome-Driven Innovation methodology, and Clayton Christensen's version, popularized through his writing and later extended into the job story format by Alan Klement and applied operationally by practitioners like Bob Moesta.
How do you write a good job statement? Name a specific situation, a motivation, and an outcome, and make sure each piece points to something observable, a behavior, spend, or workaround that already exists today rather than a wish.
Does JTBD replace personas entirely? For predicting why someone would switch, yes, situation beats demographic segment. Personas can still be useful for things like tone of voice or support staffing, but they shouldn't drive prioritization decisions.
How many interviews do I need to find the jobs? Switch interview patterns tend to repeat somewhere between ten and fifteen interviews. Beyond that, mining unfiltered demand sources like reviews and feature requests becomes more efficient than running more interviews.
How is JTBD different from a feature request? A feature request is a proposed solution; a job is the underlying situation and outcome the request is trying to solve. The same job can generate many different, even contradictory, feature requests.
Conclusion
The jobs to be done framework earns its keep only when you hold it to the same standard you'd hold any other product claim: show the demand. A job you can't observe someone already acting on, in an interview's switching story or in the unfiltered language of a feature request, isn't a job yet. Treat it as a hypothesis, write it down with enough specificity to test, and let real demand, not a persona deck, decide what gets built next.