GreenBox delivers weekly produce boxes from local farms to 200 subscribers in Perth. Recent customer interviews revealed that people stay for the convenience – not the local sourcing the team assumed. Now the team needs to find out what else they believe that they haven’t actually tested.
The JTBD interviews gave the GreenBox team a breakthrough. Subscribers don’t stay for fresh local vegetables. They stay because GreenBox eliminates weeknight dinner stress. The box arrives, dinner is decided, one less thing to worry about.
That insight reshaped the product roadmap. Recipe cards went into every box. Churn dropped from 8% to 5% in the first month. The team felt like they’d cracked it.
But the interviews also revealed something less comfortable: a lot of what the team believes about the business is assumption, not fact.
Maya believes subscribers value local sourcing. She built the entire brand around it. “Farm to table, within fifty kilometres.” It’s on the website, on the boxes, in every piece of marketing Sam has ever written. It’s the emotional core of GreenBox’s identity.
But does anyone actually care?
Tom believes the substitution algorithm is good enough. When a farm can’t deliver what they promised, the system swaps in similar items. Tom spent weeks building and refining it. He’s proud of the logic. It handles seasonal availability, dietary preferences, and farm reliability scores.
But has anyone checked whether subscribers notice or care about the substitutions?
Sam believes word-of-mouth is the main acquisition channel. He points to the referral programme and the number of subscribers who say they heard about GreenBox from a friend.
But is that actually what the sign-up data shows?
Priya believes the weekly delivery cadence is right. Every subscriber gets a box every week. That’s how it’s always worked.
But did anyone ask subscribers whether they want a box every week?
These aren’t minor details. They’re load-bearing assumptions. The business model, the brand, the pricing, and the product roadmap all rest on them. If any of them are wrong, the team could be optimising the wrong things for the next six months.
How assumptions hide
The tricky thing about assumptions is that the team doesn’t experience them as assumptions. They experience them as facts. “Subscribers value local sourcing” doesn’t feel like a guess – it feels like the foundation of the business. Maya would have said, with total confidence, that local sourcing is why people subscribe. Until the JTBD interviews showed otherwise.
That’s how assumptions work. The ones you’re most confident about are often the ones you’ve tested least. Nobody tests what they consider obvious.
Lee brings this up at the Monday standup. “The JTBD work surfaced one assumption you didn’t know you had – that subscribers stay for the produce rather than the convenience. What other assumptions are you running on?”
Silence. Then Maya: “I don’t think we have that many assumptions. We know our customers pretty well now.”
“Let’s find out,” Lee says.
Introducing Assumption Mapping
Assumption Mapping is a structured way to surface, categorise, and prioritise the things you believe but haven’t validated. It’s not a single person’s invention – it draws on Lean Startup thinking, design thinking, and the kind of risk-assessment work that good product managers do instinctively. But structuring it as a team exercise forces everyone’s assumptions into the open, which is the point.
The process is deliberately simple.
Step 1: List your assumptions. Everything the team believes about the business, the customers, the market, the product. No judgement, no filtering. Get them all out.
Step 2: Rate each assumption on two axes. How critical is it to the business? (If this assumption is wrong, how badly does it hurt?) And how much evidence do you have for it? (Have you tested it, or is it just a feeling?)
Step 3: Plot on a 2x2 grid.
Horizontal axis: Low Evidence → High Evidence. Vertical axis: Low Risk → High Risk.
- High risk, low evidence – TEST IMMEDIATELY. These are the dangerous ones. The business depends on them and you don’t know if they’re true.
- High risk, high evidence – Monitor. Critical to the business, but you’ve got data backing them up. Keep an eye on them.
- Low risk, low evidence – Park. You don’t know if they’re true, but it doesn’t matter much right now. Come back later.
- Low risk, high evidence – Fine. You know they’re true and they’re not critical. Move on.
The quadrant that matters most is top-left: high risk, low evidence. That’s where the landmines live.
Running the session
Lee facilitates. Sticky notes again – but this time, each note is an assumption, not a domain event.
Dave is here. Maya invited him after the JTBD interviews – partly because the assumptions about farms need a farmer’s perspective, and partly because Dave has a way of saying things that cut through the noise. He drove in from Margaret River this morning, two hours in his ute with ABC Country playing the whole way. He sits at the end of the table in his work shirt, arms folded, watching the team arrange their markers and sticky note pads. He hasn’t been in this office since the Event Storm months ago. The walls are different now – covered in printouts, transcripts, sticky notes from the JTBD work. Patrick’s quote is pinned above the whiteboard: “I was paying twenty-five dollars a week to feel bad about myself.”
Dave reads it. He doesn’t say anything.
“Write down everything you believe about GreenBox that you haven’t actually tested,” Lee says. “Not features. Not plans. Beliefs. Things you’d bet the business on. Things you’d be shocked to discover were wrong.”
The team writes for ten minutes. The notes pile up.
Maya’s notes:
- Subscribers value local sourcing
- Farms will scale with us as we grow
- Our price point ($25/box) is competitive
- Subscribers prefer a curated box over choosing their own items
- The brand matters more than the price
Tom’s notes:
- The substitution algorithm produces acceptable results
- The platform can handle 1,000 subscribers without major rework
- Farms will use the portal to submit availability (rather than calling Maya)
- Weekly delivery is the right cadence
Priya’s notes:
- Subscribers want more variety, not less
- Mobile is the primary way subscribers manage their account
- The signup flow conversion rate is acceptable
Sam’s notes:
- Word-of-mouth is our primary acquisition channel
- Subscribers would recommend GreenBox to friends
- Our Instagram content drives sign-ups
- Churn is mainly driven by value perception, not logistics
Jas’s notes:
- Subscribers read the recipe cards
- The unboxing experience matters for retention
- People who cancel would come back if we offered a discount
Dave writes slowly. His handwriting is large and deliberate, the kind of hand that’s used to signing delivery dockets, not writing on sticky notes. He produces three:
- Farms will scale with us as we grow
- We’re the only option for people who want this
- Farmers will keep supplying if GreenBox has a bad quarter
The second note catches Sam’s eye. He writes his own version of it, more precisely: “We don’t have a serious competitor in Perth.” He sticks it on the wall next to Dave’s.
Twenty-four assumptions. Some overlap, which is fine – duplicates mean multiple people share the same unvalidated belief, which makes them more interesting, not less.
Plotting the map
Lee draws the 2x2 grid on the whiteboard. The team takes each assumption and debates where it belongs. This is where the interesting conversations happen.
“Subscribers value local sourcing.”
Maya instinctively puts it in the top-right: high risk, high evidence. “It’s our brand identity. And the JTBD interviews showed that people mention local sourcing.”
Lee pushes back. “How many mentioned it as the primary reason they subscribe?”
Maya pauses. “I’d have to check the transcripts.”
Sam pulls up the LLM’s analysis from the interview round. He’d asked it to rank the themes by frequency. “Local sourcing” appears in nine of the fifteen interviews – but as the primary motivator in only three. In the active subscriber group, convenience and dinner-stress relief dominated. Local sourcing was mentioned, but as a secondary benefit. A nice-to-have, not a deal-breaker.
The assumption moves to the top-left: high risk (it’s central to the brand and pricing), low evidence (nobody’s actually tested whether subscribers would stay without it).
That move is uncomfortable. Maya built GreenBox around local sourcing. It’s not just a feature – it’s personal. Her family farmed. She believes in supporting local agriculture. Discovering that subscribers might not share that belief as strongly as she does feels like a challenge to her identity, not just her business strategy.
Lee is careful here. “This isn’t about whether local sourcing is good. It’s about whether it’s the reason people pay $25 a week. Those are different questions.”
“The substitution algorithm produces acceptable results.”
Tom puts it at high risk, high evidence. “We’ve been running it for months. Nobody complains.”
Priya raises her hand. “Nobody complains to us. But three of the churned subscribers in the JTBD interviews mentioned getting items they didn’t want. One of them specifically said ‘I got turnips three weeks in a row.’ That sounds like a substitution problem.”
Tom’s face changes. “The algorithm wouldn’t do that. Let me check.”
He opens his laptop during the session. Thirty seconds later: “Ah. Turnips were available in bulk from Dave’s farm for three weeks straight. The algorithm scored them as the best substitution for three different items because they were cheap and plentiful. It’s technically correct – turnips are a root vegetable and the substitution rules allow root-for-root swaps. But getting turnips three times in a row is a terrible customer experience.”
The assumption moves to the left. High risk, less evidence than Tom thought. The algorithm works by its own internal logic, but nobody had validated that customers find the results acceptable. Working correctly and working well are different things.
“Word-of-mouth is our primary acquisition channel.”
Sam is confident. “People tell me all the time that they heard about us from a friend.”
Lee asks: “Do you track acquisition source at sign-up?”
“We ask in the welcome email survey.”
“What’s the response rate?”
“About thirty percent.”
“So you know the acquisition source for thirty percent of subscribers. Of those, what percentage say word-of-mouth?”
Sam checks. “Forty-five percent of respondents.”
“So forty-five percent of thirty percent. That’s about fourteen percent of all subscribers where you actually know word-of-mouth drove the sign-up. For the other seventy percent, you’re guessing.”
The assumption moves to the left. Sam was generalising from a small, self-selected sample. The people who respond to welcome surveys might not be representative of the whole base.
“Farms will scale with us as we grow.”
Maya puts it in the top-right. “I talk to Dave and Rachel every week. They’re committed.”
Dave clears his throat. The room turns to look at him. He’s been quiet for most of the session, watching the sticky notes move around the grid with the patient expression of a man who’s seen plenty of ideas come and go.
“Last bloke who asked me to scale went bust and owed me eight thousand dollars.”
The room goes still.
Dave doesn’t raise his voice. He doesn’t need to. “Farm-to-table scheme out of Busselton. Three years ago. Promised guaranteed orders. I expanded my planting for them. Hired a casual for harvest. They folded in August and I was out the produce, the labour costs, and the eight grand they owed me. Never saw a cent.”
He looks at Maya. Not with hostility – with the kind of frank assessment you give a stock fence before leaning on it.
“I’m here because I trust you, Maya. I trust that you grew up on a farm and you know what it costs when things go wrong. But trust doesn’t plant seeds. Contracts plant seeds. And right now, you and I have a handshake.”
Lee lets the silence sit for a long count. Then he says, gently: “That’s a perfect example of why we’re doing this exercise. The assumption isn’t ‘Dave trusts us.’ The assumption is ‘farms will scale with us.’ And the evidence for that is a handshake and a history of being burned.”
The note moves firmly to the top-left: high risk, low evidence.
Maya writes “formalise farm contracts” on a fresh sticky note and puts it in her pocket. She’ll deal with that later. But Dave’s words – last bloke who asked me to scale went bust – sit in the room like weather.
“We don’t have a serious competitor in Perth.”
Sam picks up his note. “I wrote this because of something that came up in the JTBD interviews. Two churned subscribers mentioned a company called Freshly. I’ve been doing some research.”
He opens his laptop. “Freshly launched in Sydney four months ago. They’ve raised twelve million dollars in Series A. They’re VC-backed, ex-McKinsey founders, and they’re expanding to Perth. They’re recruiting delivery drivers in Perth right now – I found the job ads on Seek.”
Tom leans forward. “What do they charge?”
“Eighteen dollars a week.”
The room does the mental arithmetic. GreenBox charges twenty-five.
“Their produce isn’t locally sourced,” Sam continues. “Wholesale markets. Their technology is polished – they’ve got an app with real-time delivery tracking, algorithmic substitution, and a slick onboarding flow. They’re everything we’re not in terms of scale and funding.”
Maya’s jaw tightens. She knew about Freshly from Greg’s interview, but she hadn’t seen the full picture until now.
The note goes straight to the top-left quadrant. High risk, almost no evidence. The team had been operating as if GreenBox was the only game in town. It isn’t. A competitor with sixty times their funding is about to enter their market.
Dave, from the end of the table, says: “Freshly rang me last week.”
Every head turns.
“Asking about supply. I told them I was committed elsewhere. But they’ll ring Rachel next, if they haven’t already.”
The session continues for another twenty minutes, but the energy in the room has shifted. When it’s done, the whiteboard looks like this:
Six assumptions in the “Test Immediately” quadrant. Six things the business depends on that nobody has actually validated.
Maya stares at the board. “Local sourcing is in the red zone.”
“It is,” Lee says. “And it might turn out to be solid. But right now, you’re betting the brand on something you haven’t tested. That’s the uncomfortable part of this exercise – it doesn’t tell you your assumptions are wrong. It tells you that you don’t know whether they’re right.”
Designing cheap experiments
The team needs to test five assumptions. They can’t do five research projects – they need to ship product and grow the subscriber base simultaneously. The experiments need to be cheap: hours, not weeks.
Lee introduces a principle from Lean Startup: find the smallest, cheapest experiment that would change your mind.
“For each assumption, ask: what would we need to see to believe this is true? And what’s the cheapest way to see it?”
The team designs experiments over lunch. LLMs help – Maya and Sam use them to generate survey questions, draft A/B test copy, and prototype landing pages. Each experiment is designed to cost hours, not weeks.
Experiment 1: Do subscribers value local sourcing?
They create a short survey sent to all active subscribers. Three questions:
- Rank these five factors in order of importance to your GreenBox subscription: convenience, local sourcing, quality of produce, price, recipe cards.
- Would you consider a $20/box option that included some non-local items alongside local ones?
- What would make you cancel your subscription tomorrow? (Open text.)
Sam asks the LLM to help phrase the questions to minimise leading bias. The LLM suggests changing “Would you consider a cheaper option with non-local items?” to “If a mixed-sourcing box were available at a lower price, how likely would you be to switch?” – less emotionally loaded, more likely to get an honest answer.
The survey goes out that afternoon. It takes Sam twenty minutes to build.
Experiment 2: Is the price point competitive?
Jas creates two landing page variants. One shows the current pricing ($25/week for the standard box). The other shows a $20/week option labelled “Mixed Box” alongside the $25 “Local Box.” They run both for a week and track sign-up intent (clicks on “Subscribe” button – not actual sign-ups, just measured interest).
The LLM generates the copy for both variants. Jas adjusts the design. Total time: two hours.
Experiment 3: Will farms scale with us?
Maya calls three of their current farm partners and asks a direct question: “If we needed to double our order in three months, could you do it? What would you need from us?” This one doesn’t need technology. It needs a phone and thirty minutes.
Experiment 4: What’s the real acquisition channel?
Tom adds a single dropdown to the sign-up flow: “How did you hear about GreenBox?” with options including word-of-mouth, social media, search engine, local press, and other. Mandatory field. Takes him an hour to build and deploy.
Experiment 5: Is weekly the right cadence?
Sam adds a question to the post-delivery email: “Would you prefer to receive your box weekly, fortnightly, or on a flexible schedule?” Simple, zero-cost, answers come in with each delivery cycle.
Five experiments. Total cost: about eight hours of work across the team. Results expected within one to two weeks.
The results
Two weeks later, the team reconvenes. The data is in.
Local sourcing:
168 subscribers responded to the survey (84% response rate – much higher than the usual welcome survey, because Sam sent it as a standalone email with the subject line “Help us make GreenBox better for you”).
When asked to rank factors, the results were:
- Convenience (ranked #1 by 38% of respondents)
- Quality of produce (ranked #1 by 26%)
- Recipe cards (ranked #1 by 18%)
- Local sourcing (ranked #1 by 12%)
- Price (ranked #1 by 6%)
Only 12% ranked local sourcing as the most important factor. Convenience and quality dominate.
On the mixed-sourcing question: 60% said they would likely or very likely switch to a $20 mixed-sourcing box. 22% said they’d stay with the $25 local box. 18% were unsure.
Maya sits with this for a moment. Sixty percent of her subscribers would accept non-local produce for a five-dollar saving. The thing she believed was the core of the business – local sourcing – is a secondary factor for the majority of customers.
“I feel like I’ve been punched in the stomach,” she says.
Lee lets the silence sit. Then: “This doesn’t mean local sourcing is worthless. Twelve percent of your subscribers rank it first. That’s twenty-four people who might leave if you drop it. And the brand identity – ‘local farm produce’ – is how you differentiate from supermarket delivery services. But it does mean that 100% local sourcing at $25 a box might not be the only viable model. There might be a $20 mixed box that expands your market significantly.”
This is a massive business model insight. The team didn’t know it existed two weeks ago. It came from a twenty-minute survey, not a three-month research project.
Price point:
The landing page A/B test showed 2.3x more clicks on the “Subscribe” button when the $20 mixed option was available alongside the $25 local option. People didn’t just want the cheaper box – the presence of a choice made them more likely to subscribe at all. Having two options reduced the perceived risk of committing.
Farm scaling:
Maya’s phone calls revealed mixed news. Two of the three farms said they could increase supply by 50% within three months. The third – Dave, their biggest supplier – said he’d cap out at current levels. He’d need a full growing season to expand. This means the team can’t rely on current partners alone to reach 1,000 subscribers. They’ll need new farm partnerships, or the mixed-sourcing model that reduces dependence on local-only supply.
Acquisition channel:
After two weeks of mandatory sign-up tracking, the data was clear. Word-of-mouth accounted for 31% of new sign-ups. Google search: 28%. Instagram: 19%. Local press coverage from a Margaret River newspaper article: 14%. Other: 8%.
Sam was partially right – word-of-mouth is the biggest single channel. But it’s not dominant. Search and social together account for nearly half of acquisitions. The team has been under-investing in SEO and over-crediting referrals.
Delivery cadence:
41% of respondents wanted weekly delivery. 35% wanted fortnightly. 24% wanted a flexible schedule. The assumption that weekly is right was wrong for the majority of subscribers. More than half wanted less frequent delivery.
This explains some churn the team hadn’t understood. Subscribers who wanted fortnightly delivery but could only get weekly were either accumulating unwanted produce (and feeling guilty about waste) or cancelling entirely. A fortnightly option wouldn’t just be a nice feature – it would directly reduce churn by serving a segment the current product doesn’t fit.
The hard conversation
The team sits with the results. Five assumptions tested, plus the competitor question answered by reality rather than experiment. Three were partially or fully wrong:
- Local sourcing is not the primary value driver (it’s fourth out of five).
- Word-of-mouth is the biggest channel but not dominant (search and social are nearly as large).
- Weekly delivery is wrong for the majority of subscribers.
Two were roughly correct:
- Farms can scale, but with constraints (partially validated – needs new partners).
- The price point creates a barrier that a mixed option could resolve (validated by the A/B test).
Maya is quiet for a long time. Then she says something that takes courage: “I built this business around an assumption I never tested. I assumed people cared about local sourcing as much as I do. They don’t. They care about convenience. I’m not sure what that means for us.”
Lee is honest. “It means you have options you didn’t know you had. A mixed-sourcing box at $20 could open up a much larger market. A fortnightly option reduces churn. Neither of these kills the local brand – you can still offer a premium local box for the twelve percent who value it most. But the path to 1,000 subscribers probably isn’t ‘1,000 people who care deeply about local produce.’ It’s ‘1,000 people who want dinner stress eliminated, some of whom also care about local.’”
“That’s a different business than the one I set out to build,” Maya says. She looks at Dave. He’s been quiet since his comment about Freshly calling. His arms are folded again. He’s watching her.
“Maybe,” Lee says. “Or maybe it’s the same business, with a broader front door.”
Dave stands up. He needs to get back to the farm before dark. He shakes Maya’s hand at the door.
“You’ll work it out,” he says. It’s not a compliment – it’s a bet. The same bet he made when he agreed to supply GreenBox on a handshake. He’s still holding.
Maya watches his ute pull out of the car park. She thinks about his eight thousand dollars. She thinks about her father selling the farm. She thinks about the “pausing operations” email she hasn’t written yet but can feel forming at the edges of her mind, like weather moving in from the coast.
How assumption mapping feeds back
The assumption mapping exercise doesn’t exist in isolation. It feeds directly back into the tools the team already uses.
Impact Map: The goal is still 1,000 subscribers. But the impacts now include “subscribers choose their preferred cadence” and “potential subscribers see a price point they’re comfortable with.” These are new branches on the map that didn’t exist before the assumptions were tested.
Example Mapping: The fortnightly delivery option and the mixed-sourcing box both need to be Example Mapped before anyone builds them. There are edge cases everywhere – what happens if a fortnightly subscriber wants to switch to weekly? What counts as “non-local” produce? Do the recipe cards change for the mixed box?
Event Storming: If the team introduces mixed sourcing, the supply chain changes. New farm partners, different procurement processes, different quality checks. That’s new domain territory that needs to be Event Stormed.
The assumptions don’t just produce a grid on a whiteboard. They produce work – specific, prioritised work that feeds into the techniques the team already knows how to use.
When to use Assumption Mapping
Before major investment decisions. If the team is about to spend significant time or money on something, map the assumptions underlying it first. The cost of the mapping exercise is trivial compared to the cost of building on a wrong assumption.
After discovery work reveals surprises. The JTBD interviews surprised the GreenBox team. That surprise was a signal: if one major assumption was wrong, others might be too. Assumption mapping is the systematic follow-up.
When the team disagrees about direction. Disagreements often hide different assumptions. If Maya thinks they should double down on local sourcing and Sam thinks they should invest in SEO, they might have different beliefs about what drives growth. Mapping the assumptions makes the disagreement concrete rather than political.
Periodically, as the business evolves. Assumptions that were true at 50 subscribers might not hold at 500. The market changes. Competitors appear. Customer expectations shift. Run the exercise quarterly, or whenever the business hits a new stage.
When not to use it
When you’re already testing fast. If your team is already running experiments and validating hypotheses weekly, a formal assumption mapping session might add ceremony without adding insight. The value of the exercise is proportional to how many untested assumptions are hiding in the business.
When the team isn’t psychologically safe enough to admit uncertainty. Assumption mapping requires people to say “I believe this but I don’t have evidence.” In a team where admitting uncertainty feels dangerous, the exercise will produce a sanitised list of safe assumptions. The scary ones – the ones that actually matter – will stay hidden. Fix the safety problem first.
As a substitute for talking to customers. The assumption map tells you what to test. It doesn’t do the testing. Some teams map their assumptions, nod sagely at the “Test Immediately” quadrant, and then never run the experiments. The map without the experiments is just a wall decoration.
The bigger picture
The GreenBox team now has validated data about their customers, their market, and their business model. The JTBD interviews told them what job subscribers are hiring GreenBox for. The assumption mapping told them which beliefs are solid and which are shaky.
But there’s a bigger question looming. Maya has a board meeting in three weeks. She needs to present a credible path to 1,000 subscribers. The insights are powerful – mixed sourcing, fortnightly options, SEO investment. But do the numbers actually add up? Can GreenBox reach 1,000 subscribers with a business model that works financially – especially with a competitor about to enter the market at $18 per week?
That’s a question about the business model itself. And it’s where Lee starts to hit the limits of what he can help with.
Next week, the team zooms out to look at the whole business on one page – and discovers that the unit economics have a problem nobody spotted. That’s Business Model Canvas (coming 30 June).