GreenBox has 200 subscribers but 8% monthly churn. Over three weeks of discovery – Jobs to Be Done, Assumption Mapping, Business Model Canvas – the team has uncovered more problems than they can fix at once. Now they need to decide what changes first.
It’s Monday morning and the office whiteboard is covered. Three weeks of discovery work have produced a wall of insights, sticky notes, canvas printouts, and scribbled questions. Maya stands in front of it with a coffee that’s gone cold.
Sam arrives early, which is unusual. He puts his bag down and opens his laptop before he takes off his jacket. “We lost three subscribers over the weekend.”
Maya turns from the whiteboard. “Churn?”
“Not exactly. They switched. To Freshly.” Sam turns his laptop around. Freshly’s Perth launch page fills the screen – a clean hero image of a produce box, the $18 price tag prominent, a “Now delivering in Perth” banner across the top. “They went live on Friday. Three of our subscribers signed up over the weekend and cancelled with us. One of them – Louise, from the JTBD interviews – sent a message: ‘Sorry, but $18 is $18.’”
Maya stares at the screen. She knew this was coming. The assumption mapping had surfaced Freshly as a threat. Charlotte’s BMC questions had forced the pricing conversation. Dave had told her Freshly was calling farms. But knowing it’s coming and seeing “three subscribers switched” on a Monday morning are different experiences.
“Seven dollars a week,” Maya says. “That’s three hundred and sixty-four dollars a year. Of course people switch.”
She can see the whole picture now. It’s not a comfortable picture. The JTBD interviews showed that customers hire GreenBox for weeknight dinner stress relief, not fresh local produce. The assumption mapping revealed that local sourcing – the emotional core of the brand – isn’t as valued as the team believed. The Business Model Canvas exposed unit economics that don’t work at current pricing with 100% local sourcing. Charlotte’s questions about customer acquisition cost and lifetime value made the financial gaps painfully visible.
The insights are clear. A two-tier pricing model – $25 local box, $20 mixed-sourcing box – could fix the economics. A pause button would directly reduce churn. The value proposition needs repositioning around convenience, not provenance. SEO is an underinvested acquisition channel. The recipe cards are working but the marketing doesn’t match what customers actually care about.
Maya knows all of this. The team knows all of this. And that’s the problem.
Too much to fix
Tom arrives at nine and finds Maya staring at the whiteboard. “You look like someone who’s been up since five.”
“Six,” Maya says. “I’ve been trying to figure out what we do first.”
“Everything?”
“That’s what I keep landing on. And that’s wrong.”
She’s right that it’s wrong, but she can’t articulate why. The board meeting is in three weeks. Every initiative feels urgent. The pause button would reduce churn immediately. The two-tier pricing model is essential for the board pitch. The value prop repositioning would improve acquisition. SEO could compound over time. Mixed sourcing needs a pilot before they can commit to it in the pitch.
By the time the rest of the team arrives, Maya has written five priorities on the whiteboard, each circled in red, each with an exclamation mark.
- Ship the pause button (reduces churn)
- Launch two-tier pricing model (fixes unit economics)
- Reposition the value prop in all marketing (aligns messaging to JTBD insight)
- Run a mixed-sourcing pilot (validates the supply chain change)
- Start SEO foundation work (diversifies acquisition)
Priya reads the list. “We’re five people.”
“I know.”
“That’s five initiatives for five people. One each.”
“I know.”
Sam looks at the board. “Plus we still have to pack and ship two hundred boxes a week, manage farm relationships, keep the platform running, and prepare a board presentation.” He’s listing his own workload, though he doesn’t frame it that way. Sam never frames it that way. He has forty-three unread support emails from the weekend, including a complaint about a damaged box, a question about allergen information he can’t answer because the system doesn’t track it, and Louise’s cancellation message that he’s already read four times.
“I know.” Maya puts down her marker. “I don’t know how to choose.”
She thinks about the email draft sitting in her outbox. The “pausing operations” email she wrote after the BMC session and never sent. She thinks about how close she came to sending it. She thinks about how the decision she makes this morning – which of these five things to do first – might determine whether she ever has to send it.
Everyone has a different answer
Lee is on the call. Charlotte is on the call. The team spends thirty minutes arguing.
Tom thinks the pause button should be first. “It’s the highest-leverage change we can make. Every subscriber who pauses instead of cancelling is revenue we keep. And it’s a small engineering lift – Priya and I can build it in a week.”
Sam disagrees. “If we fix the value prop in our marketing, we’ll acquire better-fit customers who churn less in the first place. The pause button is a bandage. The messaging fix addresses the root cause.”
Jas pushes for the two-tier pricing model. “The board meeting is in three weeks. If we can show them a working two-tier model with early data, the Series A conversation changes completely. Everything else can wait.”
Priya wants the mixed-sourcing pilot to go first. “We can’t pitch two-tier pricing to the board if we haven’t validated the supply chain. What if mixed sourcing doesn’t work? What if the produce quality drops? We’d be pitching something we haven’t tested.”
Maya keeps circling back to SEO. “Word-of-mouth and search are nearly tied as acquisition channels. If we invest in SEO now, it compounds. Every month we wait is a month of compounding we lose.”
Charlotte has been listening. She lets the argument run. She lets it run past the point where it’s productive, into the territory where people start repeating themselves with increasing volume, because that’s when the real dynamic becomes visible.
After thirty minutes, she says: “Can I ask a question?”
The room goes quiet.
“If you could only change one thing in the next quarter, what would it be?”
Another round of arguments. Tom says churn. Sam says messaging. Jas says pricing. Priya says supply chain validation. Maya says SEO.
“Five people, five answers,” Charlotte says. “That’s not a disagreement about priorities. That’s the absence of a framework for deciding. You each have good instincts. But instincts pull in five directions when there’s no structure to resolve them.”
Now / Next / Later
Charlotte shares her screen. A blank document with three columns.
“This is the simplest roadmapping format I know,” she says. “Now, Next, Later. It’s not a Gantt chart. It’s not a detailed timeline. It’s a prioritisation tool that acknowledges you don’t know enough to plan twelve weeks in detail.”
She explains the three horizons.
Now is the next four weeks. These are the things you’re committing to. High impact, and either high urgency or high confidence that they’ll work. You should be able to name the people doing the work and describe what “done” looks like.
Next is four to twelve weeks out. These are things that matter but can wait. They might need more information before you can start. They might depend on something in the Now column finishing first. You’re not committing to them yet – you’re signalling intent.
Later is beyond twelve weeks. Good ideas that aren’t ready. Maybe the market needs to shift. Maybe you need more data. Maybe you need more people. “Later” doesn’t mean “never.” It means “we don’t know enough yet to commit.”
“The power of this format,” Charlotte says, “is that it makes the trade-offs visible. Everything can’t be Now. If it is, nothing is. The act of moving something from Now to Next is a decision. It forces the conversation about what matters most.”
Scoring the initiatives
Charlotte draws a simple 2x2 on the shared screen. The vertical axis is impact – specifically, impact on the metric that matters most right now. She asks the team: “What’s the one metric that, if it improved, would change everything?”
“Churn,” Maya says, without hesitation. “If we get churn below 5%, we can grow to 1,000 with our current acquisition rate. If we don’t, we can’t.”
Charlotte writes “churn reduction” on the vertical axis. The horizontal axis is effort and risk – a combined measure of how hard the initiative is to execute and how much uncertainty surrounds it.
“Score each initiative,” she says. “High impact or low impact on churn. High effort/risk or low effort/risk. Don’t overthink it. Gut feel is fine here – you’ve spent three weeks building that gut feel through actual discovery.”
The team works through the list.
Pause button: High impact on churn (subscribers who’d cancel can pause instead – directly reduces the 8% monthly number). Low effort (Tom and Priya estimate a week of engineering). Low risk (the feature is well understood, common in subscription businesses, and the Example Mapping session could be done in a morning).
Two-tier pricing model: Medium-high impact on churn (better-fit pricing means fewer subscribers who feel they’re overpaying). High effort (requires new subscription plans, updated billing, marketing materials, farm communication about order changes). Medium risk (the assumption mapping survey validated demand, but execution is complex). Critical for the board meeting.
Value prop repositioning: Medium impact on churn (better messaging attracts better-fit subscribers, but doesn’t directly help current ones). Medium effort (Sam can rework the website and email copy). Low risk.
Mixed-sourcing pilot: Low direct impact on churn (it’s a supply chain change, not a customer-facing one – yet). Medium effort (requires new supplier relationships). High risk (untested produce quality, potential farm partner friction – Charlotte flagged this in the BMC session).
SEO foundation: Low impact on churn (SEO drives acquisition, not retention). Medium effort (content strategy, technical SEO, ongoing). Low risk but slow to compound.
Charlotte plots them.
Vertical axis: Impact on churn. Horizontal axis: Effort & risk.
The picture clarifies things immediately.
The pause button is the obvious first move. Highest impact, lowest effort, lowest risk. It’s the kind of initiative that should have been built weeks ago – the team just didn’t have the data to see how much it mattered until the JTBD interviews revealed how many subscribers cancel when what they really want is a break.
The two-tier pricing model is the big bet. It’s high effort and the board needs to see it. It goes in Now too – not because it’s easy, but because the board meeting creates a hard deadline.
Priya objects to the mixed-sourcing pilot being in “Defer.” “We need supply chain data before we can commit to two-tier pricing.”
Charlotte nods. “You’re right. But you don’t need a full pilot. You need three phone calls to wholesale suppliers and a week of test orders. That’s not a separate initiative – it’s part of the two-tier pricing preparation. The full pilot, where you’re sending mixed boxes to actual subscribers, is Next.”
This is a subtle but important distinction. Charlotte isn’t dismissing the supply chain work. She’s decomposing it. The validation piece is small and urgent. The full pilot is larger and can wait until the pricing model is designed.
T-shirt sizing: estimating for the roadmap
The roadmap structure helps. It creates calm where there was chaos. But Maya keeps glancing at Sam’s laptop, still open to Freshly’s launch page, and the calm doesn’t quite reach her chest.
She raises a practical concern. “The board will ask ‘when does two-tier pricing launch?’ I can’t answer ‘it’s in the Now column.’”
Charlotte introduces t-shirt sizing – a lightweight estimation technique for roadmap items. Not story points. Not hours. Just Small, Medium, and Large.
“Small is a sprint or less. One or two people, well-understood problem, minimal uncertainty. The pause button is a Small. Medium is two to four sprints. Multiple people, some open questions, but the shape of the work is clear. Value prop repositioning is a Medium – Sam can rework the messaging, but she’ll need to test it with customers. Large is a quarter or more. Cross-functional, significant uncertainty, probably needs breaking down further before anyone starts building. The two-tier pricing model is a Large.”
Tom asks the obvious question. “If two-tier pricing is a Large and we need it for the board meeting in three weeks, are we in trouble?”
“You’re in trouble if you try to ship the whole thing in three weeks,” Charlotte says. “You’re fine if you scope what ‘done’ means for the board. The board doesn’t need working two-tier subscriptions in production. They need a validated plan and a credible timeline. Ship the financial model, the supplier validation, and the pricing page design. That’s a Small. The full engineering work is a Large that starts after the board approves the direction.”
That distinction – estimating the roadmap item, not the sprint story – is the key. Sprint stories don’t need formal estimation at five people. But roadmap items need rough sizing so that the team can promise the right things to stakeholders and avoid committing to a quarter’s work in a fortnight.
The roadmap
Charlotte fills in the three columns.
Marcus, one of the investors who joins the board call that week, sees the roadmap and asks: “Where’s the mobile app? Every competitor has an app. Freshly’s app has 4.2 stars.” Diane cuts in before Maya can respond: “Build the web experience for mobile first. The app can wait until you know what people actually do on their phones.” Charlotte writes “mobile app” in the Next column. Not because it doesn’t matter, but because building an app before understanding mobile behaviour is building a solution before understanding the problem.
Maya wants to push the value prop repositioning into Now. “Sam could do it in parallel while Tom and Priya build the pause button.”
Charlotte shakes her head. “Sam is going to be busy. He’s writing the board presentation. He’s running the customer communication for the pricing change. He’s managing the weekly marketing. If you add value prop repositioning to his plate, you’ll get three things done badly instead of two things done well.”
“Five people can’t run five initiatives,” Charlotte says. “Pick two for Now. The rest goes to Next or Later.”
Maya nods slowly. She doesn’t like it. But she sees it.
Lee adds something quietly. “This connects to the Impact Map. The Now items should trace directly from the map – goal to actors to impacts to deliverables. If an initiative doesn’t connect to the goal through a clear chain, it’s not Now. It might be good, but it’s not Now.”
Tom pulls up the team’s Impact Map. Goal: reduce monthly churn below 5%. Actors: current subscribers, the team, the platform. Impacts: subscribers can take a break without cancelling (pause button), subscribers on the right pricing plan feel they’re getting fair value (two-tier model). Deliverables: pause/resume functionality, subscription tier management, updated billing. The chain is clean.
“What about SEO?” Sam asks. “It doesn’t connect to churn reduction.”
“Exactly,” Charlotte says. “SEO connects to a different goal – acquisition. It’s important. But it’s not the goal for this quarter. It’s Next.”
The quarterly theme
Charlotte introduces one more concept. “A roadmap tells you what to work on. A quarterly theme tells you why. Every quarter should have a clear theme – not a list of features, but a single goal that everything connects to.”
She writes on the shared document:
Q3 Theme: Fix the leaky bucket – reduce monthly churn below 5%.
“Everything you work on this quarter should connect to that theme. When someone suggests a new initiative – and they will, because good teams generate ideas faster than they can execute them – the first question is: does this serve the quarterly theme? If yes, discuss. If no, it goes in Next or Later.”
Jas asks: “What about the board presentation? That’s not directly about churn.”
“It is,” Charlotte says. “The board presentation is about securing the funding that lets you continue working on churn. It’s a meta-activity – it serves the theme by keeping the business alive long enough to execute the plan.”
Tom grins. “So the theme is like a filter for sprint planning.”
“Exactly. Monday morning, first question: does this serve our quarterly goal?” Charlotte pauses. “You’ve been doing sprint planning, right?”
“Sort of,” Priya says. “We plan work on Monday, review on Friday. It’s not formal.”
“That’s fine. Formal doesn’t matter. What matters is that the plan connects to the theme. Every Monday, look at the roadmap. Look at the theme. Ask: what’s the most important thing we can do this week toward fixing the leaky bucket? That’s your sprint goal.”
This is the layer that’s been missing. The team has excellent discovery tools – Event Storming, Example Mapping, Impact Mapping – and now excellent customer and business model understanding. But they didn’t have a way to turn all that understanding into a sequenced plan. The Now/Next/Later roadmap, anchored to a quarterly theme, provides the structure. It doesn’t replace the other tools. It sits above them. The theme guides the roadmap. The roadmap guides the Impact Map updates. The Impact Map guides the sprint goals. The sprint goals guide the Example Mapping sessions that produce concrete work.
The LLM as synthesis engine
Before the prioritisation session, the team had used an LLM to prepare. Maya had asked it to do something specific: take the outputs from the three discovery exercises and produce a one-page brief that could feed a planning conversation.
She’d pasted in the JTBD interview summaries, the assumption map results (including the survey data), and the Business Model Canvas with Charlotte’s annotations. The prompt was simple: “Summarise the key findings from these three discovery exercises into a single-page brief. Focus on: what we’ve learned about our customers, what assumptions have been validated or invalidated, and what the business model implications are. Write it for a team planning session, not for investors.”
The LLM produced a clean summary in about thirty seconds. Three sections: Customer Insights, Validated/Invalidated Assumptions, Business Model Implications. Each section had bullet points with specific data – the 60% who’d switch to mixed sourcing, the 12% who ranked local sourcing first, the $3 margin per box, the 2.3x increase in sign-up intent with two-tier pricing.
Maya edited it. She added context the LLM couldn’t know – the emotional weight of the local sourcing discovery, the farm relationship risks Charlotte had flagged, the team dynamics around who felt ownership of which initiatives. The LLM produced the skeleton. Maya added the muscle and the nerve endings.
The brief saved the team about two hours of “let me find that number” and “what did the survey say again?” during the prioritisation session. Everyone walked in with the same facts. The arguments were about priorities, not about data.
Later, after the roadmap was set, Maya asked the LLM to draft the board presentation. She pasted in the one-page brief, the Now/Next/Later roadmap, and Charlotte’s financial framework. “Draft a board presentation. Ten slides. Lead with the problem (churn), show the discovery process, present the two-tier model with financial projections, and end with the roadmap. Tone: confident but honest about what’s still uncertain.”
The first draft was competent but bloodless. All the right numbers in all the right places. Clean structure. But it read like a report, not a story. Maya spent an evening rewriting it – not the data, but the narrative. She added Patrick’s kohlrabi story. She added the moment the assumption map showed that local sourcing wasn’t the primary driver. She added Charlotte’s question about farm partnerships that nobody had considered. The numbers told the board what was happening. The stories told them why it mattered.
This is the pattern with LLMs throughout the GreenBox story: fast at structure, weak at narrative. Good at synthesising data, bad at knowing which details carry emotional weight. The LLM saved Maya hours on the draft. Her editing saved the presentation from being forgettable.
Building the Now
The two Now initiatives become sprint goals. Tom and Priya take the pause button. Maya and Jas take the two-tier pricing model. Sam splits his time between the board presentation and customer communication.
The pause button gets Example Mapped on Monday afternoon. Twenty-five minutes, sticky notes on the table. The rules are straightforward – a subscriber can pause for one, two, or four weeks. During the pause, they’re not charged. Their delivery slot is held. They get a reminder email three days before the pause ends. If they don’t respond, the subscription resumes automatically.
The edge cases are where it gets interesting. What if someone pauses the day after their box has already been packed? What if they try to pause and unpause repeatedly to game the system? What if they’re mid-pause and want to extend it? The Example Mapping session surfaces twelve concrete examples and three red cards (unknowns the team needs to resolve before building).
Tom and Priya build the pause feature in six days. The Example Map means they don’t have to guess at the requirements. The red cards were resolved by Wednesday – Lee helped them think through the repeated-pause scenario, and they decided on a simple rule: maximum two pauses per quarter.
The two-tier pricing model is harder. It’s not just an engineering problem – it’s a business model change with implications for farm relationships, marketing, and customer communication. Maya spends two days on the phone with Dave, Rachel, and their third farm partner. She explains the plan. She’s honest about what it means: mixed-sourcing boxes will reduce local farm orders for about 60% of subscribers, but the local box will remain as a premium tier.
Dave is quiet for a long time. Then he asks: “Will the local box subscribers grow?”
Maya doesn’t know. She says so.
“Here’s what I need,” Dave says. “Don’t blindside me. Give me three months’ notice if the local orders are going to drop. I can find other buyers, but I need time.”
It’s a reasonable ask. Maya commits to it. She adds “quarterly farm partner review” to the Later column – it’s not urgent now, but it will be.
Jas designs the pricing page. But first, she presents something she’s been working on privately for the last week.
She’d taken the value prop repositioning – the one Maya moved from Now to Next – and done it anyway. Not as a formal initiative. Not in working hours. She’d spent three evenings at home in Leederville, Moleskine open, laptop beside her, designing mockups that reframe the entire customer-facing experience around the JTBD insight.
She connects her laptop to the office projector without asking anyone’s permission. The screen fills with her work.
The homepage: “Dinner decided.” Mrs Patterson’s words, from the interview Jas sat in on, now a headline in GreenBox’s brand typeface. Below it, not a photo of vegetables – a photo of a family kitchen, a recipe card propped against a cutting board, a half-chopped sweet potato. The message is: we deliver the moment after the decision is made.
The pricing page: two options side by side. “Local Box, $25/week – 100% locally sourced, seasonal produce from farms within fifty kilometres” and “Fresh Box, $20/week – a mix of local and market-fresh produce, same quality, more variety.” The positioning is deliberate. The mixed box isn’t framed as the cheap option. It’s framed as the variety option. The local box isn’t the default. It’s the premium choice for people who specifically value local sourcing.
The about page: not “we source from local farms” (the old lead) but “we take Tuesday night off your plate” (the new one). The farm stories are still there – halfway down the page, for the people who care. But the lead is the job.
Maya stands in front of the projector. She reads every screen twice. Then she says: “This is the first time the website matches what we actually do.”
Jas’s eyes fill. She blinks hard and looks down at her Moleskine. She’s been waiting to hear something like that since week one, when she designed the customisation interface that got thrown away, when Maya redirected the entire product without telling her first, when she sat in her Leederville flat looking at discarded mockups and thinking about quitting. She’d called her mum in Adelaide that night. Her mum had told her about her grandmother’s vegetable garden: “She never grew what she thought people should eat. She grew what they actually wanted.”
Jas had been growing what she thought people should want – beautiful interfaces for browsing produce. Now she’s growing what they actually want: the feeling that dinner is sorted. The napkin sketch from Mrs Patterson’s interview, the one with “dinner decided” underlined twice, is still in her Moleskine. It might be the most important thing she’s ever drawn.
“We can’t ship this yet,” Charlotte says, gently. “Value prop repositioning is Next, not Now. But Jas – save every one of these files. When it’s time, we won’t be starting from zero.”
Jas nods. She closes her laptop. Sam catches her eye across the table and mouths: That was brilliant.
Sam writes the customer email announcing the pricing change. He uses the JTBD insight to frame it: “We’ve been listening. Many of you told us that what matters most is having dinner sorted – knowing what to cook when you get home. Our new Fresh Box gives you that same convenience with more variety, at a lower price. For those who love supporting local farms, the Local Box is still here, unchanged.”
The board meeting
Maya presents to the board on a Thursday afternoon. Three investors on a video call. Charlotte coaches her the night before – not on the content, but on the structure.
“Don’t start with the product,” Charlotte says. “Start with the problem. Then the discovery. Then the insight. Then the plan. The board doesn’t care about pause buttons and pricing tiers. They care about whether you understand your business well enough to grow it.”
Maya starts with the churn number. “We have 200 subscribers and we’re losing 8% monthly. At that rate, we can never reach 1,000. That’s the problem we spent three weeks understanding.”
She walks through the JTBD insight – subscribers hire GreenBox for dinner stress relief, not local produce. She shows the assumption mapping results – 60% would switch to a mixed-sourcing box. She presents the Business Model Canvas with the unit economics exposed: $3 margin per box at current pricing.
Then the plan. The Now/Next/Later roadmap. Two initiatives in Now: pause button (already shipped, churn trending down in week one) and two-tier pricing (launching next week). Three initiatives in Next. Three in Later. A quarterly theme: fix the leaky bucket.
One of the investors, a woman named Diane who’s been on the board since the seed round, leans forward. “This is the first time you’ve presented something that isn’t a feature list. You’re showing me the thinking behind the choices. That matters more than the choices themselves.”
Another investor asks about the farm partnerships. Maya is ready for this – thanks to Charlotte’s question during the BMC session. She explains Dave’s ask for three months’ notice. She describes the risk and the mitigation.
The third investor asks about the metric. “You said churn below 5%. When will you know if the plan is working?”
“Four weeks,” Maya says. “The pause button is live. The pricing change launches next week. We’ll have early data by the next board call.”
The board approves the next tranche of funding. Not because the plan is guaranteed to work – nobody pretends that – but because the plan is coherent. It’s evidence-based. The team knows what they don’t know, and they’ve designed cheap experiments to find out.
Diane stays on the call after the other investors drop off. “Maya, two things. First, the roadmap format is good. Keep using it. Second – the fact that you were willing to present a plan that partially walks away from 100% local sourcing tells me you’re making decisions based on data, not sentiment. That’s what we needed to see.”
Four weeks later
The team reviews. Charlotte facilitates.
The pause button has been live for four weeks. Twenty-three subscribers have used it. Of those, nineteen resumed after their pause ended. Four extended their pause but none cancelled. Before the pause button existed, subscribers in the same situation – going on holiday, feeling overwhelmed, wanting a break – had one option: cancel. Now they have another option. And most of them take it.
Monthly churn dropped from 8% to 5.5%.
Tom puts the number on the whiteboard. The room is quiet for a moment, and then Sam lets out a breath he didn’t know he was holding.
“That’s not below 5% yet,” Maya says.
“No,” Charlotte says. “But the trend line is clear. And you did it with one feature, built in a week, based on a JTBD insight from a set of thirty-minute interviews. That’s the return on discovery work. The insight was worth more than the code.”
The two-tier pricing model has been live for two weeks. Early numbers: fourteen new subscribers chose the Fresh Box ($20 mixed). Six chose the Local Box ($25). Nobody has switched from Local to Fresh – the existing local subscribers are staying put. The new tier is expanding the market, not cannibalising the existing one.
More importantly: of the fourteen new Fresh Box subscribers, Sam checked how many had also looked at Freshly. He couldn’t tell from the data, but he asked five of them in the welcome call. Three said yes – they’d compared the two services and chosen GreenBox because of the recipe cards and the box preview emails. The $20 price point made the comparison close enough that the recipe cards tipped the balance.
“Freshly has better technology and a lower price,” Charlotte says. “You have better curation and a clearer job-to-be-done. The question is which one matters more in six months.”
“It’s early,” she adds. “Two weeks isn’t a trend. But the direction is right.”
Updating the roadmap
Charlotte pulls up the Now/Next/Later roadmap. “This isn’t a document you write once. It’s a document you update every cycle. Let’s look at what moves.”
The pause button is done. It leaves Now. The two-tier pricing model is in pilot – it stays in Now for another cycle while they gather data.
From the Next column, the mixed-sourcing pilot moves to Now. The two-tier pricing launch revealed that the mixed box is popular, but they’re currently sourcing it from a single wholesale supplier. They need two or three suppliers for resilience, and they need to validate that produce quality is consistent. That becomes the next four-week priority.
The value prop repositioning also moves to Now. Sam has bandwidth now that the board presentation is done. He’ll rework the website, the email sequences, and the social media messaging to lead with convenience and stress relief rather than local sourcing.
SEO stays in Next. It’s important, but the team is still five people and they can’t absorb another initiative this cycle.
In the Later column, the B2B offering gets struck through. Not because it’s a bad idea, but because Jas talked to two potential corporate customers and discovered that the logistics of office delivery are completely different from residential. “It’s a different business,” Jas reported. “Different customer, different delivery windows, different box sizes. We’d need to build a separate product.” Later doesn’t mean never – but in this case, it might mean “not us, not now.”
The referral programme moves from Later to Next. The two-tier model makes referrals more attractive – existing subscribers can refer friends to the $20 box, which is an easier sell than the $25 box. Charlotte suggests a simple mechanic: refer a friend, both get a free box. No points, no tiers, no app. Just a discount code.
Second city expansion stays in Later. It’s the right ambition for post-Series A, but the team needs to prove the model works in Perth first.
The updated roadmap looks different from the one they built four weeks ago. That’s the point. A roadmap that doesn’t change isn’t being used. A roadmap that changes every week isn’t stable enough to plan around. The quarterly cadence – review every four weeks, update based on data – strikes the balance.
What Now/Next/Later gives you
The format is deliberately simple. Three columns. No dates beyond the horizons. No percentage-complete bars. No dependencies diagram. No resource allocation spreadsheet.
That simplicity is the point. Detailed plans create an illusion of certainty. A Gantt chart that shows the SEO initiative starting in week 7 and finishing in week 12 implies that you know what week 7 looks like. You don’t. You don’t even know what next week looks like – a farm partner might have a crop failure, a key team member might get sick, the board might change direction.
Now/Next/Later acknowledges uncertainty by design. Now is concrete. Next is directional. Later is aspirational. The further out you look, the less you commit, because the less you know.
Charlotte has seen teams fail both ways. “Some teams plan everything in detail and then feel like failures when reality diverges from the plan. Other teams plan nothing and drift from one urgent thing to the next. Now/Next/Later is the middle ground. Enough structure to be coherent. Enough flexibility to adapt.”
When to use Now/Next/Later
When the team has more ideas than capacity. This is the classic trigger. Discovery work – JTBD, assumption mapping, Business Model Canvas – produces insights faster than a team can act on them. Now/Next/Later forces prioritisation.
When the team argues about what to do first. If five people have five different opinions about the most important initiative, you need a framework for deciding. Scoring initiatives on impact and effort, then sorting into Now/Next/Later, turns an argument into a structured conversation.
When you need to communicate a plan to stakeholders. Investors, boards, partner teams – they all want to know what you’re doing and why. A Now/Next/Later roadmap is easier to explain than a Gantt chart and more honest about what’s certain versus what’s tentative.
After a major discovery cycle. The GreenBox team ran three weeks of discovery – JTBD, assumption mapping, BMC. That produced a pile of insights and initiatives. Now/Next/Later is the tool that turns discovery outputs into a sequenced plan.
When not to use it
When you only have one thing to work on. If the priority is obvious and singular, you don’t need a prioritisation framework. Just do the thing.
When you need detailed scheduling. Now/Next/Later doesn’t tell you that the pause button ships on Thursday and the pricing page goes live on Monday. For that level of detail, you need sprint planning. Now/Next/Later sits above sprint planning – it tells you which initiatives go into which sprints. The sprint itself needs its own planning.
As a substitute for saying no. The Later column can become a graveyard for ideas the team doesn’t want to reject outright. If Later is thirty items long and nobody ever reviews it, it’s not a planning tool – it’s a guilt list. Review Later quarterly. Strike things out. Let go.
What comes next
The quarterly plan works. Churn drops. The two-tier model finds its market. The board is satisfied. Maya secures the Series A.
On the evening after the board call, Maya sits at the kitchen table in Fremantle. Nadia pours her a glass of wine.
“They said yes?” Nadia asks.
“They said yes.”
“Then why do you look like that?”
Maya opens her email drafts. The “pausing operations” email is still there – three sentences, unsent, from the night after the BMC session. She reads it once. Then she closes the draft folder without deleting it. Not yet. She’s superstitious about it, the way some people are superstitious about keeping the receipt until the thing definitely works.
“I look like this because the hard part isn’t over,” Maya says. “It’s changing shape.”
Freshly has ninety subscribers in Perth after one month. Sam tracks the number. It’s small – GreenBox has two hundred and thirty-one. But Freshly’s growth rate is steeper. They’re spending money GreenBox doesn’t have on ads GreenBox can’t match. Dave reported that Rachel got a call from them last week. Rachel told them to get stuffed, but Rachel is one farmer. There are others.
GreenBox raises its funding. The team grows from five to fifteen. Two cities. New subscribers arriving faster than at any point in the company’s history.
And then the problems change.
The codebase that five people understood in their heads becomes a system that fifteen people need to navigate. The architecture that worked at startup scale starts creaking. The substitution algorithm that Tom built – the one with the turnip problem – is now handling three times the volume across two cities with different farm networks and different seasonal patterns. New developers join and don’t know why things are built the way they are. A change in the billing module breaks the delivery scheduler because nobody realised they were coupled.
The techniques from the first two series – understanding the domain, understanding the customer, planning the roadmap – got GreenBox here. But “here” is a different kind of problem. Not “what should we build?” and not “what do customers want?” but “how do we build at scale without the system collapsing under its own weight?”
The next series, Scaling the Machine (coming 4 August), follows GreenBox as it faces the problems that success creates.