What it checks
Ten questions, each scored 1 to 5. The four core ones decide Stop and Pivot.
- Problem *. Is the problem painful, frequent and urgent? Five means a costly, recurring problem with a workaround people already pay for in time or money. One means the problem exists only in the founder’s head.
- Customer *. Who exactly buys, can you reach them, do they have budget? Five means a narrow first segment with a named buyer the founder can reach this month. One means “anyone who”.
- Market and timing. How big is the market and why now? Five means a bottom-up estimate is possible from your text and something changed recently. One means the window has closed or the slice cannot pay one salary.
- Alternatives *. What do people do today and why would they switch? Five means you know the status quo and the competitors, and a customer would repeat your reason to switch. One means an incumbent already does this cheaper.
- Monetization *. Will they pay, how, and do the numbers work? Five means the customer already pays for this category and the arithmetic closes. One means the payer has no reason to pay.
- Distribution. How do the first customers find you? Five means a first channel you already have access to, matched to the price. One means no way to reach the customer at all.
- Feasibility. Can it be built, how fast, and what is risky? Five means a thin first slice is obvious and the technology is proven. One means it depends on something that does not exist.
- Defensibility. What stops a copy from winning? Five means a data loop, a network effect or a locked channel. One means the platform you depend on can ship it as a feature.
- Founder fit. Why you, and do you have the time and runway? Five means you have lived the problem and the team covers build and sell. One means a gap that makes the idea impossible for you.
- Risks. Legal, regulatory, platform and ethical exposure? Five means no serious exposure, or a credible plan for it. One means the core activity is illegal where you sell.
* core rubric
How the verdict works
A fixed rule, printed in every report, so two ideas get the same treatment.
- Stop: three or more core rubrics at 2 or lower, with high confidence.
- Pivot: two or more core rubrics at 2 or lower, with high confidence, and a named pivot.
- Validate first: any rubric at 2 or lower, or a core rubric the text does not settle.
- Build: no rubric at 2 or lower and at least six at 4 or higher.
Sample report
An invented idea, the full output.
Read the sample report
Verdict
Validate first
Do not build yet. Run the experiments below and come back with answers.
Why this verdict. Rule: no red flags, but core rubrics are not rated with high confidence (Monetization). The idea text leaves them open.
Scorecard
| Score | Rubric | Band | Confidence |
|---|---|---|---|
| 4/5 | Problem | Solid | high confidence |
| 5/5 | Customer | Strong | high confidence |
| 3/5 | Market and timing | Unclear | medium confidence |
| 4/5 | Alternatives | Solid | high confidence |
| 4/5 | Monetization | Solid | medium confidence |
| 4/5 | Distribution | Solid | high confidence |
| 4/5 | Feasibility | Solid | high confidence |
| 3/5 | Defensibility | Unclear | medium confidence |
| 4/5 | Founder fit | Solid | high confidence |
| 4/5 | Risks | Solid | medium confidence |
Summary
FieldNote turns a landscaping crew's photos and hours into a same-evening invoice for small Polish owners. The customer, the pain and the first channel are concrete, and the founder has lived the business. The open points are whether owners pay for software at all, how big the national market is, and who builds the app at eight hours a week.
Riskiest assumption. Owners of 2 to 8 person crews will pay 29 EUR a month for faster, photo-backed invoicing instead of keeping the notebook and WhatsApp.
Cheapest experiments (next two weeks)
- Ten calls, three pre-sales (7 days)
Call ten of the forty owners. Describe the same-evening photo invoice and ask for 29 EUR up front for the first month, refundable.
Success metric: 3 of 10 pay before anything ships. - Concierge invoice (14 days)
For two owners, do it by hand for two weeks: they send photos and hours on WhatsApp, you send a finished invoice the same evening.
Success metric: Both owners ask to continue after two weeks and days-to-payment drops versus their last month. - Software spend check (7 days)
In the same calls, ask what each owner pays today for any app, tool or accountant service.
Success metric: At least 6 of 10 already pay for something monthly.
Open questions
- Who builds the app, and how many hours a week can they give it?
- What do the owners you know pay today for any app or accountant tool?
- How many landscaping companies of this size exist in Poland?
- Have you confirmed the e-invoicing timeline for companies of this size?
- Which three apps have the owners you know tried and dropped, and why?
Rubric by rubric
Problem: 4/5, Solid (high confidence)
Is the problem painful, frequent and urgent?
The text names a recurring cost: disputes over work done and late payments, both tied to invoicing a week after the job. That is weekly pain with a cash-flow cost, which matches the 4 anchor. It is not a 5 because the cost is described, not quantified; nothing says how often disputes happen or how much cash sits unpaid.
Evidence from the idea text
invoice a week later, which means arguments about what was done and late payments
write hours in a notebook or a WhatsApp group
Assumptions made
- Assuming a late invoice delays payment by one to three weeks for a crew that bills a few thousand EUR a month.
Questions to answer
- How many invoices a month end in an argument about scope?
- How many days late do customers pay on average?
What would raise the score
- Lead with the cash-flow number: days from job to money, before and after.
Customer: 5/5, Strong (high confidence)
Who exactly buys, can you reach them, do they have budget?
A narrow segment, a named buyer who is also the user, and a concrete way to reach the first forty of them. Owners at this size already pay for phones, fuel and some software, so a monthly fee is within their budgeting. This matches the 5 anchor. The only open point is whether the forty owners are representative of the country.
Evidence from the idea text
owners of small landscaping crews (2 to 8 people) in Poland
I ran a crew like this for six years and still know about 40 owners in my region
Assumptions made
- Assuming the owner both decides and pays, which is normal at this size.
Questions to answer
- Of the 40 owners you know, how many still run crews of 2 to 8?
Market and timing: 3/5, Unclear (medium confidence)
How big is the market and why now?
Bottom-up: assuming 20 000 crews of this size times 29 EUR times 12 months gives a ceiling near 7 million EUR a year, and a reachable 5 percent is about 350 000 EUR a year. That supports a small company, not a venture outcome. Why now is not stated; smartphones on site have been normal for years. The category is stable rather than growing. Score 3: real but not sized by the text.
Evidence from the idea text
small landscaping crews (2 to 8 people) in Poland
29 EUR per month per company
Assumptions made
- Assuming 15 000 to 30 000 landscaping businesses in Poland with 2 to 8 people; the text gives no count.
- Assuming 5 percent of them would adopt any app in two years.
Questions to answer
- How many landscaping companies of this size exist in Poland (GUS data by employment band)?
- What changed recently that makes owners want this now?
What would raise the score
- State the target: a lifestyle business at 300 crews, or a path to adjacent trades (cleaning, roofing) that lifts the ceiling.
Alternatives: 4/5, Solid (high confidence)
What do people do today and why would they switch?
The status quo is named precisely, and the switching reason is concrete: the invoice goes out the same evening with photos instead of a week later. Switching cost is low because the crew already takes photos on phones. The text does not name any competing app, which keeps this from a 5; the founder should check what the owners who already use software use.
Evidence from the idea text
write hours in a notebook or a WhatsApp group and invoice a week later
Notebook, WhatsApp photos, invoice in a spreadsheet template a week later.
Assumptions made
- From general knowledge: field-service tools (Jobber, Workiz) and Polish invoicing apps exist, but target larger teams or only the invoice step.
Questions to answer
- Which three apps have the owners you know tried and dropped, and why?
What would raise the score
- Position against the notebook, not against software: same-evening invoice with proof.
Monetization: 4/5, Solid (medium confidence)
Will they pay, how, and do the numbers work?
Flat per-company pricing fits how a small owner budgets and avoids per-seat friction for a crew that changes size by season. Against the pain (late payments on thousands of EUR), 29 EUR is easy to justify. To replace one founder salary at about 4 000 EUR a month, roughly 140 paying companies are needed. The text shows no evidence that owners pay for software today, which caps this at 4 with medium confidence.
Evidence from the idea text
29 EUR per month per company, flat
Assumptions made
- Assuming a crew bills 5 000 to 15 000 EUR a month, so 29 EUR is under 1 percent of revenue.
- Assuming cost to serve is under 2 EUR a month per company (storage for photos, no model calls).
Questions to answer
- What do the owners you know pay today for any app or accountant tool?
- Would they pay yearly up front for a discount?
What would raise the score
- Test 29 EUR against 49 EUR with the first ten owners; the pain may support more.
Distribution: 4/5, Solid (high confidence)
How do the first customers find you?
A named first channel the founder already has access to, matched to a cheap product: forty phone calls cost nothing. The second channel, a regional trade group, is plausible for the next hundred. What is missing is anything repeatable beyond the region: no search, referral or partner channel is named. Score 4.
Evidence from the idea text
First channel: I will call them, then the regional trade group on Facebook
still know about 40 owners in my region
Assumptions made
- Assuming the Facebook group has a few thousand members and allows posts from members.
Questions to answer
- Where do owners outside your region gather: national groups, suppliers, trade fairs?
What would raise the score
- Build a referral into the product: the owner's customer sees the photo invoice and asks who made it.
Feasibility: 4/5, Solid (high confidence)
Can it be built, how fast, and what is risky?
The first slice is obvious: a timer per job, photo upload, and a PDF invoice sent by email. All of it is proven technology and cheap to build with coding agents. Risks are operational, not technical: Polish invoicing rules (numbering, VAT, the coming e-invoicing system) and offline use on site. Evenings-only pace is the main constraint, so the first version should skip accounting integration. Score 4.
Evidence from the idea text
Stage: idea and a clickable prototype, evenings only
photographs the site before and after, taps start and stop per job
Assumptions made
- Assuming a standard mobile stack with photo upload, a timer and PDF invoice generation; no AI needed for the core.
Questions to answer
- Have you checked what a compliant Polish invoice needs, including the national e-invoicing rules?
- Does the app need to work without signal on site?
What would raise the score
- Ship invoice as PDF by email first; integrate with accounting software only after ten paying owners.
Defensibility: 3/5, Unclear (medium confidence)
What stops a copy from winning?
Nothing stops a copy on day one. What can build over time: the history of photo-backed jobs per customer (a switching cost once a year of records is inside), and trust in a small niche that larger field-service vendors do not serve in Polish. Founder insight helps but is not a moat. Score 3: none yet, plausible later.
Evidence from the idea text
I ran a crew like this for six years
Assumptions made
- Assuming an AI-free product, so the vendor-feature risk is low; a generic field-service tool could still add a Polish invoice template.
Questions to answer
- What will the app know after a year that a new entrant would not?
What would raise the score
- Make the job history useful to the owner's customer too (a shared record), which raises the cost of leaving.
Founder fit: 4/5, Solid (high confidence)
Why you, and do you have the time and runway?
Six years of running the exact customer's business is the kind of insight the market cannot read in a blog post, and the owner network is an unfair advantage for the first sales. Time is thin, eight hours a week, but runway is a year without revenue, which covers validation. The gap is build capacity: the text never says who writes the code. Score 4.
Evidence from the idea text
I ran a crew like this for six years
evenings only, no outside money
Clickable prototype, about 8 hours a week, can go a year without revenue.
Assumptions made
- Assuming the founder can build or has a developer; the text does not say who makes the app.
Questions to answer
- Who builds the app, and how many hours a week can they give it?
What would raise the score
- Decide now whether to pay for a first build or partner with a developer for equity; eight hours a week is enough to sell, not to build and sell.
Risks: 4/5, Solid (medium confidence)
Legal, regulatory, platform and ethical exposure?
Two risks matter. Invoicing is regulated in Poland, and the national e-invoicing system will change what a valid invoice is; the product must follow it or hand off to accounting software. Photos of private gardens and occasionally people raise GDPR questions, manageable with retention limits and consent in the terms. No platform dependency beyond the app stores. Nothing here stops the idea. Score 4.
Evidence from the idea text
the owner gets a ready invoice the same evening
Assumptions made
- Assuming Polish jurisdiction and the national e-invoicing obligation applying to small companies in the coming years.
- Assuming site photos may include people or private property.
Questions to answer
- Have you confirmed the e-invoicing timeline for companies of this size?
What would raise the score
- Delete or blur photos after a set period and say so in the terms.
Critic pass
- Customer: 5 to 5. Score stands. Confidence stays high: the segment, the buyer and the first forty contacts are all in the text.
The scores track the text. The one soft spot is monetization: no evidence yet that these owners pay for any software, so medium confidence there is right.
The idea
FieldNote: a phone app for owners of small landscaping crews (2 to 8 people) in Poland. The crew lead photographs the site before and after, taps start and stop per job, and the owner gets a ready invoice the same evening with photos attached. Today most owners write hours in a notebook or a WhatsApp group and invoice a week later, which means arguments about what was done and late payments. Price: 29 EUR per month per company, flat. I ran a crew like this for six years and still know about 40 owners in my region. First channel: I will call them, then the regional trade group on Facebook. Stage: idea and a clickable prototype, evenings only, no outside money.
Additional answers
- What they use today: Notebook, WhatsApp photos, invoice in a spreadsheet template a week later.
- Stage: Clickable prototype, about 8 hours a week, can go a year without revenue.
Generated with AI on 2026-10-17. AI can make mistakes, so read the scores as a starting point, not the final word. The verdict follows a fixed rule applied to those scores.
Questions
Is it really free?
Yes. The whole report, no email, no account. It runs on a daily budget, so there are three reports a day per visitor and the tool may pause when the budget is spent. The skill version has no limits because it runs on your own tokens.
What model scores the idea?
A current Claude model from Anthropic, called through Cloudflare AI Gateway. Models get swapped when better ones ship. The rubrics and the verdict rule do not change when the model does.
Why is the verdict a rule and not the model’s opinion?
So that it can be checked. The model’s judgment is in the ten scores and the reasoning under each one. The verdict is arithmetic over the scores, printed with the rubrics that decided it.
Does it search the web?
No. It reads your text. Competitors and market numbers come from the model’s general knowledge and are marked as assumptions, so check them.
Who sees my idea?
The text goes to Anthropic’s Claude through Cloudflare AI Gateway and the report is stored under a random link for 180 days. I do not read reports. You can delete yours from the report page.
Can I run it on many ideas?
Three a day here. For more, install the skill and run it in Claude Code or any agent that reads Agent Skills.
Method and limits
Ten rubric passes, each a separate model call with its rubric file and your text, returning scores and quotes in a fixed JSON shape. A second pass tries to refute the extreme scores. A synthesis pass writes the summary, the riskiest assumption and the experiments. The verdict is computed in code from the scores. This is a structured opinion from a language model, not business, legal or financial advice. It can be wrong, especially about competitors and market sizes it cannot look up.
Privacy
Your idea and the report are stored in a key-value store under a random id for 180 days, then deleted. The report page is not indexed by search engines. If you ask for the report by email, the address is used once to send the link and a copy of the request goes to me, so I can follow up if you want to talk; the address is not stored with the report. Analytics: PostHog EU through a first-party proxy, no cookies, session replay with all inputs masked.
Run it in your own agent
The same rubrics, rule and renderer ship as an open-source Agent Skill. Install it in Claude Code or any agent that reads skills, no limits, your tokens. Skill on GitHub.
