Prove the business, not just the product: evidence that convinces assessors and investors

Assessors and investors have read a thousand applications that say customers are excited. This guide replaces excitement with evidence in about twelve weeks: desk research you can cite, interviews that produce facts, letters that count, priced pilots, and where each funder lets you show them.

The reality: products with fans still fail

Steve Blank, 2009: "Startups don't fail because they lack a product; they fail because they lack customers and a profitable business model."⁶ CB Insights' reasons for 385 VC-backed shutdowns since 2023 (March 2026; causes overlap)¹:

Budget time. ICURe Explore gives research teams 12 weeks full time and up to £35,000 to test value propositions in the market⁵. Ash Maurya's "typical timeline", with his caveat that "Every startup moves at its own pace"⁸:

What a minimum viable business is, and who said so

Nobody owns the phrase. The nearest named idea is Bill Aulet's Step 22, "Define the Minimum Viable Business Product (MVBP)", then Step 23, "Show That 'The Dogs Will Eat the Dog Food'"¹⁰. In a published summary (not Aulet's words), the MVBP passes three tests: the customer gets value, the customer pays, and feedback starts to flow; an offering is not a product until a customer has bought it¹¹.

You have a minimum viable business when all five are true
1. SEGMENT  You can name the beachhead in one sentence, with a customer count.
2. PAYMENT  Members of it pay, or have signed to pay, rather than praise.
3. RETURN   They keep using it: retention measured, not assumed.
4. ROUTE    At least one channel wins them at a cost the margin covers.
5. RUNWAY   You are default alive, or have a dated plan to be.
Fewer than five: you have a product with fans. Say so, and show the plan to five.

That is this guide's working definition, assembled from the people below.

How funders let you prove it: the evidence rules

Week 1: research before you build

Free sources cover "how many" through "who, by name". Quote the year and publisher with every figure; Innovate UK wants market size "backed up by references where available"²⁹.

Innovate UK asks whether you have "identified any similar innovation and its current limitations, including those close to market or in development"; where references "are not offered or are not relevant" the answer scores 3 to 4²⁵. "No competitors" fails, and Business Connect names the strongest rival: "It is often preferable to stick with the status quo even if a better solution exists, because it is a known pain versus an unknown potential gain."²⁷

Keep the map honest: draw positioning from customers' stated perceptions, since a managers' map is "limited by not being based on consumer data"⁴⁶; mystery-shop rivals for real prices, posing "as a customer, prospect or user"⁴⁷; write a win/loss note after every lost pilot or sale.

Competitor matrix (columns = the 4 to 6 attributes customers named in interviews)
Row                                  | Price | Time to value | Accuracy | Integration
Do nothing / spreadsheet / manual    | £0    | none          |          |
Incumbent A (vendor page or quote, date)
Adjacent product B
Research-stage rival (patent number, filing date)
Us, today
Us, after the project
Mark honestly where you are worse. Cite the page or mystery-shop quote and its date.

Pace it on ICURe Explore. Its closing Options Roundabout recommended these routes for 641 projects, 2014 to 2024⁴:

For counts, the Mom Test says "three to five meetings are sometimes enough" and to stop when you stop learning⁴⁸; Moesta's range is in the table above.

A 12-week research-first sprint (modelled on ICURe Explore; stretch it if part time)
Wk 1-2   Desk: ONS/Nomis counts by SIC and size; Companies House target list;
         patent search; competitor matrix v1. Output: counts with sources and years.
Wk 2-10  Interviews, at least weekly: 10 to 20 per segment, 30 to 60 in all for a
         B2B product with two or three segments, half recent switchers, half
         current sufferers. Output: the log sheet, quantified pain, verbatims.
Wk 4-8   One or two pretotypes (fake door, concierge). Output: ILI with its
         denominator; OLI by week.
Wk 8-12  Convert the strongest conversations: dated LOIs with volume, price band
         and timing, or a paid pilot sheet. Ask everyone the four price questions.
Wk 12    Write it up as counts, ratios, dates and quotes. That is your evidence pack.

Weeks 2 to 10: interviews that produce truth

Rob Fitzpatrick's Mom Test rules: "Talk about their life instead of your idea"; ask about specific past events, not opinions about the future; listen more than you talk⁴⁸. Compliments are "fool's gold"; "I would", "I might" and "I usually" are fluff.

Discovery interview script (30 minutes; two of you, one asks, one writes)
Open     "Tell me about the last time [problem] came up. What happened next?"
Past     "How often in the last month? Who else was involved? What did it cost
         you in hours or money?"  (log numbers, not adjectives)
Attempts "What have you tried? What did that cost? Why did you stop?"
Forces   (recent switchers) "What pushed you to look? What pulled you to the new
         thing? What worried you? What did you give up?"
Buyer    "If this were fixed, who would sign it off? Which budget does it come from?"
Price    Four questions, each answered with a number: "At what price would this be
         too expensive to consider? Expensive but still worth considering?
         A bargain? So cheap you would doubt the quality?"
Close    Ask for one commitment: a diary slot with a goal, an introduction to the
         budget holder, a trial with a start date, a letter of intent, a deposit.
         Record what they said yes to, and what they declined.
Never    Pitch. Describe your idea only if asked, and only after the past questions.

Interview log (one row per conversation; the application quotes it as counts)
Date | Segment | Role | Org size | Last occurrence | Times per month
| Cost now (hours or £) | Current workaround | Price band (four answers)
| Commitment asked | Commitment given (time / reputation / money) | Verbatim
Summary line for the form: "n of N [role] in [segment] [behaviour], median [cost];
k gave [commitment] (dates)."

End every meeting by asking for a commitment; its currency says what it is worth⁴⁸.

When to narrow the segment
Ten interviews give ten different problems      -> split by role, size or trigger
Very-disappointed score under 40%               -> keep only the personas who are
                                                   very disappointed, re-score
Warm introductions convert worse than 1 in 20   -> wrong buyer or wrong promise
Pilots succeed but do not renew                 -> value went to the user, not
                                                   the economic buyer

Narrowing is not retreat: Superhuman's score rose from 22% to 33% by keeping only the personas found among the very disappointed, before any product change¹⁶.

The evidence ladder: lead with your highest rung

Nine rungs, weakest first; sources in the table below:

UKRI project partner letters are not on this ladder: they evidence delivery capacity and co-investment, not demand²⁸.

Letter of commitment (fits UKRI's rules: two sides of A4 at most, no headed paper
or wet signature needed, one PDF for all partners; keep the same content for the
EIC's Letters of Intent)
[Organisation], [date]
Re: [project title], [lead organisation], [funder and competition]
1. Who we are and our role: [named role: pilot site / data partner / first customer]
2. The problem as we experience it: [hours, £ or incidents per month, from our records]
3. What we will do: [activities, named staff, dates, site access, data]
4. What it is worth: cash £[x]; in-kind £[y] ([days × day rate], equipment, data),
   as entered on the form
5. If the project meets [success threshold]: [purchase order at £z / renewal /
   rollout to n sites / reference]
6. Why this matters to us: [value, relevance and benefit, in our words]
Signed: [name, job title with budget authority], [email, phone]

Weeks 8 to 12: pilots and prices that turn "say" into "do"

Pilot design sheet (agree in writing before day one)
Segment and site     [role, organisation, size]; one site, one named decision-maker
Baseline             [current cost or time, from their records, dated]
Success threshold    [metric] from [baseline] to [target] by week [n]
Duration             [weeks]; review dates [d1, d2]; end date fixed
Price                £[x] for the pilot (a token fee still counts), or free with a
                     signed next step
Decision at the end  Threshold met: [purchase order at £y / annual licence / rollout
                     to n sites], signed now, conditional on the result
                     Threshold missed: what we both learn; no renewal by default
Report               Against this sheet, with the numbers and the customer's words

A free, open-ended pilot is rung 5. A paid pilot with a conversion clause is rung 6: even a token fee moves the evidence from stated intention to observed behaviour⁴⁹. Either way, early traction may come "from the subset of the market that are early adopters and not the actual market itself"², so say which segment piloted.

Public buyers can pay for the pilot: Contracts for Innovation fund development as a contract, and the battery phase 1 call offers "a share of £2 million, inclusive of VAT" (closes 4 November 2026)⁵². See guides/innovation-procurement-and-nhs.

Price in three steps. Van Westendorp's four questions (1976, in the script above) give a range; the method assumes respondents can picture prices and does not estimate demand, so treat the range as a hypothesis⁵¹. A pre-order, deposit or paid pilot fixes a point; recurring payment confirms it. Simon-Kucher: "72% of innovations fail to meet their financial targets" or fail entirely³. Sean Ellis would have most startups charge only once they have fit, but enterprise sooner, because a paying customer has "somebody on the inside" who is "motivated to work with you"¹⁵.

Strategyzer's rule for the Business Model Canvas: "Treat it as a claim, not a plan"; "Date and version the canvas, then redraw it as evidence comes in."⁵⁰ Label each claim with its evidence tier: strong (observed actions), moderate (stated intentions) or weak (opinions)⁴⁹.

Give a pre-revenue assessor a milestone to score: a baseline, an end-of-project target and the decision if you miss it, using metrics the sources name: ILI and OLI²⁰, time to complete a sale¹³, warm-introduction conversion and regretted churn⁵³, months default alive¹⁸.

Innovation accounting a grant assessor can read (pre-revenue)
Metric                        Baseline (date)      Target, month 12    If missed
Interest ratio (ILI)          32 of 410 visitors   10% of a cold list  change the promise
Interviews, quantified pain   24 of 31             n/a                 n/a
Warm-intro conversion         1 in 14              1 in 10             change the buyer
Time to complete a sale       n/a                  under 90 days       change the channel
Pilot threshold               5.5 h/week           under 3.3 h/week    pivot or stop
Renewals / regretted churn    none yet             2 of 3 renew        fix onboarding
Months default alive          9                    18 (with grant)     cut hires
The numbers are placeholders. Replace every one with yours, dated.

Traction benchmarks, with their labels on

These are US software benchmarks. Borrow the shapes (a few payers before scaling spend; retention by cohort; acquisition cost against margin; honest runway), not the dollar figures.

ARR is annual recurring revenue, NRR net revenue retention, CAC customer acquisition cost; burn multiple is net burn divided by net new ARR⁵⁵.

Paul Graham's Y Combinator growth bands (September 2012): 5 to 7% a week is good, 10% exceptional, 1% suggests the founders "haven't yet figured out what they're doing"⁵⁶. Sequoia's Arc framework (April 2024) explains why traction differs by market: "Hair on Fire" problems reward the best product fast, "Hard Fact" problems need market education first, "Future Vision" companies need commercial "pit stops"⁵⁷. Name your archetype so the reader judges your numbers in context.

Put each rung where the reader looks

Question numbers differ between competitions; the headings do not. A hypothetical rewrite, with invented numbers:

Examples: what the evidence looked like

Pitfalls

• Compliments logged as evidence. They are "fool's gold"; count commitments instead⁴⁸.
• Prices taken from "How much would you pay?". Hypothetical pricing is weak-tier evidence: use the four-question range, then a paid point.
• Letters sent where they cannot count. Non-partner letters are ignored at UKRI²⁸; Innovate UK's briefs do not invite them²⁹.
• LOIs without volume, price band or timing. Make each one specific or leave it out.
• Pilots with no baseline, threshold or end decision. They produce anecdotes, not rungs.
• Early-adopter traction read as the market². Name the segment every number came from.
• Scaling before fit. 70% of startups did², and "Hiring too fast is by far the biggest killer of startups that raise money"¹⁸.
• "No competitors". Innovate UK scores your grasp of the nearest similar innovation, and "do nothing" is always a rival²⁵²⁷.
• Different numbers in different answers, or cumulative charts.

Checklist: the evidence pack

• Beachhead segment named in one sentence, with a customer count from ONS, Nomis or Companies House and its year
• Competitor matrix includes "do nothing", an adjacent product and a research-stage rival from a patent search
• Interview log: at least 10 per segment, with counts, quantified pain, dates and verbatims
• Every interview ended with a commitment asked for, and the answer recorded
• One pretotype run, with ILI and its denominator
• Price range from the four questions, with the respondent count
• At least one signed letter of intent or commitment stating volume, price band, timing and the signatory's role
• Pilot design sheet agreed in writing: baseline, threshold, price, end decision
• Highest rung placed first in each answer, lower rungs as support
• This call's letter rules checked: partners only (UKRI), none invited (Innovate UK), LOIs attached (EIC)
• Innovation-accounting table: baseline, target, decision rule, dated
• Runway and months default alive stated honestly
• Same figures in the need, market, route-to-market, finance and risk answers
• Evidence pack filed for due diligence: logs, letters, pilot sheets, survey exports

Questions

What is the difference between a minimum viable product and a minimum viable business?

An MVP is the smallest product that teaches you something about customers. A minimum viable business exists when a named segment pays, keeps using the product and can be won through a channel whose cost the margin covers: the summary of Aulet's tests: the customer gets value, pays, and feedback flows. Assessors and investors score the second.

How many customer interviews do I need before I apply for a grant?

Bob Moesta suggests about ten, or 15 to 20 if you are new to interviewing; the Mom Test says three to five are sometimes enough and to stop when you stop learning. For a B2B product with two or three segments, aim for 30 to 60 over eight weeks, reported as counts: "24 of 31 practice managers".

Do letters of support help an Innovate UK application?

The three live Innovate UK briefs checked for this guide (October 2026) do not mention letters of support and allow appendices only on named questions (two A4 pages, PDF, 10MB). Put customer evidence inside the answers as names, numbers and dates. In Investor Partnerships, an investor's term sheet or indicative offer counts as traction.

What counts as traction for a pre-revenue company?

Behaviour, in writing, with dates: pretotype interest ratios, signed letters of intent with volume and price, paid pilots, pre-orders or deposits, and an investor's written pre-commitment. First Round's Nascent level starts at three to five customers; Sam Altman's bar for B2B is ten paying customers.

Should my first pilot be free or paid?

Paid if at all possible, even at a token fee, with a conversion clause: it turns stated intention into observed behaviour and meets the test that the customer pays. A free pilot still counts if it has a written baseline, a success threshold, an end date and a signed next step.

Does the EIC Accelerator need letters of intent?

Yes. The 2026 guide lists "Letters of Intent" among the full-proposal documents, and some Challenges require one from an end-user or regulator. An investor pre-commitment, non-binding, must come from a single qualified investor for at least 20% of the target round.

How do I prove demand if I have no customers yet?

Climb the ladder: sourced counts of people with the problem, interview findings with quantified pain, a pretotype people clicked or signed up to, then letters of intent and a paid pilot. Lead with the highest rung you reach.

Written for businesses, founders and researchers in the North West of England and checked against the sources listed above on 18 October 2026. Funders change their rules, so always read the current call before you apply. Spotted something out of date? Tell us.