What a funder is paying for

Whether you call it an accelerator or an incubator, we think a funder weighing a renewal is asking one question: what changed in the teams it paid for?

In our experience, programmes drift toward reporting activity because activity is easy to count: sessions held, mentors who turned up, introductions made, workshops run. And when a programme measures activity, founders learn to produce activity. We built the stages below so that each one leaves evidence of change behind it.

The cycle at a glance

1. Selection: Score evidence before people

We think selection is half of what a programme can achieve. Choose well and you are halfway to results. Choose badly and, in our view, no curriculum or mentor bench makes up for it, for your portfolio or for the ecosystem you report to.

In our experience, the usual method hands experienced mentors and investors a spreadsheet and asks them to rank the applicants. Each of them reads the same sheet through their own lens, and their biases and shortcuts end up in the cohort. We would put a standard assessment first, answered by each co-founder separately, and let the panel's judgement sit on top of it.

The scorecard we would use, with written anchors and six steps for running the panel, is in How to score accelerator applications so a second panel would pick the same startups. One step from it belongs here as well: write one sentence on what the programme is selecting for before the call opens. That sentence is also the first line of your funder report, the reason these teams got the seats.

2. Intake: Say what you can give and what you cannot

In programmes we have run, good founders, the ones we were glad to select, have started dropping away in the middle. In our experience the cause is expectations. A programme promises funding, networks, space and mentors. When what a founder finds falls short of what they expected, they stop showing up, and it hits hardest in programmes that give no investment.

So we would start with an honest conversation, and hold it before the offer. Ask each team what it needs from the programme and take the answer at face value, because what a founder says they need is usually what they need. Then write down what the programme can give, what it cannot, and when. A gap you name at intake can still be talked through. Found in the middle of the programme, it tends to end with the founder gone.

Intake is also where you learn what each team brings in. Melody Chang and Valentina Assenova studied 6,723 startups and their founding teams for a paper in the Strategic Management Journal, published online in October 2025. They found that founders with substantial pre-entry knowledge (the education, industry experience and entrepreneurial exposure a team builds up before it launches) tended to gain more from accelerators than novices did, and their findings suggested that the returns depended on how well that knowledge matched the programme's design.

We would ask about all three at intake. Use the answers to decide which lane a team starts in; whether it gets in is the selection panel's decision. One cohort can run a structured foundation lane for first-time teams and a specialist lane, heavy on access to industry mentors, for teams that know their market. We set out how to run both in One cohort, two lanes.

3. Mentors: A small bench with one lead per startup

We would keep the active mentor list to 10 to 15 people per cohort. Ask each mentor for at least three structured sessions, with a short written summary after each one, and take off the list anyone who does not meet that. Mentors who have been on your list for years may not make the cut.

Then give every startup one lead mentor whose job is continuity, and bring specialists in when a specific decision needs them. Techstars has described a similar split. Its Mentor Manifesto, published in 2020 and still on its site, divides its mentors into lead mentors, who spend more time with companies, and mentors, who help out where they can. The same page asks mentors to commit clearly to mentoring or to say no.

Match mentors to what each team needs. A patient generalist can carry a first-time team a long way, while a team that knows its industry needs a mentor who knows it at least as well.

Ask every mentor to leave one written output after each session: one risk they see, one bet they would place and one next step they would test. Those notes, filed against each team, are what a funder can read later. A count of mentor hours shows that sessions happened and says nothing about what they changed.

4. The middle of the cohort: Catch the drift early

In our experience the middle is where the mismatch from intake comes out, and where good founders quietly disappear. Two light habits catch it before that.

The first is a weekly learning note that fits on one screen. Each team answers four prompts: what did we learn this week that changed our mind, what evidence supports it, what decision did we make because of it, and what is the next test. Teams producing activity with no evidence behind it stand out quickly.

The second is a midpoint review. Run the same assessment the teams took at intake, then go through the lanes with the mentors in the room and move any team that started in the wrong one.

When a team falls behind, we separate a founder who is struggling from a founder who has stopped trying, because each needs a different response. The struggling founder gets one honest conversation, with direct feedback delivered so they can act on it, and help that stops where the founder's own accountability begins. For a founder who has stopped trying, decide whether to keep spending programme resources on the team, tell them plainly, and treat them well even as the programme steps back. We covered both cases in the first episode of Connectors Beacon.

5. Measuring change: The same instrument at the start and the end

Run the same assessment at intake and at the end, and compare each team with itself. That difference is the centre of your impact story.

HM Treasury's Magenta Book, the UK government's central guidance on evaluation (updated in May 2026), explains why the start matters. It says baseline data has to be collected early, that comparison data from groups the intervention did not reach has to be negotiated, and that without that planning an evaluation may be impossible, severely limited or unnecessarily expensive. Its account of impact asks what would have happened without the intervention, and how far the observed changes can be put down to it.

For a programme, that means two things. Take the baseline before week one, because a baseline taken in week four already contains three weeks of the programme. And say plainly in the report what a before-and-after shows. It shows what changed in each team. To say how much of that change the programme caused, you need a comparison, and the time to plan one is before the cohort starts.

Compare by lane as well as by cohort. A cohort average hides the gap between teams that arrived experienced and teams that arrived new. Your funder learns more from "first-time teams moved this far, experienced teams moved that far" than from one number.

6. Reporting: Movement, patterns and what happened after

We would build the funder report from three parts.

1. Movement on evidence. Show how teams moved from unknown to tested on their core assumptions, next to the change in their assessment scores.
2. Patterns across the cohort. The false assumptions teams started with, what surprised them, and where they kept running into the same friction. That is useful to the funder and to your next cohort.
3. What happened after. Short check-ins at month 6 and month 18 that ask whether the company is still active, whether it has revenue or meaningful usage, what its runway is, whether it has raised money and what kind, and what its next constraint is. In our experience founders answer when the form takes three minutes and the programme still feels human.

We set out this scoreboard in the second part of our piece on accelerator KPIs.

Where we fit

The programme edition of our startup assessment runs the assessments in stages 1, 2, 4 and 5. Each programme gets its own instance with its name on the reports. Every co-founder answers separately, in about 90 minutes, and can save and come back. Your team sees scores in four dimensions, the overall level, team alignment and the co-founders' answers side by side, compares up to five evaluations, and exports CSV, XLSX, PDF and DOCX.

For EU-funded and government-funded programmes it runs one fixed assessment for every applicant, the same assessment again mid-programme and at the end, and gives you the exports your report to the funder draws on. It does not replace your application tool, sends nothing to the funder and does not write the report. The question set is fixed, which keeps results comparable across applicants, cohorts and years. It starts from 3,000 EUR in the first year for up to 25 startup teams, with renewal from 1,500 EUR a year, VAT included.

For stages 3 and 4, our mentors and cohort training are designed to complement the programme you already run: 60-minute one-to-one sessions that founders book through a link your team shares, online or in person, and 90 to 120 minute workshops on customer discovery and product-market fit, fundraising and growth, founder resilience, and AI and automation. Both are quoted per programme.

If you are not ready to talk yet, the free cohort evaluation check asks six questions in two minutes on whether your scoring would hold up across judges and cycles. For EU-funded and publicly funded programmes, the free selection accountability check looks at your judging process and your follow-up after the programme the way a sponsor would.

Why two panels pick two different cohorts

We think selection is half of what a programme can achieve. Select well and you are halfway to results. Select badly and, in our experience, no curriculum or mentor bench produces them, either for your portfolio or for the ecosystem you report to.

The usual method looks sensible. You invite mentors and investors who have done this for years, give them a spreadsheet, ask them to rank the applicants, and take the top ten. They all read the same sheet. But each of them reads it through their own lens: the sector they know, or the founder who reminds them of someone. In our experience that is where the bias comes in, along with shortcuts and misreadings, and it cuts both ways. A false positive takes a seat from a team that would have used it. A false negative sends home the founder you should have backed.

Research on accelerators shows how much weight the panel carries. Berna Beyhan, Semih Akçomak and Dilek Cetindamar studied ten accelerators in Ankara and Istanbul, with interviews done between March 2016 and January 2017, and published the work in the Entrepreneurship Research Journal in 2021. The authors describe a staged process in which, at least in the second stage, a jury or committee decided; their own table shows one of the ten choosing in a single stage. One of them used a jury of 40 people, including professionals from corporations, investors, entrepreneurs and academics. The authors found that the programmes tended to select teams that were coachable, willing to collaborate with the accelerator, mentors and others, and passionate enough to cope with the pressure of building fast.

Those are the qualities a ranking spreadsheet is worst at seeing. A judge can feel that a founder is coachable in a ten-minute pitch. But a programme only knows it by watching what the founder does with advice.

Daniel Kahneman, Dan Lovallo and Olivier Sibony named the wider problem in MIT Sloan Management Review in 2019. One of their opening examples of an evaluative judgement was this: "A venture capital investment committee chooses among an array of startups to fund." They separated errors that come from known biases from random errors, sometimes called noise, and drew on research into hiring and job interviews for a structured approach they called the Mediating Assessments Protocol. Choosing a cohort is the same kind of judgement, so it should respond to the same kind of structure.

The scorecard

Five criteria. The first four are the four dimensions of the assessment we built for programmes, so the evidence for them can come from founders' own written answers. The fifth is the one panels guess at.

Weights are the programme's call, and they'll follow the programme's job. A pre-accelerator that exists to create first revenue reads the second row differently from an investment-led programme that has to show follow-on funding. Write the weights down before applications open, while nobody has met the founders yet.

How to run the panel with it

1. Write one sentence on what the programme is selecting for, before the call opens. Every judge gets it with the scorecard.
2. Collect the evidence before anyone pitches. Each co-founder answers the same questions separately; in the assessment we built that takes about 90 minutes per founder, and they can save and come back. Differences between co-founders are part of the evidence, so keep them visible.
3. Judges score alone, criterion by criterion, against the anchors, before any discussion. No one sees anyone else's scores until all are in.
4. For each startup, look at the spread: the highest judge score minus the lowest. Where judges agree, the discussion can be short. Spend the panel's time on the startups with the widest spread, and ask each judge which evidence moved their score.
5. Between the first and second stage, give each shortlisted team one small piece of homework built from their weakest row. Score what comes back. That turns coachability from a feeling into evidence.
6. Keep every score. At the end of the programme, run the same instrument again.

Step 6 is what you show your funder. The same scorecard at intake and at the end shows what changed in each team, and over two or three cycles it shows whether your selection picked the teams that grew.

Before you change anything

Test the scoring you run today. If a different panel judged the same applicants next month, would they pick the same cohort? Our cohort scoring check answers that in six questions and two minutes, and it is free. For EU-funded and publicly funded programmes, the selection accountability check looks at your process the way a sponsor would.

Where we fit

The programme edition of the startup assessment we built runs steps 2 and 6 for every startup: each co-founder answers on their own, and your team gets readiness scores on four dimensions, team alignment and flagged answers, with up to five evaluations compared side by side. It starts from 3,000 EUR for the first year for up to 25 startup teams, VAT included. Our mentors and cohort workshops can then work on the weakest rows inside the programme you already run.