Every accelerator on the continent runs the same conversation at intake, and almost none of them run it on evidence. The debate is familiar: open the doors wide because access is the point, or narrow them because scarce seats deserve serious founders, and both sides argue from conviction because neither has looked at what happened to the people they already admitted. We looked. Across 99 alumni of one Ugandan programme, tracked from entry through completion, the business a founder held on the day they walked in predicted whether they finished with a force that no interview impression came close to matching: founders arriving with a registered, trading business dropped out at 15 percent, founders with an informal business at 25 percent, and founders arriving with no business at all at 45 percent. Ten of the programme’s twenty-four total drop-outs came from the twenty-two who arrived with nothing trading. That is not a moral finding and it should not be read as one. It is a completion probability, and a programme that ignores it is not being generous. It is selecting people into a failure it could have predicted.
Key Takeaways
- Business status at entry is the strongest completion predictor in the dataset: 15 percent drop-out for registered ventures, 25 percent informal, 45 percent no business.
- The finding is about probability, not worth. Ten of twenty-four drop-outs came from a group of twenty-two, which is a design signal rather than a judgement.
- Programmes that admit first-come-first-served are allocating scarce paid places by typing speed, and typing speed correlates with nothing.
- Ability to pay and venture substance are separable criteria, and conflating them loses exactly the founders scholarships exist for.
- The honest response to a low-scoring applicant is a different product, not a rejection: a lighter tier costs a tenth as much and matches the stage.
- Selection criteria that cannot be shown to a rejected founder are not criteria. They are preferences with a spreadsheet.
Why does business status predict completion so strongly?
Because a four-month evening programme is not a course, it is an obligation layered on top of an existing life, and an existing business supplies three things that carry a founder through the hard middle. It supplies a reason: the module on unit economics is abstract to someone with no units and urgent to someone whose margin is wrong this month. It supplies a rhythm: a trading founder already organises weeks around customers, so a Tuesday evening class competes with other commitments rather than creating the founder’s only one. And it supplies cash, which matters less as a fee question than as an attention question, because the founder who is job-hunting in month two is not absent through weakness but through arithmetic.
The founder arriving with an idea has none of those. Their enthusiasm at intake is genuine and often higher than the trading founder’s, which is exactly why interviews mislead. Enthusiasm at week one is uncorrelated with attendance at week eleven, and the dataset says so plainly: the group that looked most excited on arrival is the group that halved.
What does this mean for how places are allocated?
That first-come-first-served is indefensible, and that ability to pay should be heavy without being decisive. When a programme awards a place the moment someone completes a form, it is allocating scarce, subsidised, paid seats by response speed. Response speed measures phone access and free time. It does not measure whether a founder will be in the room in November.
The workable alternative is a weighted rubric, published internally, where ability to pay carries real weight because a place funded by nobody is a place that empties, venture substance carries comparable weight because the completion evidence demands it, and the contribution a founder makes to the cohort counts as its own line. A founder who employs four people and trains apprentices brings something the room consumes; that is worth scoring. Faith integration, for a Kingdom programme, functions as a screen rather than a ranker, because nearly every applicant passes it and a criterion everyone passes cannot separate anyone.
The design test is simple. If a founder who pays cash and brings nothing else can score their way in, the rubric is a payment gate wearing a costume. Money should buy a hearing, not a seat.
How do you avoid losing the founders scholarships exist for?
By scoring ability to pay and venture substance separately, then reading them together. The instinct when budgets are tight is to filter on payment first and rank whatever survives, which feels prudent and quietly destroys the programme’s best argument for existing. In our own review the highest-community-contribution applicant in the entire pool, a registered and trading agro-processing venture with a clear jobs story, had written a careful request for a full scholarship. A payment-first filter removes her at step one. A weighted rubric places her fifth.
That is the whole case for scoring rather than filtering. A scholarship budget spent on the strongest founder who cannot pay is the most productive subsidy a programme can make, and it is only findable if the process rates her venture before it reads her wallet. Filters lose her silently. Rubrics surface her by name.
The second half of the answer is having a real product for the founders who score low. A rejection from a flagship programme is a dead end. An offer of a lighter tier at a tenth of the fee, matched honestly to the stage they are at, is a route. The founder with an idea and no trading history is not being turned away, they are being told the truth the completion data already told us: the intensive four-month version is not where they convert, and proof of one paid customer is the qualification that changes their odds.
FAQ
Is this saying founders without businesses should not be supported?
No. It says the four-month intensive is the wrong instrument for them, and that admitting them into it at a 45 percent drop-out rate serves nobody. Support them with a lighter product and readmit them when something trades.
Does 99 alumni justify a rule?
It justifies a weighting, not a law. Ninety-nine is enough to see a three-fold gap that holds across four categories, and small enough that any single cohort could shift it. Treat it as the best available evidence, revisited annually.
Should the criteria be published to applicants?
The weights should be defensible to a rejected founder even if they are not printed on the form. A criterion you would not say out loud to the person it excluded is not a criterion you should be using.
What about founders who register a business just to qualify?
The signal is trading, not paperwork. In the data, registered-and-trading and registered-in-programme both outperform; a shell registered the week before intake carries none of the rhythm or reason that produced the effect.
Related Reading
- Selection versus treatment: what accelerators actually change
- Dignity has a price architecture: tuition, subsidies and scholarships without shame
- The missing middle is a management problem, not a capital problem
- What we learned charging African founders instead of giving acceleration away
Sources and Evidence
- Internal alumni outcomes dataset, 99 participants, 2022 to 2026, recording business status at entry against programme completion. Categories: registered business pre-programme (n=41), informal business (n=32), no business (n=22), registered in-programme (n=4).
- Clevenger, M. R. and Fortunato, M. W-P, Entrepreneurial Communities and Ecosystems, on the community conditions that sustain founder participation over time.
- Blacksmith, N. and McCusker, M. E., Data-Driven Decision Making in Entrepreneurship: Tools for Maximizing Human Capital, on selection science and the limits of interview-based judgement.