The Counterfactual: Selection, Signaling, and Fit
The hardest question about a startup accelerator is not whether its graduates succeed. It is what those ventures would likely have done without the program, given the resources, founders, market, and ecosystem they already had. This chapter turns that question into a counterfactual and fit diagnosis. The aim is not to calculate an impossible exact share for “program value,” but to separate the routes that can produce an observed advantage and improve the quality of a real decision.
A counterfactual is not an abstract comparison with a nonexistent world. It is the best realistic alternative available to this venture at this time: raising money directly, finding a specialist mentor, joining another network, running a smaller experiment, delaying a decision, or doing nothing. If the alternative is never specified, “the accelerator worked” can mean only that the venture did better after joining than it did before joining. Before-and-after improvement is not enough because the market, the founder, and the venture would have changed even without participation.
Three routes to one observed success
Created capability
The program creates value when it changes what the venture can do relative to its realistic alternative. A mentor may help the team design a better customer test. A cohort may reveal a distribution problem before the venture spends its runway. Funding may make it possible to run an experiment that would otherwise be delayed. A network may provide access to a buyer or specialist who would not otherwise respond. The common feature is an internal or relational change that remains useful beyond the moment of contact.
This route requires more than an input. It requires a chain from resource to action to outcome. If the venture receives a mentor introduction but does not learn anything, no capability has been created through that interaction. If it learns a useful method but cannot implement it because the team lacks time or technical capacity, the program may have supplied knowledge without producing a changed venture. Evidence of value should therefore be strongest when it shows a plausible change in capability, not only a later event such as fundraising.
Selected quality
The program can also be associated with success because its admissions process selects ventures that had stronger prospects from the start. Prior funding, a capable founder, a promising market, a supportive location, or a more developed product can all make later performance more likely. A selective accelerator may identify those attributes without causing them.
The meta-analysis raises this concern by noting that many programs favor ventures that have already obtained funding. Yet its subgroup comparison finds approximately the same positive effect for ventures with and without prior funding—r = 0.103 in both groups, with no significant difference. These findings are not contradictory. The admission pattern remains a reason to worry about selection, while the aggregate prior-funding moderator does not establish that prior funding explains the pooled relationship. Selection can operate through other characteristics, through a combination of traits, or through criteria the studies do not measure.