Does Startup Acceleration Work? What the Evidence Says

Does startup acceleration work? Yes — and the same evidence base shows that most individual accelerators do not. A decade of Global Accelerator Learning Initiative (GALI) data covering more than 23,000 ventures finds that accelerated companies outgrow rejected applicants on revenue, employment, and capital raised; a 2025 NBER analysis of roughly 750,000 US startups and 329 accelerators finds that most programs have negative value-added against a no-accelerator benchmark, with a small right tail of excellent programs generating the large average gains. Both findings are true simultaneously, and together they redefine the industry’s question: not “should accelerators exist?” but “is your program in the right tail — and can you prove it?”

Key Takeaways

• GALI — built by ANDE and Emory University on 23,000+ tracked ventures — finds accelerated ventures outperform rejected peers: nearly 75% of accelerated ventures grew revenue within a year versus roughly 50% of rejected applicants (1, 2).
• Acceleration effects hold across geographies: average gains in equity and debt raised are nearly identical in emerging markets and high-income countries, though emerging-market ventures convert acceleration into revenue growth more than into investment (3).
• The 2025 NBER working paper “Beyond Demo Day” (Baek and Hegde) finds accelerator value-added is highly dispersed: most programs are value-negative relative to no acceleration, while a small right tail produces large gains in acquisition, employment, revenue, and valuation (4).
• Selection is systematic — better ventures sort into better programs — meaning unadjusted alumni statistics overstate nearly every program’s true effect (4).
• Design evidence is specific: cohort selection and peer learning, capital linkage, and network quality move outcomes; generic mentorship and classroom curriculum show little measurable effect on funding outcomes (2, 5, 6).
• High value-added programs also accelerate the shutdown of weak ventures — fast, candid failure is part of what good acceleration produces (4).

What Does the GALI Evidence Actually Show?

Begin with the strongest dataset the industry has. The Global Accelerator Learning Initiative, co-created by the Aspen Network of Development Entrepreneurs (ANDE) and Emory University in 2015, did what almost no program does for itself: it tracked both accepted and rejected applicants over time, creating a comparison group that turns alumni anecdotes into estimable effects (1). Across its core analyses, the pattern is consistent — accelerated ventures report more revenue, more full-time employees, and more capital raised than comparable ventures that applied and were turned away. In the headline figure, nearly 75% of accelerated ventures increased revenue over the following year, against about 50% of rejected ventures (2).

Two refinements matter for East African readers. First, geography does not erase the effect: GALI’s emerging-markets report with Deloitte, drawing on 2,400+ ventures across 43 programs, found average acceleration effects on equity and debt nearly identical between emerging markets and high-income countries, with emerging-market ventures just as likely to achieve rapid growth (3). Second, the channel differs: ventures in low- and middle-income countries convert acceleration into revenue and employment more readily than into investment, because the surrounding capital market — not the program — caps what fundraising support can deliver (3, 5). Accelerators narrow the investment gap between emerging-market and high-income ventures; they cannot close it alone (3). For a region where the post-program capital desert is the defining condition — the subject of the post-accelerator valley of death — that finding should drive design: programs here should be engineered to produce revenue-sustained companies first and fundable companies second.

So the affirmative case is real, replicated, and cross-market. If the evidence stopped there, the industry could declare victory. It does not stop there.

Why Does the NBER Paper Change the Conversation?

Because it separates two things GALI’s averages blend: whether acceleration works on average, and whether the typical program adds value. In “Beyond Demo Day: Sorting and Value Added in Startup Accelerators” (NBER Working Paper 35063), Youn Baek and Deepak Hegde adapt the teacher value-added methodology from education economics to a sample of roughly 750,000 US startups linked to 329 accelerators — estimating each program’s value-added (AVA) while explicitly accounting for which ventures sort into which programs (4).

Three findings deserve to be quoted at every industry conference. First, selection is systematic: observably stronger ventures are more likely to enter acceleration at all, and more likely to sort into higher-value-added programs — so raw alumni outcomes flatter every program, and flatter the best programs most (4). Second, performance is brutally dispersed: most accelerators show negative value-added relative to a no-accelerator benchmark, while a small right tail generates large gains. The industry’s positive average is an artifact of excellence at the tail, not competence at the median (4). Third, the right tail is real and durable: high-AVA programs predict better long-term outcomes — acquisition, employment, revenue, valuation — and, tellingly, also accelerate the shutdown of weaker ventures. Good programs kill bad ideas faster, which is value creation the sector’s celebratory culture rarely counts (4).

Read GALI and NBER together and the synthesis is clear: acceleration is a high-variance intervention whose median instance may be worth less than nothing and whose best instances are among the most effective enterprise-development tools ever measured. The honest question for any program director, funder, or ministry is therefore distributional. The industry’s own funding reckoning — examined in the ESO funding crisis — has made the question existential: when capital is scarce, only programs that can evidence right-tail performance deserve to survive.

One caveat cuts both ways. The NBER sample is American; East Africa has no equivalent census. It would be convenient to assume the region’s programs cluster in the right tail. The structural facts — donor funding that paid for activity rather than outcomes, imported curricula, almost no counterfactual measurement — suggest the opposite is at least as likely. We do not know. That is precisely the problem.

What Separates Right-Tail Programs From the Rest?

The evidence will not hand any director a complete recipe, but it is far more specific than the industry’s habits suggest. Four design variables carry the measurable weight.

Selection quality. The single most consistent finding across GALI-affiliated research is that the ability to select growth-capable ventures drives downstream outcomes, including subsequent capital attraction (6). Selection is not gatekeeping vanity; it is the program’s first product. And the NBER sorting result adds a discipline: a program must know whether its outcomes come from picking winners or improving them — both are legitimate, but only one is value-added, and funders are learning to ask which (4).

Capital linkage. Programs embedded in real networks of investors and follow-on capital show stronger funding outcomes; programs that teach pitching without proximate capital show little (2, 6). GALI’s funding-flow analysis is blunt that investment benefits concentrate where the surrounding capital market can respond (5). The design implication for East Africa is not to abandon capital linkage but to redefine it: link ventures to the capital that exists here — revenue-based financing, trade finance, angel networks and domestic LPs, corporate procurement — rather than rehearsing for venture rounds the market rarely supplies.

Peer architecture. Among the “soft” components, structured peer learning is the one with evidentiary support. Village Capital-affiliated GALI research found a flipped curriculum — more entrepreneur working time and peer interaction, less classroom — associated with better outcomes, and cohort composition itself shapes whether transparent peer exchange happens at all (6). Generic mentorship hours and standardized curriculum, by contrast, show little measurable effect on funding outcomes (2, 6) — a sobering result for the industry’s most photographed activities. Trust density inside the cohort appears to be the active ingredient, which is the design logic explored further in values-based accelerator design.

Duration and dosage fit. The three-month sprint is an administrative convention, not an evidence-based one; emerging-market ventures, growing through revenue rather than fundraising milestones, plausibly need longer arcs, milestone-based progression, and post-program accountability windows. The right-tail programs’ willingness to accelerate shutdowns (4) points the same direction: a program confident in its value tells a founder the truth early, rather than graduating everyone on schedule.

What Would a Program Built From the Evidence Look Like?

Every claim above converts into an audit any funder, government buyer, or director can run. Call it the Right-Tail Audit — five questions, each answerable with data a serious program already collects or can begin collecting this quarter:

1. Counterfactual: Do you track rejected or waitlisted applicants alongside alumni? Without this, every outcome claim is unfalsifiable theater.
2. Selection ledger: Can you decompose your results into picking effects versus improvement effects — and would your model survive the comparison?
3. Capital conversion: What share of graduates closed any external capital or major commercial contract within 12 months — and from sources your program’s network actually touched?
4. Peer density: What fraction of program time is structured founder-to-founder work, and do alumni transact with each other (deals, referrals, hires) after graduation?
5. Survival honesty: What are your 24-month survival and revenue-growth numbers — and how quickly do your weakest ventures find out they should stop?

A program that can answer all five is either in the right tail or on a credible path to it. A program that can answer none is — on the best available evidence — more likely destroying value than creating it, regardless of how full its event calendar looks. The audit is also a procurement tool: as governments become the largest remaining buyers of acceleration in East Africa, these five questions are precisely what tenders should require, an argument developed fully in how governments should buy acceleration.

The standing objection deserves an answer: measurement of this kind sounds expensive, and most East African programs run lean. Two responses. First, the costly version is optional. A counterfactual does not require a randomized trial — it requires keeping contact with the applicants you rejected, a spreadsheet, and two follow-up surveys a year. GALI built the world’s best acceleration dataset substantially on voluntary applicant surveys (1); a single program can replicate the method at the scale of its own funnel for the cost of one staff-week per cohort. Second, the comparison is wrong. The relevant alternative to measuring is not saving money; it is continuing to spend the entire program budget without knowing whether the program works — which the NBER distribution says is, for the median program, the most expensive possibility available (4). Measurement is not overhead on the product. In an industry where most products fail, it is the product’s quality-control function, and the first budget line a serious director protects.

The composite right-tail program for this region, built strictly from the evidence, looks something like this: ruthlessly selective intake against explicit growth-capability criteria; a flipped, peer-dense format with founders doing real work on real numbers; capital linkage rebuilt around instruments the regional market redeems — revenue-based finance, structured debt, procurement pipelines — rather than demo-day equity theater; duration set by milestones, not calendars; and a measurement system with a counterfactual at its core, reporting survival and revenue at month 24 as the headline metrics. Nothing in that design is exotic. What is rare is the institutional discipline to run it — and the willingness to publish the numbers.

Where Should the Industry Go From Here?

The optimistic reading is the correct one, and it is worth stating boldly: the evidence has matured to the point where excellence is now legible. A decade ago, a great program and a busy program were indistinguishable to funders, founders, and governments alike; today, GALI’s methodology and the NBER’s value-added framework give the industry the measurement instruments that teaching and medicine took generations to acquire. That changes the competitive game in East Africa’s favor — because in a field where quality was invisible, money flowed to incumbency and announcement volume, and in a field where quality is measurable, it flows to whoever performs.

The region is saturated with programs and starved of proof. The first East African accelerators to adopt the Right-Tail Audit voluntarily — tracking counterfactuals, publishing 24-month survival and revenue, decomposing selection from improvement — will not merely survive the sector’s funding reckoning. They will define the standard that everyone else is eventually procured against. Acceleration works. The evidence is in. Now the work is making sure each individual program deserves the average.

Frequently Asked Questions

Do startup accelerators actually work?
On average, yes. GALI data tracking 23,000+ ventures shows accelerated companies outperform rejected applicants on revenue, employment, and capital raised — nearly 75% grew revenue within a year versus about 50% of rejected peers. But averages conceal enormous variance between individual programs (1, 2).

What did the NBER “Beyond Demo Day” study find?
Analyzing ~750,000 US startups and 329 accelerators, Baek and Hegde found most accelerators have negative value-added relative to no acceleration, while a small right tail of excellent programs generates large gains. Stronger ventures also systematically sort into better programs, inflating raw alumni statistics (4).

Does acceleration work in emerging markets like East Africa?
Yes — GALI’s emerging-markets research found average effects on equity and debt nearly identical to high-income countries, with emerging-market ventures equally likely to grow rapidly. The difference is channel: gains arrive more through revenue and jobs than investment, because local capital markets limit fundraising outcomes (3, 5).

What program features does the evidence support?
Selection quality, structured peer learning (flipped curriculum with more founder working time), embeddedness in real capital and commercial networks, and honest outcome measurement. Generic mentorship hours and standardized classroom curriculum show little measurable effect on funding outcomes (2, 5, 6).

How can you tell if an accelerator is in the right tail?
Ask five questions: Does it track a counterfactual (rejected applicants)? Can it separate selection effects from improvement? What share of graduates close capital or contracts within 12 months? How much time is structured peer work? What are its 24-month survival and revenue numbers? Programs that answer all five with data are the exception — and the standard (2, 4).

Related Reading

• The ESO funding crisis: what the aid collapse revealed
• How governments should buy acceleration: a procurement manifesto
• The post-accelerator valley of death in East Africa
• Values-based accelerator design: the industry’s most underrated choice

Sources and Evidence

1. Global Accelerator Learning Initiative (GALI / ANDE / Emory University). “Does Acceleration Work? Five years of evidence from the Global Accelerator Learning Initiative.” https://www.galidata.org/assets/report/pdf/Does%20Acceleration%20Work_EN.pdf — The sector’s largest longitudinal dataset (23,000+ ventures), with rejected-applicant comparison groups; the foundational affirmative evidence.
2. GALI. “Initial Insights from GALI.” https://www.galidata.org/insights/ — Source for the ~75% vs ~50% revenue-growth comparison and findings on which program components do and do not move funding outcomes.
3. GALI / Deloitte. “Accelerating Startups in Emerging Markets.” https://www.galidata.org/assets/report/pdf/Accelerating%20Startups%20in%20Emerging%20Markets.pdf — Comparative analysis of 2,400+ ventures across 43 programs; emerging-market vs high-income effects and the investment-gap finding.
4. Baek, Y. and Hegde, D., 2025. “Beyond Demo Day: Sorting and Value Added in Startup Accelerators.” NBER Working Paper 35063. https://www.nber.org/papers/w35063 — Value-added analysis of ~750,000 startups and 329 accelerators; negative median AVA, right-tail concentration, systematic sorting, and shutdown acceleration.
5. GALI. “Accelerating the Flow of Funds into Early-Stage Ventures.” https://www.galidata.org/publications/accelerating-the-flow-of-funds-into-early-stage-ventures/ — Evidence on where acceleration’s investment benefits concentrate and the role of surrounding capital markets.
6. Porticus, 2024. “When, how and who accelerators accelerate: 10 years of findings.” https://www.porticus.com/latest/lessons-and-research/2024/when-how-and-who-accelerators-accelerate-10-years-of-findings/ — Funder synthesis of a decade of GALI and Village Capital research: selection effects, flipped curriculum, network quality, and the limits of mentorship and curriculum.