Counting applicants and graduates isn't evaluation. Baseline every founder at intake, track the same outcomes through the program, and compare cohorts on identical measures so you can prove what actually changed.
In short: Evaluating an accelerator or startup cohort means measuring what actually changed for founders between intake and demo day and beyond — readiness, traction, capability, and survival — not just counting how many companies applied and graduated. A real evaluation baselines every founder at intake, tracks the same measures through the program, and compares this cohort to the last one on identical questions. Sopact keeps each founder on one record under a persistent Contact ID, so intake, in-program milestones, and post-program follow-up stay comparable across cohorts.
Most programs already collect two things: an application at the front and a demo-day survey at the end. The problem shows up the moment someone asks a harder question. Was this cohort stronger than last year's? Which founders gained the most confidence, and on what? Where are the companies from two cohorts ago now? The application and the exit survey were built in different tools, use different questions, and were never linked to the same founder — so the answer is a guess. You can report outputs. You cannot yet report outcomes.
Evaluation starts before the program does. Capture each founder's starting point at intake: current revenue and users, funding raised to date, a self-rated readiness or confidence score, and the specific capability gaps they name in their own words. This baseline is the "before" half of every claim you will make later, so collect it the same way for every founder. In Sopact each founder becomes one Contact record at intake, and that Contact ID persists through the whole program — every later survey, milestone, and follow-up attaches to the same record instead of living in a separate spreadsheet.
Decide what your accelerator is actually trying to move, and separate it cleanly from what you happen to count. Outputs are activity: applications received, companies admitted, mentoring sessions delivered, demo-day attendance. Outcomes are change in the founder or the company: readiness and confidence gained, traction (revenue, users, funding raised), new capability, and, later, company survival and follow-on investment. Write two or three outcome measures you will ask about at intake and at exit using the same wording, so the before/after comparison is honest rather than reworded.
Don't wait for demo day to learn how the cohort is doing. Track a short set of milestones at fixed points — a mid-program check-in, a mentor assessment, a metrics update — against the same measures you baselined. Because every check-in writes back to the founder's persistent Contact ID, you can see movement per founder over the program instead of one snapshot at the end, and you can flag a founder who is stalling while there is still time to intervene.
A cohort evaluation only means something next to another cohort. Ask the identical intake and exit questions of every cohort, so this year's readiness gain, traction change, and completion sit beside last year's on the same scale. When the questions drift year to year, the comparison quietly breaks; keeping one measure set on one data model is what lets you say "this cohort raised more follow-on funding than the last two" and defend it. Watch for confounds; a stronger applicant pool, not a better program, can lift the numbers.
The outcomes funders care about most land after the program ends: which companies are still operating a year later, which raised a priced round, which hired their first employees. This is longitudinal work, and it only functions if the post-program survey reaches the same founder record you baselined at intake. Because the Contact ID persists after graduation, a six-month or twelve-month follow-up sits on the same timeline as intake and demo day — so survival and follow-on investment become a measured trend, not an anecdote you chase down over email.
Founders and mentors write the most useful things in the comment boxes: what unblocked them, what the program missed, what they still can't do. Theme that qualitative feedback the same way each cohort: pull the recurring barriers, the capability gains founders describe unprompted, and the mentor observations, so the story lines up with the numbers instead of sitting in a folder no one reopens.
Three prompts you can run against your own cohort data once it sits on one record:
Compare pre-program and post-program founder readiness scores for [COHORT YEAR A] and [COHORT YEAR B]. Report the average gain per cohort, the share of founders who improved, and any measure where the newer cohort did worse.
Using post-demo-day and [N]-month follow-up data, list the founders in [COHORT] by traction change (revenue, users, funding raised) and flag any company with no activity since graduation as a survival risk.
Theme the open-ended mentor and founder feedback from [COHORT] into the top 5 recurring barriers and the top 5 capability gains founders describe in their own words. Quote two verbatim responses per theme.
GRADE: green | one founder record from intake through post-program, same measures every cohort | a few questions reworded between years; amber | application and exit survey linked but no mid-program or post-program tracking | cohorts comparable on outputs only; red | applications and demo-day survey in separate tools, no shared founder ID | counting applicants and graduates and calling it evaluation.
Ask the outcome question at intake, not just at exit. The most common evaluation failure is having a great demo-day survey and no baseline to compare it against. If you only measure readiness at the end, you can report a level but never a gain. Add the same two or three outcome questions to your application form.
Freeze your measure set before you launch the cohort. Every time a question changes wording between cohorts, that measure drops out of your year-over-year comparison. Lock the core intake and exit questions, and add new ones as extras rather than by rewriting the ones you need to trend.
Separate a stronger cohort from a stronger program. If this year's numbers jumped, check whether your applicant pool got stronger before you credit the curriculum. Compare intake baselines across cohorts first; a higher starting point explains a lot of "improvement." See how to compare cohorts without being fooled by confounds.
Don't let post-program follow-up depend on chasing email. Founders scatter after demo day. If the follow-up survey isn't tied to the same record you built at intake, response rates collapse and your longitudinal data gets thin. Keep everyone on one persistent record and schedule the follow-up as part of the program, not as an afterthought.
Track both outputs and outcomes, and keep them labeled separately. Outputs are activity you can count: applications received, companies admitted, sessions delivered, demo-day attendance. Outcomes are what changed for founders: readiness and confidence gain, traction (revenue, users, funding raised), new capability, company survival, and follow-on investment. A useful accelerator evaluation reports the outcomes as before/after changes, with the outputs as context, not the other way around.
Ask every cohort the same intake and exit questions on the same scale, and keep each founder on one persistent record so the numbers line up. Comparison breaks when questions get reworded year to year or when data lives in separate tools. In Sopact, holding one measure set across cohorts under a stable Contact ID is what lets an accelerator say a cohort improved and show the evidence, while still checking whether a stronger applicant pool, not the program, drove the change.
Outputs describe what the accelerator did: how many startups it admitted, how many mentoring hours it delivered, how many founders showed up to demo day. Outcomes describe what happened to the founders because of it: they got more investment-ready, grew revenue or users, raised follow-on funding, or kept the company alive a year later. Outputs are easy to count and easy to game; outcomes require a baseline and follow-up, which is why most cohort reports stop at outputs.
Send a scheduled follow-up survey — commonly at six and twelve months — to the same founder record you created at intake, and measure survival, traction, and follow-on investment against that starting baseline. The follow-up only works longitudinally if it attaches to a persistent ID rather than a fresh, unlinked form. Treat it as a built-in program milestone, and see longitudinal data collection for how the same record carries across time.
Sopact keeps each founder on one record under a persistent Contact ID, so a cohort's intake baseline, in-program milestones, demo-day results, and post-program follow-up all sit on the same timeline instead of in separate tools. Because every cohort answers the same measure set on that shared model, Sopact can compare cohorts on identical outcomes, track traction and survival after graduation, and theme the open-ended founder and mentor feedback alongside the numbers. See accelerator software, the broader practice of impact measurement, and how to analyze pre, mid, and post survey data for the mechanics behind each step.
ChatGPT, Claude, and Gemini are fine for a quick test — but not for an answer you'll put in front of a funder or board. When it has to hold up, run it in Sopact Sense.
Try it in Sopact →
