About
A free, honest read on early startup momentum
CohortWatch is a free, open-source view of public startup momentum for pre-seed to Series A accelerator companies. It reads a dated export, compares each company within a supported peer pool, and carries confidence and observable evidence with every reading.
CohortWatch is an independent open-source research project. It is not a prediction of success, a valuation of equity, or a measure of team quality. Its narrower job is to make observable public activity comparable, dated, and inspectable.
Startups disclose unevenly, so CohortWatch does not pretend to read fundamentals it cannot see. The current dated export contains hiring and frozen code signals. Capital is a designed signal family, but no capital collector data ships yet. A reading is not a verdict on the company or a recommendation to join, invest, or buy.
What it is not
- Not a prediction of success
- Not a valuation of equity
- Not a measure of team or job quality
- Not a recommendation
The wedge
One honest framework, across more than one accelerator.
Public startup data is fragmented and uneven. CohortWatch applies one normalization and evidence standard across multiple accelerator directories, including companies with sparse or quiet public footprints.
The hard part is sustained normalization: keeping fragmented public sources comparable, dated, and reproducible over time. The method is open. The defensibility comes from operational discipline and an auditable history of doing the work.
How people use it
One public reading. Different questions.
CohortWatch supplies inspectable context. The decision built on that context still belongs to the person making it.
Founders and operators
See how a company's public footprint reads against its supported peer pool, then inspect or correct the underlying evidence.
Early employees
Compare observable activity while treating momentum as one input. CohortWatch does not measure culture, runway, compensation, mission, or job quality.
Accelerator and portfolio teams
Triage where public evidence warrants a closer look without turning the score into a leaderboard.
Researchers and investors
Reproduce a dated, peer-relative view of public momentum. It is context, not a recommendation.
What stays true
The rules that keep the reading honest.
Peer-relative, with the pool named
Every published comparison names its supported peer pool. When the available fallback crosses accelerators, public peer statistics are withheld.
Confidence travels with the score
The pre-seed window is the thinnest-signal zone. A thin-data company reads as low confidence, never as low momentum.
The quiet ones stay in
Companies with sparse or quiet public footprints stay in the dataset. Missing evidence lowers confidence instead of becoming a confident negative.
Free proxies only
Public sources only. No paid databases, no LinkedIn scraping, and no founder names in evaluative public artifacts.
Where this is headed.
A guiding direction, recorded to steer decisions over time. It is not a scheduled phase, a committed deliverable, or in-scope today. When it conflicts with what ships now, what ships now wins.
The current export uses accelerator cohorts as a bounded comparison set. A later research direction is to test whether the same public-signal framework can surface relevant off-list companies. That would require separate validation and would remain distinct from any assessment of business quality.
Any off-list discovery layer would need to keep two questions separate: what public momentum is observable, and what deeper evidence says about the business itself. The current score answers only the first.
Everything is open. Inspect the work.
Read the methodology, explore the tracked companies, or inspect the code. No account is required.