# Open-source signal review

Use this before adding a GitHub or open-source field to the firm's company records, and again at each annual review of the field. Fill it from the firm's own records and the public archive, not from a vendor's description. The definitions follow Conti, Peukert, and Roche (*Organization Science* 36, no. 4, 2025, pages 1551 to 1573; figures from the authors' November 2024 manuscript on SSRN); the lookback is the firm's own evidence.

**Firm / scope (funds, stages, sectors covered):** ____________________

**Prepared by / date:** ____________________

**Field owner (who can explain any value it shows):** ____________________

**Source: built from GH Archive and the GitHub API / bought from a vendor (name):** ____________________

## Definition

- [ ] The field counts a company organization's activity on repositories it does not own (forks, pushed commits, pull requests), per quarter.
- [ ] Stars and README or documentation edits are excluded, or shown separately and never summed into the count.
- [ ] The field shows the change over the last two quarters, not only a lifetime total.
- [ ] A company with no linked organization reads "no public record," never zero.
- [ ] The field notes that open source pulled in through package managers (pip, npm) leaves no public event, so a low count is not evidence of low use.
- [ ] The field covers organization accounts only. Tracking individual developers is out of scope unless decided separately and recorded here: ____________________

**Event types counted:** ____________________

**Event types excluded, and why:** ____________________

## Linkage

| Check | Method | Result |
| --- | --- | --- |
| Organization linked to company by the website domain on the GitHub organization |  |  |
| Sample of 20 linked records checked by hand for a wrong match |  | Wrong matches found: |
| Link method shown on the record (domain match / manual / vendor) |  |  |
| Companies with more than one organization handled (summed / primary only) |  |  |

## Vendor questions (if bought)

| Question | Vendor's answer | Evidence seen | Acceptable? |
| --- | --- | --- | --- |
| Which event types feed the number? |  |  |  |
| Do stars or documentation edits count? |  |  |  |
| How are accounts linked to companies, and can the firm see each link? |  |  |  |
| How current is the data, and how is a missing account shown? |  |  |  |

## Lookback

Run on companies the firm met in the past two to three years, before the field is shown to the team.

**Companies in the lookback / period:** ____________________

**Share with any linked organization:** ____________________

**Share with outside-repository activity:** ____________________

| Company | Segment (novel / crowded / other) | Engagement began (quarter) | First round (quarter) | How the firm first met it (quarter) | Would the field have prompted an earlier look? |
| --- | --- | --- | --- | --- | --- |
|  |  |  |  |  |  |
|  |  |  |  |  |  |
|  |  |  |  |  |  |
|  |  |  |  |  |  |
|  |  |  |  |  |  |
|  |  |  |  |  |  |

- [ ] Results are read by segment. Crowded-market companies without public activity are not treated as weaker.
- [ ] At least one decision the field would have changed is named, or the field is not built.

**Decisions the field would have changed:** ____________________

## Placement

- [ ] The field appears on the company record with its date, link method, and a link to the underlying events.
- [ ] It is presented as a prompt to look, not a score, and is not used to rank or filter out companies.
- [ ] It is reviewed on the same cadence as the rest of the pipeline.

**Where it appears (record view, pipeline report, alert):** ____________________

**Decision:** build / buy / run the lookback again later / do not add

**Approved by / date:** ____________________

**Next review (at least annually, and after any change in data source or link method):** ____________________

## Common failure modes

- "Has a GitHub organization" is shipped as the signal, which the study found weaker than working with outside code.
- Stars are counted, and the field measures attention instead of building.
- A blank is stored as zero and pushes companies without public code down a sorted list.
- A company is matched to someone else's organization by name, and nobody checks the domain.
- The field is never tested against the firm's own history, so no one knows whether it would have changed a decision.

This review decides whether a signal belongs in the firm's records and how it is defined. It does not assess a company's technology, and it does not replace diligence.