Builder Field Guide

How many teams use your open-source project in production?

For a self-hosted tool, count the outside organizations that can confirm a specific production task in a stated period. A founder can often substantiate a lower bound even when the total number of installations is unknown.

Download the free adoption worksheet (CSV)

What counts as one production adopter?

Imagine you maintain an open-source data sync tool. A package download can happen on a build server that never touches live data. For your October update, define a production adopter as one independent organization whose operator confirms a successful sync for a real workflow between October 1 and October 31, 2026, UTC. An organization can run ten installations and still contribute one to this count.

Give operators a question they can answer without sharing private data. “Did a sync finish against a live source or destination in October? Which date can we record, and may we count your organization anonymously?” Give each responding organization a consistent alias. Record the run date and environment. Leave unconfirmed fields blank. Naming the organization or publishing a quote needs separate permission.

Six reports, two confirmed organizations

These fictional reports illustrate the rule. The report date identifies the evidence. The date of the successful run determines whether it falls in the window. Count the first qualifying report for each outside organization.

October 2026 · synthetic operator reports
Report receivedWhat the report establishesNew orgs
Team A · Oct 6An operator confirmed a successful sync against live records on Oct 5.1
Team A · Oct 19A second operator confirmed another production sync on Oct 18. Same organization.0
Team B · Oct 24An operator confirmed a successful sync to a live destination on Oct 23.1
Team C · Oct 26The Oct 25 sync succeeded in staging.0
Maintainers · Oct 28The maintainers ran their own demonstration on Oct 27.0
Team D · Oct 30A developer said they were evaluating the tool. The environment and outcome are unknown.0

A reportable sentence. In October 2026, operators at two independent organizations confirmed that our tool successfully synced live data. We reviewed six voluntary reports. We don’t know how many other organizations self-host the tool.

Another run at Team A leaves the count at two. If Team D later confirms an October production sync, the count becomes three. The CSV worksheet includes the date boundaries, aliases and decisions. To reuse it, replace the sample rows with your own anonymous reports and sum the last numeric column after reviewing duplicates.

What do GitHub and PyPI actually measure?

A GitHub star can be a bookmark. The traffic graph records full clones and visitors over a rolling 14-day window. A repository’s “Used by” count reflects public package references that GitHub could identify. Private repositories are absent from that list, and a reference alone supplies no date for a successful run. GitHub stars · Repository traffic · Public dependents

A PyPI download is a file request. Caches can serve later installations without another download. Repeated CI jobs can generate many requests for one organization. On August 24, 2026, PyPI narrowed its counts to distribution files and stopped including metadata-only requests. For a chart spanning that date, use the same distribution-file filter on both sides or mark the break in the series. Report the download count on its own. Dividing two confirmed organizations by file requests cannot tell you what percentage of installers adopted the product. Python Packaging User Guide · PyPI’s measurement change

How three open-source teams described their evidence

  • Cerbos, November 2021. The company said multiple customers were already using its open-source authorization software in production. It left their exact number and payment status unspecified. Cerbos announcement
  • dltHub, November 2025. In a Bessemer interview, the founders reported developers created more than 50,000 custom connectors in September. One developer could create several connectors. Bessemer’s founder interview
  • Dylibso, March 2023. Seed investor Felicis credited the team with co-creating Extism and reported over 1,800 GitHub stars and support for more than 16 languages. Those figures describe interest and language coverage. Felicis announcement

The First Check Atlas comparison connects each claim to its dated financing announcement. Pick the unit you can support in your own update, then keep that unit and its reporting window consistent next month.

Share what you’re building

Send a working title and a short description for private review. Add a deck if you have one.

Open the short application

Applying creates no commitment.