What actually worked

Thirteen sourced cases where a useful answer, tool, integration, or public artifact earned users before any ask, each with a confidence line and its limit.

You built the thing and the launch was quiet. Before you post anywhere, it helps to see what a useful contribution looked like for products that found their people.

The thirteen cases below share one shape. Someone gave an answer, an artifact, an integration, or a partnership that was useful before any ask. Then something closer to product value than applause happened: a paying customer, a retained cohort, a working database, a duplicated template.

Read the numbers with care. Nearly all of them are first-party and multi-causal. Every case carries a source, an access date, and a confidence line, so you can weigh it yourself.

How to read a case

• First-party versus independent. Every source here is first-party: a company blog, a founder retrospective, or a founder's own post. "The company reports" means exactly that. None of the thirteen has been independently audited, and the confidence line says so where it matters most.
• What confidence means here. Confidence rates the strength of the public evidence, not the odds that you'll reproduce the outcome. High means the operator published a specific milestone or an explicit product test. Low-medium means a detailed self-report we couldn't verify.
• Why no case is a recipe. Company histories can substantiate reported actions and metrics, but they rarely isolate causation. Launch weeks, integrations, and free tools happened alongside hiring, price changes, timing, and luck. Treat each case as a pattern to test, not a result to expect.
• What to look for instead. Notice the need that existed before the promotion, the format that fit the venue, and the measure that sits closer to product value than raw attention.

The cases

C01 — Zapier: answered one explicit integration question in a support forum

What to copy: Start with a stated need and answer it fully, without requiring a click. Mention the product only when it materially completes the solution, and only where the venue allows affiliation.

C02 — Zapier: built integrations that partners would put in front of their own users

What to copy: Contribution and channel fit multiply. Before building, get a named partner owner, a documented user need, and a credible commitment to distribute; the useful thing alone wasn't enough.

C03 — Buffer: taught social-media practice where its readers already were

What to copy: Turn one hard-earned product lesson into a complete, non-promotional guide for one publication whose readers have that problem. Stop pitching a venue when your topics don't fit its editorial needs or you can't keep the piece useful without a product link.

C04 — Plausible: published inspectable technical and business evidence in the open

What to copy: Publish one technically inspectable artifact, whether a benchmark, a migration analysis, code, or a transparent decision, with the caveats a peer would need. Don't turn a technical community into a recurring announcement feed.

C05 — Supabase: gave developers something real to inspect, then made launches a recurring program

What to copy: Bundle one meaningful release with a technical explanation, a working demo, a place for people to contribute, and a staffed response window. Don't manufacture a launch week when there's no meaningful release or no capacity to answer feedback.

C06 — PostHog: released usable open-source code and answered developers in public

What to copy: Release one artifact that lets a skeptical user verify your claim without a sales call, then answer implementation questions in public. Don't use "open source" as a launch label if the project can't support contributors or users.

C07 — HubSpot: turned a diagnosis into a free tool anyone could run

What to copy: Build the smallest diagnostic whose output tells the user what to do next, with inputs genuinely connected to your product's domain. A score without validation, actionable reasons, or a way to measure whether the advice helped is just a number.

C08 — Ahrefs: built free utilities around tasks people were already searching for

What to copy: Pick one recurring task adjacent to the paid product, publish a genuinely complete free utility, and measure qualified next actions rather than visits. Stop if the tool draws a crowd with no plausible product fit, or if support and compute costs outrun what you learn.

C09 — Notion: let users publish templates that lower the next user's starting cost

What to copy: Find one repeatable user output another user can adopt safely, then preserve attribution, quality controls, and an easy path back to the product. Don't launch a gallery without curation, rights clarity, a maintenance owner, and a way to retire stale artifacts.

C10 — Figma: tested whether community assets changed how teams collaborate

What to copy: Place a vetted community contribution where the user needs a starting point, then measure the downstream product behavior it should change, not the download count. If reuse rises but core success or creator outcomes don't, redesign the surface.

C11 — Baremetrics: made real operating data a public benchmark

What to copy: Publish one decision-relevant metric or operating artifact with definitions, caveats, and a consistent update cadence. Selective vanity metrics aren't transparency, and disclosure that creates customer, privacy, or security risk isn't worth the credibility.

C12 — Ghost: made an open project's governance and finances inspectable before asking for backers

What to copy: Make the governance, the artifact, and the promised public benefit concrete enough that supporters can inspect what they're joining. Mission language without governance or financial choices behind it is just copy.

C13 — An independent founder: turned subreddit research into an early cohort

What to copy: Ask a narrow research question only where the community permits it, publish what you learned, invite a small relevant cohort, and track retained use. Don't treat these metrics as a benchmark, and don't run research in communities that prohibit pain-point extraction.

What the useful cases have in common

Across the thirteen, the same six moves show up. The middle column is what you'd actually observe; the right column is a better thing to count than impressions.

This is a cross-case pattern, not proof that doing all six is sufficient. It tells you what to look for and what to measure; it doesn't tell you the result.

Weigh it against the weeds

Every case above has a counterpart that produced bots, votes, or sessions and no customers, and the weeds collects those with the same sourcing, alongside the tactics to avoid outright. When you want to see these moves applied to a specific kind of product, with real venue rules and a stop condition for each path, the portfolios work through five hypothetical examples.