Transparency as a feature, not a press release

Most marketplaces publish a transparency report when they get big enough that someone external is asking pointed questions. Uber's first one came in 2019, eight years after launch. Airbnb's followed under pressure from regulators. The pattern is: hide the numbers until you can't anymore, then publish a curated subset framed for damage control.

We made an early call to publish the report from launch — the first one is live at /safety/transparency-report — and to publish the figures that come out of production directly, without curation. This post is why.

Numbers we'd rather not show, shown anyway

Helpward at launch will have ugly figures alongside good ones. Response time during the helper-supply ramp will be higher than we want. The cancellation rate in cities with thin supply will spike. Some quarterly disputes will resolve in ways our reviewers second-guess.

Publishing those numbers anyway is the point. If we only publish when the figures flatter us, the report stops being useful — it becomes a marketing artefact. If we publish every quarter, including the ones where a metric got worse, the report becomes a contract between Helpward and the people using it.

What's in the launch baseline

• Helper-supply pipeline: approved helpers, avg time to approval, approval rate, suspensions for cause.
• Booking outcomes: total bookings, success rate (completed with no dispute), avg match time, cancellation rate.
• Disputes: filed counts by category, % refunded, avg resolution time.
• Safety incidents: total reports, severe incidents, helpers removed for safety cause, avg response time.
• Commitment delivery: % of safety reports acked within 4h, % of disputes acked within 24h, % of refunds within 5 business days, % of insurance claims acked within 24h.

Commitments we're making about future reports

1. Quarterly cadence. First business day of the new quarter, every quarter. If we miss a date, we publish the report alongside an explanation of why it was late.
2. Production-direct figures. No hand-curation, no PR pre-screen. The methodology paragraph below every section explains exactly how the figure was computed.
3. Written explanation for any metric that worsens by more than 25% quarter-over-quarter. Buried regressions are how trust dies; we'd rather name them.
4. Adding law-enforcement transparency Q3 2026. Counts of LE information requests received, complied-with, and rejected, separated by jurisdiction and request type. Same format Cloudflare uses.
5. Third-party attestation at 50,000 quarterly bookings. Once we cross that threshold an independent firm verifies the figures annually.

What this gives us internally

There's a side benefit we didn't initially plan for. Publishing the report on a fixed cadence creates an internal forcing function — we can't ship a Q3 report with the same blanks as Q2. So the metrics become real operational targets, not vanity dashboards.

When the next quarter's metrics are due, we either improved them or we wrote down why we didn't. Both are productive states; what's not productive is the state where the metric exists in a slide deck and nobody outside the company can challenge it.

Read the launch report

The Q2 2026 launch report is live at /safety/transparency-report. Most figures show '—' because we're publishing the table structure before we have a full quarter of data — Q3 is the first report with real numbers. We thought publishing the empty baseline first was more honest than waiting three months and showing up with a polished report nobody had any prior expectation of.