Error tracking for startups (2-15 people)

Error tracking for startups here means teams of roughly 2-15 people shipping a SaaS or B2B product to paying users. You share on-call. You already have an official Sentry SDK. You need issues without turning error storage into a second platform team.

Team size changes the bill

Sentry Developer is one user. The moment a second engineer needs the project, you are on Team (about $26/mo annual default with included errors then PAYG, fetched 2026-10-05) or on a host that does not meter seats. Self-host seats are your IAM, not a vendor line item. That is often the first startup-specific fork.

Alert routing, not dashboards

A startup on-call wants a Slack or webhook fire on a new issue fingerprint, not a dozen observability dashboards nobody owns. Epure stores exceptions and stacks. Wire alerts to the issue. Keep logs and uptime where you already have them. When error tracking is enough.

Lived 2am

Two founders, one API, one worker. Tracing was still sampled because "we might need it." A bad migration threw in a loop. The Team projected bill moved while both founders were in the same Zoom reading one stack. We zeroed sample rates, dual-sent to Epure on a small VPS, and cut the metered DSN after Issues matched. The product lesson was migrations. The ops lesson was scope.

Shape vs indie page

Indie essay: solo cafe incident, invoice fear. This page: seats, shared pager, alert routing. Keep both. Error tracking for indie hackers.

How to stand it up for a small team

• One person runs Compose prod overlay and documents EPURE_PUBLIC_URL.
• Put the DSN in the shared secrets store under one name.
• Set traces/replay/profile rates to 0 across web and workers.
• Verify with a shared staging throw; confirm Issues and alert delivery.
• Schedule a monthly pg_dump restore test on a calendar, not in chat lore.

Migrate from Sentry. Backups.

Limits

If your customers demand session replay in support, keep a suite. If you are enterprise-sizing Sentry self-host, budget ~16 GB and read COMPOSE_PROFILES=errors-only. Epure Cloud list prices are not live checkout. Partial compatibility only.

Seed the on-call doc

Paste three links into the team doc on day one: Issues URL, Invalid DSN, Events not showing. Add the Compose restore command you actually tested. Startups lose hours when the only person who knows the DSN is on a flight.

Pricing conversation with a co-founder

Bring three numbers: Sentry Team base (about $26), a calculator spike scenario from this site, and the VPS SKU you already pay. Decide whether spikes should hit a meter. That conversation is shorter than a feature bake-off. Cheapest error tracking. Why the Sentry bill jumped.

Hiring boundary

When you hire an SRE who lives in traces, revisit suite spend. Until then, do not buy dashboards for a role you have not hired. Error tracking for startups fails when the tool assumes a platform team that does not exist yet.

Partial compatibility in plain words

Can you just change the DSN? Yes for exceptions. No for APM, replay, and profiling as Epure features. Say both sentences in the migrate PR description so nobody files "tracing broken on Epure" as a Sev-1.

Takeaway

For 2-15 person product teams, seats and spikes usually bite before suite features do. Put one DSN in secrets, zero unused sample rates, alert on issues, test restores. Revisit the suite when someone is hired to live in traces.

Staging and production split

Give staging its own project and DSN. Startups that share one project across staging and production spend their mornings filtering noise. Two projects on one Epure box is normal. Document both DSNs in secrets with clear names. Verify each environment chip once after cutover.

Contractor and agency access

Startups often grant a contractor temporary access during a crunch. On a metered free tier with a one-seat cap, that grant is a plan change. On self-host, it is an account you can revoke on Friday. Write the revoke step next to the invite step. Security theater is inviting and forgetting.

Takeaway expanded

Seats and spikes usually bite before suite features. One DSN name, zero unused rates, alerts on issues, restore tests on a calendar. Revisit the suite when traces become someone's job.