A status page reads, from the outside, like a weekend project. It is a page with a list of components, a table of incidents, a form to post an update, and an email send when something changes. Most engineers who look at it estimate a few days, and the first version usually does take a few days.
The estimate leaves out everything that has to be true for the page to do its actual job. That job is being believed during the one moment your customers are deciding whether to trust it.
Building your own is not wrong. Plenty of teams should, and the reasons are specific. This post is an argument for estimating the real thing, not the demo.
What does the first version of a status page get you?
A static status page, a database table and a form cover the smallest possible case. That case is a solo developer or a two-person team who wants somewhere to type "API is down, investigating" during an outage instead of tweeting it. If that is the entire requirement, build it.
A single HTML page with a JSON file you edit by hand and deploy is a legitimate status page at that scale. It will outlast most engineering debates about whether it should exist.
The estimate stops being accurate once any of the following becomes true. For a company with paying customers, that is usually within the first year.
The page needs to update itself. A status page someone has to remember to open and edit stays green through outages nobody was awake for, or too busy firefighting to narrate. The fix is wiring the page to something that actually checks the service. That means a cron job hitting your health endpoint, comparing the result to what it expects, and writing to the incidents table when it changes.
A status page wired to a checker is no longer just a page. It is a small monitoring system, with its own failure modes:
- What happens when the checker itself goes down.
- How many consecutive failures count as an incident instead of a blip.
- What a flapping check does to your incident history.
Subscribers want to be told, not asked to check. The moment someone asks "can I get an email when something changes," you are building an email system. It needs signup, double opt-in confirmation (skip this and your sending domain gets flagged as spam within a month), and one-click unsubscribe. The unsubscribe header has a standard, RFC 8058, and mailbox providers increasingly expect it.
A subscriber system also needs a queue, so a burst of incident updates does not fire five hundred emails synchronously and time out your request. And it needs enough restraint in what triggers a send that people do not learn to ignore you. None of this is hard individually. All of it together is a real subsystem, not a feature.
More than one thing can be down at once, independently. "API," "Dashboard," and "Webhooks" failing on different schedules is normal. A status page with one overall state cannot represent that without lying in one direction (calling the whole system down over one degraded component) or the other (calling everything fine because the homepage still loads). Per-component status is a real data model decision, not a UI tweak.
A customer asks to put it on your own domain. status.yourcompany.com instead of a subdomain of whatever you built it on is close to a requirement once customers screenshot your status page into their own incident reports. That means TLS certificate provisioning per customer domain, which is a small infrastructure project if you have never automated it.
The pieces are the ACME protocol (RFC 8555), DNS or HTTP challenge handling (Let's Encrypt documents the challenge types), and renewal before expiry. All of it has to work unattended.
Someone asks "what was our uptime last quarter." History that survives longer than your database's default retention, rendered as an honest percentage rather than a number someone eyeballed, is a reporting feature most homegrown pages never get around to. That lasts right up until a customer's procurement team asks for it in writing.
Why does a status page need to be correct under pressure?
A status page's entire value depends on it being right exactly when the rest of your systems are not. A bug in a normal internal tool causes an annoyance. A bug in a status page causes people to stop trusting your incident communication at all, which is the one thing the page exists to earn.
Three examples of that kind of bug:
- An incident that opens for a check that was actually fine.
- A maintenance window that fails to silence the alert it was supposed to silence.
- A page that shows green because a query timed out and the frontend swallowed the error.
Maintenance windows are more subtle than they look. A window needs to silence alerts for the components it names, without also hiding the underlying measurement. If the migration runs long, you need the monitor to still notice. The window must also leave alone components outside it that have nothing to do with the planned work.
Getting that scoping wrong in either direction is the kind of bug that only shows up during an actual incident. Either alerts fire during announced maintenance, or the page stays silent on an unrelated real outage. An incident is the worst possible time to discover it.
What does a status page cost after launch?
The first version shipping is not the end of the cost. This is the part that trips up teams who estimated the build honestly and shipped something solid.
A status page is one of the few pieces of internal software whose bug reports arrive during your worst days. Three typical reports:
- The maintenance window that silenced the wrong component during a real migration.
- The email queue that backed up during a genuine incident and delivered the "resolved" notice four hours late.
- The TLS certificate that quietly expired on a customer's custom domain, because the renewal cron job failed silently three weeks earlier and nobody was watching a system that was supposed to need no watching.
None of these show up in a demo. All of them show up eventually, and by definition they show up when the page has the least margin for error.
The plan "we will build the simple version now and add the rest later" is reasonable for the first year and risky to leave permanent. The simple version optimizes for the case where nothing goes wrong with the status page itself. That is backwards, because a status page exists to be correct on the day something else has already gone wrong.
The honest cost comparison
Building the full version is closer to several weeks of focused engineering time than a weekend. The full version means checks, per-component status, subscriber email with proper deliverability handling, a custom domain with automated TLS, and history that survives a quarter. Add ongoing maintenance every time an edge case surfaces, and it will: a maintenance window that overruns, a check that flaps, a bounce-handling edge case in the email sends. For a team that is not in the business of building monitoring infrastructure, that is real opportunity cost against the team's actual product.
Buying starts at free. It includes checks from multiple regions, subscriber emails with confirmation and unsubscribe already handled, per-component status, and a custom domain on paid tiers. None of that requires estimating an email deliverability project as a side quest to your actual roadmap.
The honest case for building anyway has three forms:
- You have specific requirements a vendor's page will not fit, such as an unusual auth model for a private page or a data residency requirement.
- You are building status pages as your actual product.
- The volume of pages you would run makes the vendor's per-page or per-subscriber pricing genuinely worse than the engineering time. That happens at a scale most companies asking this question have not reached.
Outside of those cases, the weekend-project estimate is almost always wrong by an order of magnitude. Subscribers, components and a custom domain all enter the picture, and none of those three stays optional for long.
If you want a gut check on where an existing page stands against this list, the status page grader can help. It scores the URL of a public status page, yours or a vendor's, against what a reader needs during a real incident. It is free and needs no signup.