How do you tell whether your app is working for customers?
Check the address to see whether the app answers. Try one action a customer depends on to see whether they can finish it. If a failure affects people, update a status page with what you know and when you will check again.
What does a green homepage check tell you?
It tells you the address gave the expected reply when the check ran. Someone could still open the homepage and get stuck when they try to reserve a time.
| Question | What to look for |
|---|---|
| Can someone reach the page? | Ask the address for a response. If it times out or returns an error, check whether it is reachable from another place. |
| Can someone finish a booking? | Use an isolated test copy with sample records and test connections for any emails or payments. In that copy, use a browser check to select a time and look for the confirmation. An API check can separately verify the reply from the save action. |
| Do people know what is affected? | Post an update where they can see it even if they cannot open the app. State which action is failing and when you will update them. |
How do you practice catching a failure before you set up a monitor?
Use this example on paper. The homepage opens and shows three appointment times. A customer picks Tuesday at 2 pm, presses Reserve, and never sees a confirmation.
- Write down what passedMark the address check as passing because the page opened. Mark the booking check as failing because the confirmation never appeared.
- Write the check you would runUse one made-up appointment. Write the result you would check for. After reserving Tuesday at 2 pm, the page should confirm that time. If your app sends an email or saves a record, decide which of those you would verify separately.
- Choose who needs a messageWrite down who can look into the failure and what they need to know. If the same problem affects real bookings, draft an update such as “Some bookings aren’t confirming. We’re investigating. Next update by 14:30 UTC.” Keep it on paper.
- Decide what proves recoveryWrite down what would prove recovery in this example. You would see the confirmation and one saved booking. If you later try it on a test copy, confirm where it saves records, sends email or takes payment before you submit anything.
Compare tools
Scroll the table sideways to read every column.
| Tool | What it checks or showsThe first problem this tool can help with. | What you give itWhat you configure before it can do that job. | Who hears about a failureWhere an alert or update goes. | What your plan countsThe unit that can change your bill or capacity. |
|---|---|---|---|---|
| UptimeRobot | Checks whether an address replies and can check for expected words in the response. | Enter the address and choose a check type and interval. | Assign an alert contact to the monitor. Channels and extra people depend on the plan. | Monitor count and interval. Phone credits and extra contacts can matter. |
| Checkly | Checks an address, an API reply or a sequence of actions in a browser. | Give it a URL for a reachability check. For a browser check, supply the actions and the expected result. | Connect an alert channel to the check or group. Choose when a failure needs a message. | Uptime monitors count by monitor. API and browser checks count by runs. |
| Atlassian Statuspage | Shows visitors a status update about the part of your app that is affected. | Create a page and name the parts customers use. A person or a connected monitor supplies the updates. | Customers can subscribe to published updates through the channels the page offers. | Page type, plan and limits such as subscribers and people who manage the page. |
When should a failed check wake someone up?
Choose the failure you can act on and a person who can respond. A repeated booking failure may deserve a faster response than a brief delay on an information page.
A check from another location can help you decide whether a failure is local or wider. Keep location in the alert and look at the failed result. Two failures tell you more about reachability, but they still do not tell you which customer records or actions were affected.
Before depending on the alert, send a safe test through the channel you intend to use. Make sure the person sees the test on the device they have with them. Keep the alert about a broken action separate from the public update to customers.
How do you tell whether a scheduled job stopped running?
Have the job send a small check-in after it finishes successfully. If a nightly report should finish by 2 am, leave time for normal delays and alert when the check-in still has not arrived.
This kind of check-in is often called a heartbeat. Put it after the job has produced the result you need. A signal at the start only proves that the job began. Keep the check-in address private and record what the alert means so someone can tell a missing report from a slow one.
What should your status page say while you are still investigating?
Say which part customers cannot use, what you know so far and when you will update them again. Use names customers recognize, such as Booking or Sign-in.
Pick a page people can reach when the app is down. Give someone permission to update it and check which visitors or subscribers can see the message. Keep customer names, addresses and private diagnostics out of public updates.
When you have a fix, check the same action that failed. Tell people what works again and what they still need to do. If a confirmation arrived late or a booking may have been saved twice, explain how they can check before trying again.
What should you keep when you change monitoring tools?
Keep the addresses you check, the steps and outcomes for each important flow, and who should get alerts. Save incident updates and any history your team needs before turning off the old setup.
Set up the new checks with the same safe test cases and verify that alerts arrive before switching over. Check whether the plan counts saved monitors, scheduled runs, people who can sign in or subscribers. Extra locations and retries can count as runs on some services.