Red flag metrics that tell you when users get stuck

Red flag metrics are usage signals that suggest a user is stuck: they started a task and have not finished it after much longer than normal, or they repeated a step several times. You set a threshold for each key task, alert your team when someone crosses it, and a human writes to that person. The tactic suits products with a setup or first task that takes a few minutes and where each new trial or customer matters, because the outreach is manual.

How to do it

1. Pick two or three tasks that predict activation, such as connecting a data source, creating the first project or sending the first campaign. Skip the rest for now.
2. Measure how long successful users take. Use your event data to get the typical range for each task, from start event to done event. A median and a 90th percentile are enough.
3. Set the red flag. For example: the user started the task and has not completed it after twice the 90th percentile, in more than one session. Two sessions filter out someone who just got a phone call.
4. Wire the alert. In an email or messaging automation tool, start a flow on the "started" event and wait for the "completed" event with a maximum time. In Customer.io, a Wait Until step holds someone until a condition is met, and its Max time path is "the maximum time that a person waits before moving on in the workflow." When the timeout path fires, post an alert to Slack, your help desk or a task list.
5. Write the message once and send it from a person. Keep it short and offer help. A template: "Hi [first name], I saw you might be having trouble setting up [task]. If so, I would like to help. Reply to this email and tell me where you are stuck." Sign it with your own name.
6. Log every alert and every reply. After a month, review which tasks produce the most alerts. That list is your product roadmap for onboarding fixes.

What it looks like in practice

Say you run a reporting tool. Successful trial users connect their data in under 5 minutes. You set the flag at 15 minutes across two sessions. An alert fires for a trial user who has opened the connection screen three times. You email an offer to help. If the reply says their database is behind a firewall, you send the allowlist address and then add that sentence to the connection screen, so the next user does not need an email.

When it backfires

• Alert fatigue. Thresholds set too low flood the team and people stop reading. Start strict and loosen.
• It feels like surveillance. An email that says "I watched you for 14 minutes" is creepy. Word it as an offer to help, and describe the tracking in your privacy policy.
• Wrong benchmarks. If your median is based on 20 users, the threshold is noise. Wait for enough data, or use a round number you can defend.
• It only works if someone replies fast. An alert answered three days later is worse than none. Assign an owner.

Pair this with an onboarding checklist so that the tasks you watch are the same ones you ask users to complete. If you want to find your key task first, see key activation metric.

Questions people ask

What is a red flag metric? A usage measure with a threshold that signals a customer may be struggling or about to churn.

What tools show red flags in customer success? Product analytics tools flag drops in usage. For stuck-in-a-task alerts, an automation tool with event triggers and wait steps, such as Customer.io, can do it.

What are examples of early warning signs? Long time on one setup step, repeated failed attempts, no login for a set number of days after signup.

Sources

• Customer.io: Wait Until, read Oct 8, 2026