The server responds; the customer cannot finish
Imagine a store whose server is available but payment fails after several minutes. The technical dashboard shows activity; people abandon checkout. Start by mapping a critical journey from intent to outcome. “Can someone complete an order?” is a more useful monitoring question than simply “Is the machine running?”
Define what counts as success and where you observe it. An accepted HTTP request does not by itself prove a document arrived or payment was confirmed. When work continues in the background, track completion and waiting time for that stage too.
Averages can hide the people struggling
Latency, traffic, errors, and saturation provide four useful starting perspectives. Average latency can look comfortable while a minority waits too long. The 95th percentile is a value below which approximately 95% of observations fall; the remaining cases still deserve attention.
Compare like with like: region, operation type, and success or failure, where volume and privacy allow. Do not conclude a change improved the service if users with slower connections simply disappeared during that period. The observed population is part of the interpretation.
An alert needs an actionable response
A useful alert identifies impact, an owner, and an initial procedure. “The server looks odd” forces someone to start from scratch. “Order completion dropped; here is the diagnostic query” reduces uncertainty. Set thresholds using your service’s data; copying another system’s numbers can create noise.
Connect events with correlation identifiers to reconstruct an operation. Record only what is necessary, control access and retention, and keep passwords, tokens, and sensitive content out of logs. Collecting everything can increase cost and risk without improving diagnosis.
A dashboard for one journey
Choose one journey and track attempts, completions, completion time, and failures for a week. Add an owner and a link to the response procedure. Review one real case with the team. If the data cannot explain the user’s experience, improve measurement before adding more charts.
Sources & further reading
Sources accessed October 24, 2026. Numerical examples and scenarios are illustrative.
Ready to put this knowledge into practice?
We can work together on your team’s challenge or your next project.
Talk to Calembe ↗