Performance
Synthetic monitoring
Synthetic monitoring is a scheduled robot that loads a page or runs a scripted flow from a chosen location and records timings and pass or fail. It runs whether or not anyone is visiting.
How it is measured
A probe visits a URL every 1 to 15 minutes from set regions, on a defined device and network profile, and stores TTFB, load time, and error status. Scripted checks click through a login or checkout and assert that a key element appears. Alerts fire after a number of consecutive failures.
Because conditions are fixed, trend lines are clean. A 400 ms jump is a change you or your host made, not a different mix of visitors. Use the median of several runs, since a single run can vary by 10 to 20%.
Worked example
A Berlin bakery's online shop is checked every 5 minutes from Frankfurt and Virginia. At 03:10 the Frankfurt probe reports a checkout assertion failure three runs in a row, while homepage TTFB stays at 280 ms. Nobody is shopping at that hour, so there is no visitor data to show it.
A payment plugin update broke the Pay button. The alert reaches the owner at 03:25, and the fix lands before the morning orders.
How it differs
Synthetic monitoring is a controlled test on a schedule. Real user monitoring records whatever visitors experience. Synthetic cannot see real devices or unpredictable networks. RUM cannot see quiet hours or pages nobody has visited. Uptime checks are a simpler synthetic type that only ask whether the page responds.
Common errors
Testing only the homepage. Monitoring from one region. Alerting on a single failure. Using a warm browser profile. Not scripting the revenue path. Letting the probe get blocked by your own bot protection.
In practice
Add one check for the homepage and one scripted check for your money path, such as signup, checkout, or the contact form, from the region your customers are in. Alert after two consecutive failures. Review the median trend monthly beside your RUM data.