Digital Experience: RUM, Session Replay and Synthetic Monitoring
Digital Experience tests how you measure what users actually experience: real user monitoring data, session-level evidence, synthetic checks, and the segmentation that turns a vague complaint into a specific finding.
Real user monitoring
- OneAgent injects the RUM JavaScript into HTML responses automatically. Where injection is impossible you deploy the agentless tag manually. Content security policy must permit the beacon endpoint or data silently disappears.
- Captured data becomes user actions grouped by action naming rules, user sessions enriched with session properties, and experience metrics such as Apdex and web vitals (LCP, INP, CLS, TTFB).
- Session Replay records interactions for triage; masking rules decide whether input and text are captured. Treat replay data as personal data and check residency and opt-in settings before assuming anything is recorded.
- Mobile RUM adds app start time, crash reporting and network performance for iOS and Android.
Synthetic monitoring
- Browser clickpath monitors cover single-URL and multi-step journeys; HTTP monitors check endpoints; network availability monitors check reachability. They run from public locations or private locations via ActiveGate.
- Synthetic results are a separate data set from RUM. Use them for availability, SLA evidence and pre-production checks, not as a substitute for real user metrics.
Worked example
After a release, LCP regresses for mobile users in one region while desktop is flat. Segment user actions by device type and geolocation; the regression tracks a third-party script loaded conditionally for that region. Reasoning: LCP is a rendering metric driven by resource loading, so correlate the timing change with third-party requests and release timestamps in the same window rather than re-checking servers.
Traps that cost marks
- Comparing synthetic latency with RUM action duration as if they were the same measurement.
- Forgetting that injection failures (CSP, proxies) look like traffic drops.
- Assuming Session Replay always contains text and inputs regardless of masking configuration.
- Setting Apdex or action naming rules without considering that they restructure historical reporting.