Identifying Outlier Values in Session Calculations
In the screenshot I shared, there are some outlier values. I also wrote down how the session is calculated.
In the screenshot I shared, there are some outlier values. I also wrote down how the session is calculated.
Hi Rachel V., Please stand by as I search the Mixpanel knowledge base for an AI-powered answer. I normally take about 15 seconds or so. In case AI cannot answer your question, feel free to reach out to Mixpanel support at mixpanel.com/get-support. If you have follow-up questions for this thread, please tag me again with your question: @kapa.ai <follow-up question> 👋
Feel free to ask a follow-up question once I've answered by mentioning me with @kapa.ai
Thanks for flagging the outliers. A few things to check to validate session calculations and explain spikes:
- Confirm which session type is configured (Timeout-, Event-, or Property-Based) and its constraints. Sessions reset at midnight in the project timezone and max length is 24 hours, which can create edge cases around day boundaries. [Sessions; Deep-Dive; Event-Based; Property-Based]
- For Timeout-Based sessions, a non-excluded event within the timeout extends the session; the end time is the last event before inactivity. Midnight reset can split or truncate sessions and produce 0-length or unusually short/long durations in charts. [Deep-Dive; Examples]
- For Event-/Property-Based sessions, midnight reset can break start/end pairing (e.g., Login before midnight, Logout after) resulting in missing durations or unexpected counts. [Event-Based; Property-Based]
- Verify excluded events and inherited session properties; background/system events that fire periodically can keep sessions “alive” and inflate durations if not excluded. [Deep-Dive; FAQ exclusions; Session properties]
- In Insights, ensure you’re measuring “Total Sessions” vs “Total Events” or “Unique Users.” Mismatched measurements often look like outliers. Use the Sessions aggregation (“Sessions with event”) if you’re asking “how many sessions contained X?”. [Measurements; Total Sessions; Session controls in reports; Data discrepancies]
- Timezone differences can cause apparent spikes at day boundaries; confirm project timezone. [Data discrepancies]
Quick diagnostics you can run:
- In Insights, chart Session Start over time and break down by “Session Start Event Name” or “Session Duration (Seconds)” to see which sessions drive the outliers; look for clusters at exactly midnight or very long durations. [Session controls in reports] - Filter out known background/API events and re-check. [Deep-Dive] - In Funnels, switch counting to Sessions and review Time to Convert to sanity-check your timeout setting. [Session controls in reports; Deep-Dive]
If you can share which session type and timeout you’re using, plus the exact metric in the screenshot (e.g., Total Sessions, Avg Session Duration per user), I can point to the precise cause.
kapa.ai is still learning and improving, please let me know how I did by reacting below.
ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.