// wiki / flag analytics

Turn flag decisions into answers.

A rollout tells you what shipped. Flag Analytics tells you who received it, why they received it, and whether the experiment earned its victory lap.

Start wide, then follow the interesting signal

Open Test → Flag Analytics for a portfolio view of every flag: evaluation volume, unique users, on/off outcomes, variant allocation, experiment status, and data-quality warnings. Search or filter the portfolio, then select one flag to open its decision workspace. Your flag, environment, time period, and other filters stay in the URL, so a useful investigation is easy to bookmark or share.

Every evaluation leaves clues

See the decision

Inspect the served value or variant, enabled state, decision reason, matched rule and conditions, environment, client, geography, user, and targeting context.

Follow a user

Expand user history to verify stable assignment, understand rule changes, and connect evaluations with conversion events for the selected flag.

Spot audience patterns

Break results down by variant, reason, rule, geography, client, and context. This is where “some users saw off” becomes a precise, fixable explanation.

Trust the warning lights

Allocation drift, crossover users, missing context, dropped exposures, unattributed conversions, and stale data are called out before they can masquerade as insight.

For experiment leads

Make the decision, not just the chart

Compare each treatment with its control using exposures, conversions, conversion rate or mean value, absolute and relative lift, confidence interval, and p-value. Choose 90%, 95%, or 99% confidence. When sample size or data quality is not ready, Maxlona withholds winner guidance and explains why—because “promising” and “proven” should never share a name tag.

1. Check qualityResolve crossover and attribution warnings.
2. Read the intervalJudge plausible impact, not only the p-value.
3. Decide togetherExport a review-ready evidence package.

Take the evidence with you

Exports honor the same flag, environment, date, audience, and decision filters shown on screen. Choose PDF for a polished review, XLSX for an analysis-ready workbook, CSV for filtered evaluation rows, or JSON for downstream tools and reproducible analysis.

Privacy travels with the data. Editors receive masked user identifiers and redacted custom context. Admins can explicitly reveal or export sensitive data after confirmation, and successful access is audited without recording the sensitive values themselves.

A useful five-minute investigation

  1. Select the flag and production environment, then compare the current period with the previous one.
  2. Check allocation and data-quality warnings before interpreting lift.
  3. Filter a surprising outcome by reason or matched rule, then inspect a few user histories.
  4. Review confidence intervals and sample readiness with the experiment owner.
  5. Export the filtered evidence for the launch review, experiment log, or follow-up analysis.

Login to your account, open Test → Flag Analytics to begin.

Related
A/B testing →Feature flags →Segments →Audit log →