SAP Tutorials Platform

Appearance

Reading branch telemetry ​

When you author a [BRANCH BEGIN] block (see branched-tutorials.md), the platform records which branch each learner picks. This doc covers how to read that data and when to act on it.

Where to find it ​

Open the Missions Fiori app → pick a mission → Branch Performance section near the bottom of the ObjectPage. The section is empty until learners have actually encountered a branch in your mission.

The same data is queryable via the Analytics Explorer (/analytics-ui/) — pick AnalyticsBranchPerformance or AnalyticsBranchTopPick from the entity list.

Two service surfaces, same data. Admin users (Tutorials Admin role-collection) read these views via /admin/analytics/AnalyticsBranchPerformance. Authors (Tutorials Author role-collection) read the same views via /author/AnalyticsBranchPerformance. CAP cannot grant a single service to both audiences (@requires AND-combines with entity @restrict), so the platform projects each view into both services. The Mission ObjectPage uses the admin path; the branch-staleness lint rule uses the author path.

What the columns mean ​

  • Total Decisions — how many times the branch was rendered (one row per learner per visit).
  • Click Rate — clicks / total. The fraction of renders where the learner explicitly chose a branch (vs walking past).
  • Follow Rate — followed / clicked. Of the learners who clicked, how many took the recommendation. Low follow-rate means your recommendation logic is suggesting the wrong path.
  • Top Pick — the most-picked branch and its share of all picks. Format: hana (96%).
  • By Condition / By Ranker / By Default — breakdown of how the recommendation was determined. "By Default" means no condition matched and no ranker was registered.
  • Via Joule / Via Page Load — surface breakdown. High Joule share means learners are asking the chatbot for guidance instead of using the page.

When to act on the staleness lint ​

The branch-staleness lint rule fires (severity: notice) when:

  • The branch has been live ≥30 days
  • It has ≥50 decisions logged
  • One option has been picked >95% of the time

That's a strong signal that the branch isn't earning its keep — readers consistently pick the same path. Options:

  1. Inline the dominant path and remove the branch entirely.
  2. Rephrase the choice so the underrepresented path is more attractive (or more clearly relevant).
  3. Move the choice up or down the tutorial — maybe learners are over-fixated by the time they reach it.

The lint never blocks the build. It's a quarterly review prompt, not a CI gate.

Privacy and retention ​

Telemetry rows carry no learner-identifying data in the views — BranchDecisions itself includes user_ID, but the analytics views aggregate it away. Authors with the Tutorial.Author role collection see only the aggregated counts.

Raw BranchDecisions rows participate in the standard CAP @PersonalData anonymization cascade (see docs/developers/architecture/anonymization-cascade.md).