[Labelling Health] Labelling Health Report – 2026-04-06

[mnkiefer]

Summary

Overall status: Mixed — system newly operational, trending positive but no prior-window baseline

The auto-labelling pipeline became active on 2026-03-31. Daily runs have been fully reliable since then, with strong throughput. However, the previous 7-day window (2026-03-23–2026-03-29) contains no data, so trend direction cannot yet be confirmed. The correction backlog was small and has been fully resolved.

---

Key Metrics

Metric	Value
Discussions reviewed — last 7 days	~130 (across 4 active days: Apr 5, Apr 4, Apr 3, Mar 31)
Label changes applied — last 7 days	~66
Change rate — last 7 days	~51% (66 ÷ 130)
Comparison with previous 7-day window	⚠️ No data in previous window — baseline not yet established
Correction-collector runs — last 7 days	3 (all success, all on 2026-03-31)
Open correction signals	0
Correction signals created — last 7 days	2
Correction signals created — last 30 days	2
Oldest open correction signal	N/A — none open

Note on precision: Daily summary issues do not have pre-parsed reviewed/changed fields; counts are extracted from issue body text. Three summary issues were generated on 2026-03-31 (manual + scheduled runs); the earliest two may overlap in scope. The figures above use the latest-per-day run as the primary data point for Apr 3–5, and aggregate conservatively for Mar 31. The previous-7-day window is empty — the system had not yet been deployed.

---

Correction Pressure

Correction pressure is minimal and geographically concentrated: both signals originated in a single intake batch (Batch 01, closed 2026-04-03), both created on 2026-03-31.

The two resolved signals reveal two specific labelling gaps:

• Discussion #98 ("How do I debug GitHub Actions matrix builds failing only on arm64?"): The auto-labeller assigned Code Search and Navigation but missed Actions and the question type label. The trusted correction added both. Category: Other Feature Feedback, Questions, & Ideas.
• Discussion #118 ("Zero support from Github"): The auto-labeller did not assign bug despite the discussion being filed in the Bug discussion type within the Enterprise category. The trusted correction added bug.

Both corrections are labeled events (additions only, no removals observed). Pressure is currently low, concentrated in one batch, and fully resolved.

Open correction signal details

No open correction signals at time of report.

Closed signals (last 30 days):

Issue	Discussion	Category	Labels corrected	Closed
#29	#98 — Actions/arm64 debug	Other Feature Feedback, Questions, & Ideas	Added: Actions, question	2026-04-03
#27	#118 — Zero support	Enterprise	Added: bug	2026-04-03

---

Open Instruction Debt

The correction backlog is currently at zero open signals. The single intake batch (#26, Batch 01) and both sub-issues were closed on 2026-04-03, approximately 3 days after creation.

• Open parent intake issues: 0
• Open correction signal issues: 0
• Oldest open signal age: N/A

The backlog is fully cleared, but it is very early in the system's lifecycle (active for ~6 days). The two resolved corrections indicate specific instruction gaps around (a) Actions-category recognition within the Other Feature Feedback category and (b) bug label application in Enterprise discussions filed as Bug type. These have not yet been reflected in updated labelling instructions, creating latent instruction debt.

---

Recommendations

1. Update community-discussion-labeling.md to address the Actions label gap. Discussion #98 shows that GitHub Actions content in the Other Feature Feedback, Questions, & Ideas category is not reliably picked up by the auto-labeller. Add or strengthen rules for detecting Actions-related keywords (matrix, runners, self-hosted, workflows) and mapping them to the Actions label.

2. Clarify bug label application in Enterprise discussions. Discussion #118 shows that discussions filed under the Bug discussion type in the Enterprise category are not receiving the bug label automatically. Update instructions to explicitly handle category × discussion-type combinations, particularly Enterprise + Bug type → bug.

3. Investigate the duplicate summary issues on 2026-03-31. Three daily summary issues were generated that day (issues #25, #28, #31), suggesting the Label Discussions workflow was triggered multiple times (manual dispatch + scheduled). Confirm trigger configuration is idempotent or add deduplication guards to avoid inflated historical counts.

4. Establish a 7-day baseline before evaluating trend direction. The system has been live for only ~6 days. At the next health report cycle (2026-04-08), a prior-window comparison will be possible. Until then, treat the current rate (~51% change rate, 7/7 successful daily runs) as the baseline rather than a confirmed trend.

---

Recent daily summary issue breakdown

Issue	Date	Discussions reviewed	Label changes applied	Change rate
#38	2026-04-05	40	23	57.5%
#36	2026-04-04	40	17	42.5%
#34	2026-04-03 (evening, scheduled)	40	20	50.0%
#33	2026-04-03 (morning, manual?)	40	20	50.0%
#31	2026-03-31 (evening, scheduled)	10	6	60.0%
#28	2026-03-31 (midday)	40	26	65.0%
#25	2026-03-31 (morning)	~30	28	~93%*

*Issue #25 body did not contain a standard "Discussions reviewed" line; count estimated from context.

Multiple runs on 2026-03-31 are likely due to manual workflow dispatches during initial system setup. The evening scheduled run (#31 with only 10 reviewed) may reflect a smaller batch after earlier runs consumed most of the backlog.

Recent workflow run references

Workflow	Runs (last 7 days)	Success	Skipped/Failed
Label Discussions	7	7	0
Labelling Correction Collector	3	3	0
Labelling Correction Feedback	13	1	12 skipped

The high skip rate for Labelling Correction Feedback is expected when no correction signals are pending resolution.

References

• §29 — Label Discussions run #29 (2026-04-05, scheduled)
• §28 — Label Discussions run #28 (2026-04-04, scheduled)
• §17 — Labelling Correction Feedback run #17 (2026-03-31, success — Batch 01 feedback)

Generated by Labelling Health Report · ● 369.7K · ◷

• expires on May 6, 2026, 3:37 AM UTC

[mnkiefer]

This issue is being closed as outdated. A newer issue has been created: #41

View newer issue

---

This action was performed automatically by the Labelling Health Report workflow.