Window: 7d · Generated:
Trends compare 24h cohorts derived by differencing the 1d/2d/3d windows: they measure the items that arrived in each bucket, not the KPI as it read 24h ago. Backfill lands in the cohort of its own timestamp, so the two diverge. Panel values above still reflect the selected 7d window.
| geo_source | Count | Share |
|---|---|---|
| coord | 223 | 40% |
| text | 166 | 30% |
| none | 60 | 11% |
| neighbor | 52 | 9% |
| street | 48 | 9% |
| nominatim | 2 | 0% |
Sizes the responsibility mapping from the live store (lage-p6pi.2). A window dominated by <none> means the stored items carry no service_code, not
that the mapping is short — the two need different fixes. It does not mean the
source dropped the field: anliegen.bonn.de was measured clean on 2026-08-14 (354/354 open
records carried a code), so a spike points downstream of the fetch (lage-pjm3).
| service_code | Count | Share |
|---|---|---|
| 1.5 | 47 | 59% |
| 2.3 | 15 | 19% |
| 3.1 | 5 | 6% |
| 3.2 | 3 | 4% |
| 3.4 | 3 | 4% |
| 3.3 | 2 | 3% |
| 2.5 | 1 | 1% |
| 2.7 | 1 | 1% |
| 3.5 | 1 | 1% |
| 3.7 | 1 | 1% |