Both files are ready. Everything was processed locally — no data left this environment.
Results
Table
Copy
team	open	resolved	median resolution (h)
Accounts	2	4	8.50
Billing	2	5	7.00
Technical	2	5	4.00
Exceptions: one record — T05 (event E05-2) is counted in Technical's resolved total but excluded from the median because its final resolved snapshot has a blank resolved_at_utc.
Method
Deduplication — removed one exact repeated row (E06-2 for T06 appeared twice, byte-identical).
Cutoff filter — dropped snapshots after the inclusive cutoff 2026-09-23T18:00:00Z. This removed E09-3 (T09, 19:00Z), so T09's final state is its 17:00Z open snapshot, not resolved.
Latest snapshot per ticket by absolute instant — timestamps were parsed as ISO 8601 and compared as instants, never as strings. This mattered twice: T10's 07:00:00+02:00 (= 05:00Z) open snapshot lost to the 06:30Z resolved snapshot, and T11's 06:00:00-03:00 (= 09:00Z) resolved snapshot beat the 07:30Z open one. Final team and status come from that winning snapshot — e.g. T04's resolution is attributed to Accounts (its final team), not Billing.
State and counts — current state per ticket, not event counts. T03 was resolved then reopened, so it counts as open in Billing. All 20 tickets were opened before the cutoff, including T08 (opened 20 Sep, before the reporting label window) — the "21–23 September" label was not used as a filter.
Durations and median — only final-resolved tickets with valid opened ≤ resolved ≤ cutoff contribute (resolved − opened) in elapsed hours. Median uses the middle value, or the mean of the two middle values for even-sized sets (Accounts: mean of 8.0 and 9.0 = 8.50; Technical: mean of 3.0 and 5.0 = 4.00), rounded to two decimals. Open tickets (including reopened T03) were excluded from durations silently, per the dictionary — they are not missing-data exceptions.
Operational limits
Small samples — each team's median rests on 4–5 durations, so a single ticket moves it substantially; treat these as point-in-time indicators, not stable benchmarks.
Missing data is invisible in the median — T05 counts as resolved but contributes no duration, so Technical's median describes only the 4 tickets with complete timestamps.
Snapshot granularity — states between snapshots are unknown; a ticket resolved and reopened between two events would be invisible. Reopened tickets also retain their original opened_at, so if T03 resolves later its duration will span the full lifetime including the open gap.
Fictional fixture — per the dictionary, all records are synthetic; figures are valid only within this exercise.
summary.csv
Preview file
Preview


exceptions.csv
Preview file
Preview