An internal request takes two hours to reach accepted delivery. That does not tell an operations manager how long someone spent preparing it. Before promising a faster service, measure the preparation separately from review, documented waiting and time nobody has classified.
The MAJLS timing-boundary worksheet below is an original operating recommendation for that task. It is not a deployed feature, productivity benchmark or staffing assessment. Every request, person, timestamp and calculated result in the example is synthetic and illustrative.
Choose the question and the clock
The Lean Enterprise Institute distinguishes processing time, when a product or order is actually being worked on, from production lead time, its passage through a process from start to finish.[1] These definitions come from its Lean lexicon; we are borrowing the distinction, not imposing manufacturing terminology on office work.
GOV.UK's Service Manual advises choosing measurement methods for the question and using several data sources rather than digital analytics alone.[2] Its context is UK public services, not a requirement for your business. Our recommendation: combine source-event timestamps with authorized interval observations before making an effort claim.
Use plain labels. Here, elapsed duration means received-to-accepted-delivery calendar minutes. Observed hands-on time includes preparation, review and rework. Hands-on is not automatically value-creating time. A gap in the record stays unclassified; it does not become waiting merely because subtraction leaves a remainder.
Timing-boundary worksheet template
Complete this boundary block with the measurement owner before observing real work:
- Define one item, its included service class and the received event. Do not mix routine formatting with a research request.
- Define accepted delivery and its evidence locator. A first draft or a returned draft is not a finish.
- Record the observation window, fixed cutoff, timezone and clock basis. A business-hours clock needs its own working calendar; this example uses calendar minutes only.
- Name the measurement owner, authorized data sources and interval-observation method. Collect minimum work-item evidence, not individual surveillance or staff rankings.
- Record each interval's state, start, end and source-event locators. Keep missing events, exclusions, resolver and next check visible.
Copy these CSV headers into a worksheet; use one row per item and another sheet with one row per interval:
case_id,service_class,received,accepted,received_event,accepted_event,cutoff,measurement_owner,boundary_approval
interval_id,case_id,state,start,end,start_event,end_event,evidence_note,synthetic
Leave unknown ends blank. Do not enter zero. Confirm boundaries and classifications with the people doing the work, and keep the source log unchanged when preparing calculations.
Filled example: five formatting requests
The fictional team formats internal reference sheets. All times below are on 7 October 2026 in UTC; the fixed cutoff is 12:00. Alex Morgan is the fictional measurement owner. The example deliberately uses sequential, non-overlapping single-handler intervals for each item.
P means observed preparation; R means observed review; C means observed corrections/rework; W means documented waiting. Each interval has separate start/end locators in the companion event log, such as T1-E04/T1-E05 for T1 preparation.
| Item | Received → accepted | Recorded intervals, in chronological order |
|---|---|---|
| T1 | 09:00 → 11:00 | W 09:00–09:20; P 09:20–09:35; W 09:35–10:50; R 10:50–11:00 |
| T2 | 09:00 → 10:30 | W 09:00–09:10; P 09:10–10:10; W 10:10–10:20; R 10:20–10:30 |
| T3 | 09:00 → 12:00 | W 09:00–09:15; P 09:15–09:45; W 09:45–10:30; R 10:30–10:40; C 10:40–11:00; W 11:00–11:50; R 11:50–12:00 |
| T4 | 10:00 → still open | W 10:00–10:20; P 10:20–10:40; W 10:40–12:00 cutoff |
| T5 | 09:00 → 10:00 | W 09:00–09:10; P starts 09:10, end missing; R 09:50–10:00 |
T3's first draft is ready at 09:45 and returned at 10:40. Its original received time remains 09:00. Do not reset the clock after corrections. T4 has an age at cutoff, not a completed duration. T5 has accepted-delivery evidence but an incomplete preparation record.
For this sequential example, calculate:
elapsed = accepted delivery - received
open age = cutoff - received, only when accepted delivery is absent
observed hands-on = sum of complete P, R and C intervals
unclassified = elapsed (or open age) - hands-on - documented waiting
Validate chronology, nonnegative durations, source bindings, coverage and no overlap first. If handlers work concurrently on the same item, distinguish elapsed interval unions from summed person-effort; do not apply this partition unchanged.
| Item | Completed elapsed | Open age | Observed hands-on | Documented wait | Unclassified |
|---|---|---|---|---|---|
| T1 | 120 | — | 25 | 95 | 0 |
| T2 | 90 | — | 70 | 20 | 0 |
| T3 | 180 | — | 70 | 110 | 0 |
| T4 | — | 120 | 20 | 100 | 0 |
| T5 | 60 | — | 10 observed only | 10 | 40 |
All values are calculated calendar minutes. T5's 09:10–09:50 remainder is unclassified, not idle time; its preparation subtotal is unavailable. Ask Alex for the missing end observation. Compare completed, fully observed T1–T3 only; show T4 and T5 separately with their exclusions. This small synthetic cohort does not establish typical service performance.
Sensitivity example: make one preparation interval zero
For T1 only, hypothetically remove its 15-minute preparation duration. Keep every other interval duration unchanged and shift downstream timestamps consistently:
| T1 interval | Original | Hypothetical |
|---|---|---|
| Initial waiting | 09:00–09:20 | 09:00–09:20 |
| Preparation | 09:20–09:35 | 09:20–09:20 |
| Review waiting | 09:35–10:50 | 09:20–10:35 |
| Review and acceptance | 10:50–11:00 | 10:35–10:45 |
Elapsed duration becomes 105 minutes rather than 120. Documented waiting remains 95 minutes and observed review remains 10. This is an assumption-bound sensitivity illustration, not realized savings, a forecast or a universal upper bound. It models neither queue interactions nor behavior changes. Zero preparation is a hypothetical input, not a claim that automation can achieve it.
Pending human decision
| Owner | Next review | Proposed work | Approval state |
|---|---|---|---|
| Alex Morgan, fictional measurement owner | 13:00 UTC | Confirm the boundary; seek T5 evidence; choose whether to examine T1 preparation or its documented review wait | Pending; no change authorized |
Long waiting does not establish its cause or justify removing review. Missing ends, inconsistent clocks, overlaps and undeclared concurrency limit the affected calculation; a model confidence score cannot clear them. The named human process owner retains consequential approval for workflow changes, control removal, staffing actions, circulation and customer commitments.
The private review package includes reusable CSVs, source-event locators and the executable calculation with failure probes. These are companion review files, not advertised live downloads. You can use the boundary block, filled intervals and formulas above directly. For admission limits, use Limit AI work to what your reviewers can finish; this worksheet measures intervals rather than changing that contract.
GOV.UK material: Crown copyright. Contains public sector information licensed under the Open Government Licence v3.0. No LEI chart is reproduced.