Throughput time
Elapsed production flow time: boundaries, processing and waiting, calendar conventions, batch completion, WIP relationships and improvement measurement.
Tags: production, planning, flow-time, lead-time
On this page
A lot released 08:00 on day 1 and fully accepted 11:00 on day 2 takes 27 hours elapsed. Only 3.5 h are processing. The sequential example also includes 2 h queueing, 0.5 h setup, 0.5 h moving, 20 h waiting including overnight closure, and 0.5 h inspection. Retain both boundary events when comparing flow time with cycle time or customer lead time.
LEI uses production lead time, throughput time and total product cycle time for elapsed passage. Processing time counts work on the item. Lean Enterprise Institute, Cycle Time lexicon
State both endpoints and the object. Throughput rate instead counts units per time.
Define the object and boundary
Choose whether the observation represents an individual piece, a production lot, an order or another object. Identify the relevant revision, route, quantity and completion rule. The first completed piece and the last accepted piece of a lot generally have different timestamps.
| Measurement decision | Example of a precise definition |
|---|---|
| Object | One identified production lot |
| Start | Lot released to the controlled shop-floor route |
| Finish | Last required piece accepted and posted to finished stock |
| Time basis | Continuous elapsed hours, including nights and weekends |
| Exceptions | Rework remains within the span; cancellations reported separately |
Use the same measurement boundaries for comparable observations.
LEI separately describes order lead time and order-to-cash time. Neither should be substituted for a production-only measurement without changing the stated boundary. Its machine and operator cycle-time definitions also concern repeatable work at a station rather than the entire door-to-door journey. LEI, related time definitions
Separate activity time from elapsed time
A route can include waiting for a resource, setup, processing, teardown, movement, inspection and mandatory waiting such as drying. SAP's production-order scheduling lesson explicitly models queue, setup, processing, teardown, wait and move time, with additional production floats and goods-receipt processing. These are scheduling model elements; planned values are not automatically actual observations. SAP, Configuring Production Order Scheduling
For a sequential timeline, mutually exclusive segments can be added to reconcile elapsed time. For parallel work, adding every resource's occupied hours can exceed the item's elapsed duration. Distinguish labor-hours, machine-hours and item-hours. A two-hour activity using three operators is not six hours of elapsed time.
Decide whether inspection counts as processing, a separate interval or the finish event. Count each elapsed interval once and mark unexplained time.
Original timeline example
Assume a hypothetical lot released at 08:00 on day 1 and fully accepted at 11:00 on day 2. Continuous elapsed throughput time is 27 hours. Its sequential, nonoverlapping history is:
| Interval | Duration | Classification |
|---|---|---|
| Day 1, 08:00–10:00 | 2.0 h | Queue |
| 10:00–10:30 | 0.5 h | Setup |
| 10:30–12:30 | 2.0 h | Processing |
| 12:30–13:00 | 0.5 h | Move |
| Day 1, 13:00–day 2, 09:00 | 20.0 h | Wait, including overnight closure |
| Day 2, 09:00–10:30 | 1.5 h | Processing |
| 10:30–11:00 | 0.5 h | Inspection/acceptance |
| Total | 27.0 h |
Processing occupies 3.5 / 27 = 12.96% of that elapsed span. This is a processing-time fraction, not automatically a value-added fraction: that classification would require its own definition of customer value.
If the reporting calendar instead counts only 08:00–17:00 on each day, the same start and finish span 9 + 3 = 12 working hours. Both figures can be arithmetically correct, but they answer different questions. Removing nights from a dashboard does not shorten the customer's elapsed wait.
As an original sensitivity example, reducing the 20-hour waiting interval by four hours, with all other intervals unchanged, reduces throughput time to 23 hours, a 14.81% reduction. Saving 10% of the 3.5 processing hours instead saves 0.35 h, reducing elapsed time to 26.65 h, or 1.30%. These are assumed scenarios, not evidence that the waiting interval can actually be removed.
Relate WIP, rate and time carefully
John D. C. Little's 1992 chapter states the steady-state relationship L = λW: average items in a system equal average arrival rate multiplied by average time in that system, under its applicable conditions. Use the same boundary, item population and time basis for all three quantities. Little, Are There “Laws” of Manufacturing?, National Academy of Engineering
For an original stable-flow example, suppose a defined route contains an average 120 units and its sustained arrival and completion rates are both 30 units per working day, with no scrap or population changes. The corresponding mean time is 120 / 30 = 4 working days. This is a system-average consistency relationship, not the promised duration of the next individual order.
Do not divide a single WIP snapshot by one unusually productive shift and call the result a measured mean. Growth or depletion of inventory, scrap exits, rework loops and mixed units require explicit treatment. Nor does the equation prove that an arbitrary WIP reduction can preserve the same throughput rate; the production system still needs a feasible operating policy.
Measure actual flow without selection bias
Capture reliable entry, exit and intermediate events, including corrections and their time zones. Separate planned dates from actual events and distinguish “operation finished” from “quantity accepted.” For split lots, preserve relationships between parent and child quantities so the final-completion measure remains meaningful.
Report the observation period, sample size and distribution, not only the mean. An average can conceal a long tail of delayed orders. If reporting only completed lots, also show aging open WIP: excluding everything still stuck in the process can make performance appear better while the backlog worsens.
Segment comparisons by product, route, lot size and priority where those affect flow. A shift toward easier jobs can lower the aggregate average without improving any route. Keep expedite handling visible because moving one order ahead can delay another.
Use the result for improvement and planning
Map where time is spent and investigate the largest controllable delays. Check whether apparent waiting represents material shortage, unavailable capacity, batch accumulation, required process dwell or delayed administrative posting. Improving the timestamp alone does not improve physical flow.
NIST's supply-chain flow-time research uses manufacturing inventory and inter-industry interactions at an industry-level boundary. It illustrates that flow time can be analyzed beyond one factory; its aggregate results should not be treated as measured lead times for an individual shop or order. Thomas and Kandaswamy, NIST publication record, December 4, 2017
A planning handoff should include the object/boundary, calendar, timestamp sources, quantity and route scope, actual time distribution, open-WIP aging and known exclusions. Use those observations alongside capacity, material availability and scheduling rules when estimating future completion; historical average throughput time alone is not a finite-capacity schedule.
Source versions
Reviewed 2026-11-02 PST. Little's cited chapter is from the 1992 National Academy of Engineering volume. NIST's record identifies publication December 4, 2017 and webpage update June 24, 2021. No separately verified technical issue date is assigned to the retrieved LEI lexicon or SAP lesson. Numerical examples and measurement recommendations are original, bounded guidance.