Three plants in the same group report dock-to-line time. One reads 41 minutes, one reads 3.2 hours, one reads 1.8 hours. Nothing is wrong with any of them and none of them can be compared, because the first starts its clock at the door and stops at goods receipt, the second starts at gate-in and stops when the part reaches the station, and the third excludes anything that went into a quality hold. The group scorecard averages all three into a number that describes nothing. This is the normal condition of plant logistics measurement, and it is why KPI programmes tend to produce meetings rather than improvement — the definitional argument consumes the time that should have gone to the actual problem. What follows is five metrics that matter in a plant, each with its clock boundaries stated, its exclusions named, and its gaming risk written down, because a KPI without a documented gaming risk is one somebody is already gaming. See it on your own data to pressure-test your definitions before they are published.
2026 FRAMEWORK · PLANT LOGISTICS
A Plant Logistics KPI Framework That Holds Up
Five metrics with honest definitions — where each clock starts and stops, what is excluded and by whom, how each one gets gamed, and what a defensible target looks like.
Same Metric, Three Plants, Three Answers
Before the definitions, the reason they matter. This is one metric measured three defensible ways, and it is the single most common cause of a group logistics scorecard that nobody trusts.
Plant A
41 min
Clock runs door-in to goods receipt posted. Measures the dock, not the flow.
Plant B
3.2 hrs
Clock runs gate-in to part available at the station. Includes yard, dock and internal transport.
Plant C
1.8 hrs
Same as B, but excludes anything that entered a quality hold. Roughly a fifth of loads vanish from the denominator.
The rule that fixes most of this
Publish the clock boundaries and the exclusion list with every metric, in the same document, at the same time. Not the target — the definition. A target attached to an undefined metric is an invitation to redefine the metric, and that is exactly what happens the first quarter somebody misses.
Metric 1 — Dock-to-Line Time
The inbound flow metric. Note that this is deliberately not dock-to-stock, which is the warehouse version and stops at putaway — in a plant the question is when the part is available where it gets consumed, which is a different and longer clock.
DefinitionElapsed time from trailer arrival at the door to the part being available at its point of consumption — line-side rack, kitting cell or sequence buffer.
Clock startsTrailer spotted at the assigned door. Not gate-in — yard time is a separate metric with a separate owner.
Clock stopsPart scanned available at point of consumption. Not goods receipt posted, which can happen well before the part is physically usable.
ExcludesQuality holds, but the excluded count is published alongside. An exclusion that is not counted is a deletion.
Gamed byPosting goods receipt early to stop the clock while the material sits on a dock apron. Detect it by comparing receipt timestamp against the first consumption scan.
ReferenceWarehouse dock-to-stock benchmarks put strong performance under four hours for standard shipments; a plant measuring to point of consumption should expect to sit above that and should set its own baseline rather than importing the number.
Metric 2 — Sequence Adherence
The plant-specific metric that no generic logistics KPI set contains, and the one most directly tied to whether the line runs. Everything else on this page is a leading indicator of this.
DefinitionShare of sequenced units delivered to the station in the correct order, at the correct position, within the window — all three conditions, not any of them.
Clock startsSequence window opens at the station. The frozen build order is the reference, not the plan as amended afterwards.
Clock stopsUnit consumed at the station, or the window closes without it. A late-but-correct delivery is still a miss.
ExcludesNothing. Every exclusion proposed for this metric is an argument to stop measuring the thing that matters.
Gamed byCounting a manual sort as adherence because the right part eventually reached the station. Record sorts separately — recovered misses are misses that cost labour.
ReferenceReport as parts per million rather than a percentage. At sequenced volumes a 99.5% adherence rate sounds respectable and represents thousands of interventions a year.
Every one of these definitions will be argued with. Better in a document than in a review.
We will run your existing metric definitions against your own event data and show where the clocks, exclusions and denominators are producing numbers you cannot compare across sites.
Metric 3 — Premium Freight Rate
The cost metric finance already tracks and logistics rarely owns properly. Its value is not the total — it is the split between avoidable and unavoidable, which almost nobody maintains.
DefinitionPremium and expedited freight spend as a share of total inbound freight spend, split by root cause and by owning function.
Measured overRolling twelve months. Monthly readings are too volatile to act on and quarterly readings hide the pattern.
Split requiredSupplier-caused, plant-caused, plan-caused and genuinely unavoidable. Unsplit, the number is a complaint rather than a metric.
ExcludesPremium freight contracted as standard for a lane. If it was planned and priced, it is not premium — it is the rate.
Gamed byClassifying plant-caused expedites as supplier-caused, since nobody in the room represents the supplier. Require the root cause to name an internal owner or a specific supplier failure, never "supply chain".
ReferenceMid-market industrial manufacturers have been observed running around six per cent of inbound spend on expedites with no ability to say which were avoidable. If you cannot produce the split, assume the avoidable share is large.
Metric 4 — Dwell
Three distinct dwells get reported as one number at most sites, which is why dwell improvement programmes so often move the number without changing anything physical. Separate them or do not measure them.
DefinitionTime an asset spends stationary on site, reported as three separate clocks — gate queue, yard staging, and at-door — never blended.
Report at90th percentile, with the median alongside. The mean is structurally blind to the outliers that generate every detention invoice and every sequence exposure.
Segment byCarrier, load type and hour of arrival. An unsegmented dwell figure cannot tell you whether you have a carrier problem or an arrival-shaping problem.
ExcludesOff-site holding, which is tracked as its own clock. Merging held and staged time hides which of the two is failing.
Gamed byAdmitting trucks through the gate quickly to stop the queue clock, then holding them in the yard where the clock is less visible. Compare gate dwell against yard dwell — if one falls while the other rises, nothing improved.
ReferenceYard management deployments have been reported to cut average truck turnaround by 30–50% through sequencing arrivals and pre-staging dock slots — most of which shows up in these clocks rather than in unload speed.
Metric 5 — Container Cycle Time
The most neglected metric on this list and the one that quietly funds a large share of yard congestion. Returnable containers are an asset pool, and an unmeasured pool always shrinks.
DefinitionElapsed time for a returnable container to complete a full loop — supplier fill, inbound, line-side consumption, collection, return, refill.
Clock startsContainer despatched full from the supplier. Starting at plant arrival measures your half of the loop and blames the supplier for the rest.
Clock stopsSame container despatched full again. If you cannot identify individual containers, you are estimating a pool, not measuring a cycle.
Report withPool size, unreconciled count and yard positions occupied by empties. Cycle time alone does not show what the pool is costing you today.
Gamed byBuying more containers when the cycle lengthens. This makes the symptom disappear and the cost permanent — check pool growth against volume growth before accepting an improved cycle figure.
ReferenceSet the target from your own loop geometry — supplier distance, fill frequency and collection cadence. Imported benchmarks are meaningless here because the loop is physical and site-specific.
Rules That Apply to All Five
Definitional discipline that travels across every metric. Most KPI frameworks fail on these rather than on metric selection.
1Baseline before you target. Measure the existing process for a full cycle first — importing an industry benchmark without knowing your current position sets a target nobody can defend.
2Publish exclusions with counts. Every excluded record is visible or the exclusion list becomes a tuning dial.
3Name one owner per metric, by role. Shared ownership means the metric is discussed and never acted on.
4Percentiles for time metrics, rates for count metrics. Means belong nowhere near a distribution with a long tail.
5Set a volume floor below which no figure is published. Small denominators swing wildly and discredit the whole set.
6Track four to six aligned to your current pain, not every metric available. A framework nobody reads is not a framework.
7Improvement targets in the range of five to fifteen per cent per quarter for operational metrics. Larger commitments produce redefinition rather than change.
8Write the gaming risk down when you write the definition. It is far easier to do honestly before anyone is being measured.
Metrics That Look Useful and Are Not
Each of these appears on plant logistics scorecards and each has a specific defect. Removing them frees attention for the five above.
← Swipe to see all columns →
| Metric | Why it appears | Why it misleads |
|---|---|---|
| Total moves per shift | Easy to count, feels like productivity | Rehandles and empty running count the same as productive moves |
| Average dwell | Single number, fits on a slide | Structurally blind to the outliers that generate every actual cost |
| Dock utilisation, site-wide | Available from any scheduling tool | Averages saturated tier-restricted doors with idle general ones |
| Exceptions acknowledged | Looks like responsiveness | Acknowledgement is not resolution and measuring it flatters everyone |
| Supplier OTD, self-reported | Costs nothing to collect | Measured on the supplier's clock, so it disagrees with yours by design |
| Dashboard adoption | Demonstrates rollout success | A tool everyone opens and nobody acts on is the failure case, not the win |
| Goods receipt cycle time | Available directly from the ERP | Measures a posting, not a physical availability the line can use |
Who Sees What, How Often
The same metric at three cadences serves three purposes, and collapsing them is how a shift-level number ends up in a board pack where it means nothing. Our analytics and reporting module handles the distribution.
Shift · Operational
Live dwell clocks, door state, move queue, open exceptions. Read by yard supervisors and dock leads to act inside the shift. Never aggregated upward in this form.
Week · Tactical
Dwell percentiles by carrier, sequence adherence in ppm, exception counts by code, slot adherence. Read by plant logistics management to find patterns and assign work.
Month · Strategic
Premium freight split by root cause, container cycle and pool size, dock-to-line trend, supplier performance against the escalation ladder. Read by plant leadership and group.
Each tier should be derivable from the tier below it without manual reconstruction. If your monthly pack requires someone to rebuild numbers from spreadsheets, the definitions are inconsistent somewhere underneath — and the reconstruction is where the inconsistency gets quietly resolved in whatever direction suits.
Pressure-Test Your Definitions Before You Publish Them
Bring your current metric set and a month of event data. We will show where clock boundaries, exclusions and denominators differ across your sites, which of your metrics can be reconstructed from existing events, and which need capture that does not exist yet. The output is a definition document you keep either way.
Clock boundary mapping
Exclusion audit
Cross-site comparability
Three-tier reporting
Frequently Asked Questions
Is dock-to-line the same as dock-to-stock?
No, and conflating them is the most common definitional error in plant logistics. Dock-to-stock measures from arrival at the receiving dock until goods are available as warehouse inventory, covering unloading, inspection, documentation and system updates — a warehouse question. Dock-to-line runs further, to the point where the part is physically available where it will be consumed, which is what a plant actually needs to know. Track both if you like, but never report one under the other's name.
How many KPIs should a plant logistics function track?
Four to six aligned to your biggest current pain rather than every available metric. Long lists produce dashboards that are read and not acted on, and they dilute ownership until nobody feels responsible for any single number. Choose the ones that map to a decision somebody makes weekly. When the pain moves — say premium freight falls and sequence adherence becomes the constraint — change the set rather than adding to it.
Should we adopt published benchmarks as targets?
Not initially. Adopting industry benchmarks without understanding your current performance is a well-documented mistake — measure your existing process across a full cycle first, then set improvement targets from your own baseline, typically in the five to fifteen per cent per quarter range for operational metrics. Benchmarks are useful for sanity-checking whether your number is in a plausible range, not for setting a commitment. This applies doubly to container cycle time, where the loop geometry is entirely site-specific.
Why report dwell at percentile rather than average?
Because the average is blind to exactly the events you are trying to eliminate. A mean of two hours can comfortably contain a handful of nine-hour trailers, and those outliers are what generate detention invoices, carrier complaints and sequence exposure. Writing the target as a percentile — ninety per cent of trailers under four hours — moves the conversation to the worst cases immediately, which is where all the recoverable cost sits. The median alongside it shows whether the typical case is also drifting.
Truck turnaround looks like a transport metric. Why is it on a plant scorecard?
Because most of its variation comes from your side, not the carrier's. Dock scheduling, paperwork handling and unloading speed drive turnaround far more than anything happening on the road, which is why treating it as a purely transportation metric misassigns the ownership. Yard management deployments have been reported to reduce average turnaround by 30–50% through sequencing arrivals and pre-staging dock slots. Giving carriers visibility of their own waiting time also exposes dock bottlenecks that internal teams tend not to see.
What do we do when two sites disagree on a definition?
Let them keep both, but publish only one at group level and label the other as a local operational measure. Forcing definitional convergence in a single step usually fails, because each site's version reflects a genuine difference in layout or process. What must not happen is two definitions carrying the same name in the same report. Name them differently, state the boundaries for each, and converge over a couple of quarters if it is worth doing at all.
Can these be built from data we already have?
Mostly. Dwell, premium freight and dock-to-line are usually reconstructable from existing gate, dock and ERP events, though the clock boundaries often need correcting. Sequence adherence typically needs scan events at the station that many plants do not yet capture, and container cycle time needs individual container identity rather than pool counts — those two are the common gaps. Start with what exists, and treat the missing capture as a scoped project rather than a blocker. Our integrations overview covers the event surface.
Define It Once, Then Argue About the Result
Five metrics with stated clock boundaries, published exclusions, named owners and documented gaming risks — reported at three cadences that derive from one another instead of being rebuilt every month.
Builds on existing gate, dock and ERP events · Site-level configuration · Cross-site comparability
August 12, 2026
By Emily Parker
All Articles
