Warehouse KPIs: Formulas, Worked Examples, and Which Benchmarks Actually Compare

Which warehouse KPI statistics and reference figures matter?

A 99% line fill rate can coexist with a 90% order fill rate on the same 100 orders. This reference’s constructed example shows why warehouse KPIs need a named denominator before their percentages can be compared; it is not a survey result. See the complete example and inputs.

1. A 99% line fill rate can sit on top of a 90% order fill rate for exactly the same orders. In Cincinnati Dock Door Repair Research’s constructed September 12, 2026 dataset, 10 incomplete lines spread across 10 of 100 orders produce 99.00% line fill, 99.00% unit fill and 90.00% order fill across 1,000 lines. This is an arithmetic illustration, not an industry survey. Inputs and calculation.

2. 15.0 hours is APQC’s public median for dock-to-stock cycle time. Its displayed sample is 3,560, labelled “All Companies,” and its public page does not disclose the observation years; accessed September 12, 2026. This median is not a best-in-class threshold. 1

3. Four component rates of 95% produce an identical 81.450625% perfect-order component index whether the observed joint success rate is 95% or 80%. Cincinnati Dock Door Repair Research’s two constructed 100-order scenarios, checked September 12, 2026, use APQC’s multiplication method and differ only in which orders fail. These are synthetic examples, not measured industry results. 3 Inputs and calculation.

4. General warehousing and storage averaged 5.2 recordable nonfatal cases per 100 full-time-equivalent workers over 2020–2024, against 2.6 for all private industry. The corresponding DART rates were 4.4 and 1.6. OSHA publishes these five-year averages in its 2026 warehousing directive, calculated from BLS data; verified September 12, 2026. 7

5. 88.0% is APQC’s public median for perfect order performance, with a displayed sample of 13,590 labelled “All Companies.” The measure multiplies four component rates; it is not a direct count of orders passing all four tests. Observation years are undisclosed; accessed September 12, 2026. 3

6. 94.0% is APQC’s public median for order fill rate, with a displayed sample of 3,218 labelled “All Companies.” Its formula counts complete orders, not complete order lines. Observation years are undisclosed; accessed September 12, 2026. 2

7. Six APQC measure pages displayed a median and sample size while their public 25th- and 75th-percentile cells showed dashes. Cincinnati Dock Door Repair Research checked all six pages on September 12, 2026; this describes the public-page display, not a finding about warehouse performance. 1 2 3 4 5 6

8. OSHA’s 2026 directive reports seasonally adjusted warehousing and storage employment rising from 882,100 in December 2015 to 1,836,200 in December 2025—more than double. These are the BLS-derived figures reproduced in the directive, not a newly extracted or revision-adjusted employment series; verified September 12, 2026. 7

9. OSHA’s revised warehousing National Emphasis Program took effect July 31, 2026, with a five-year expiration provision. That provision points to July 31, 2031, unless the instruction is changed or superseded. It supersedes the instruction effective July 13, 2023; verified September 12, 2026. 7

10. WERC’s 2026 DC Measures Report tracks 36 operational metrics. MHI, WERC’s parent organization, identifies this report scope in its public 2026 summary; the number counts metrics, not respondents, and does not supply the report’s numerical benchmark thresholds. Verified September 12, 2026. 13 16

11. The “as much as 55%” order-picking share of warehouse operating expense appears as an estimate in a 2007 literature-review abstract—not as a 2026 measured average. The abstract does not identify an underlying sample for that figure; de Koster, Le-Duc and Roodbergen’s paper was received August 29, 2005 and accepted July 17, 2006. Original page checked September 12, 2026. 17

12. Travel 50%, search 20%, pick 15%, setup 10% and other 5% is a historical time distribution shown in Figure 4 of de Koster and colleagues’ 2007 review. The figure attributes the distribution to Tompkins and colleagues’ 2003 textbook and describes a typical picker-to-parts warehouse. The review’s figure—not a separate copy of the textbook—was checked September 12, 2026. 17 18

13. The two safety KPIs in this 24-metric reference use a federally published rate base of 200,000 employee-hours. BLS defines the nonfatal and DART numerators, and OSHA explains the 100 workers × 40 hours × 50 weeks base. These are cases per 100 full-time-equivalent workers, not percentages; verified September 12, 2026. 9 12

14. OSHA reports more than 1,700 violations and approximately 37,410 workers removed from hazards during the first 18 months of its program launched in July 2023. These are OSHA’s reported enforcement results in the directive effective July 31, 2026, not independent counts by this publication; verified September 12, 2026. 7

15. The Cincinnati OH-KY-IN metropolitan area contains 15 counties across three states under OMB’s July 2023 delineation. Our September 12, 2026 crosswalk identifies federal OSHA as the main private-sector authority in its Ohio counties, Kentucky OSH in its Kentucky counties and IOSHA in its Indiana counties, subject to jurisdiction exceptions. The incidence-rate formula itself does not change at the state line. 22 25 26 27

A number without its denominator, clock and population is not a usable benchmark. APQC’s public dock-to-stock median is 15.0 hours; a median and an upper-performance threshold answer different questions. The original WERC numerical tables were not inspected in this verification pass, so this page does not convert figures from public reproductions into verified WERC benchmarks. The reference below keeps those access gaps visible instead of manufacturing a comparison. 1 13

By Cincinnati Dock Door Repair Research.

Cincinnati Dock Door Repair Research is the independent research and reference section of cincinnatidockdoorrepair.com.

Last verified: September 12, 2026. Source data years are listed separately from our verification date.

Where are the formulas, benchmarks and worked examples?

This reference contains 24 metric definitions, a benchmark-and-exclusions table, five reproducible examples and a nine-check definition crosswalk. The links below go directly to the relevant section or table; each metric and headline statistic also has its own permanent anchor.

Key figures · What a KPI means · 24 formulas · Inbound and docks · Inventory and storage · Fulfilment and picking · Cycle time and cost · Safety formulas · Dataset files · Attribution reference · Benchmark comparisons · Benchmark table · Federal safety table · What the data shows · Methodology · Five examples · Fill rates · Perfect orders · Inventory counts · Labour hours · Dock time · Nine comparison checks · Definition table · Historical figures · Provenance table · Choosing a dashboard · Ohio and Hamilton County · County/authority table · Why dates matter · Limitations · Questions · Sources

Metric anchors run from K01 to K24; headline figures from statistic 1 to statistic 15. Table and comparison IDs remain independent of any future wording changes.

What is a warehouse KPI?

A warehouse KPI is a measurement of how well a facility receives, stores, picks, packs and ships goods. The useful ones name four things: what is being counted, what it is being divided by, which clock or population the count is drawn from, and over what period. Strip any of the four away and the number stops being comparable — which is why APQC’s 15.0-hour dock-to-stock median is not interchangeable with an upper-performance threshold or a differently timed receiving measure. 1

A few plain definitions used throughout this page:

• Order line. One entry for a product and requested quantity on an order. In the examples here, an order with one entry for each of three products has three lines; a real system can carry separate lines for the same product. 17

• Elapsed hours. Wall-clock time. Three people working an eight-hour shift produce eight elapsed hours.

• Person-hours. Labour time summed across workers. The same three people produce 24 person-hours.

• Median. The 50th-percentile value. With ties, half the observations need not be strictly higher and half strictly lower; for an even-sized ordered list, a conventional sample median averages the two middle observations.

• Best-in-class. A label for a high-performing comparison group, not a universal standard. WERC presents benchmarks on a quintile scale; a quintile represents one fifth, or 20%, of an ordered distribution. The original threshold values are not reproduced here. 13 16

• Incidence rate. Cases normalised to 200,000 employee-hours, conventionally expressed per 100 full-time-equivalent workers. State the case period; this is not the percentage of staff injured. 9 12

What are the formulas for these 24 warehouse KPIs?

This reference contains twenty-four warehouse KPIs, each with the formula, the unit, and the boundary that has to stay visible for the number to mean anything. The two safety measures use federally published incidence-rate definitions; the Basis column identifies the named source or our explicit operational definition for every row. None is assigned a universal target: this source set does not establish a performance requirement that applies across product mix, order profile and labour model. 9 12 23

How to read the Basis column. S — the core formula comes from a named source, listed in the row. E — the formula is this publication's explicit operational definition, written down so it can be argued with. It is not a claim that an industry standard exists. A zero denominator means undefined, never 0% and never 100%. S identifies the source’s core measure, not an assertion that every cohort rule or implementation detail in our row is prescribed by that source; those choices stay visible in the boundary column.

Receiving and dock measurements

Source: APQC Measure 100677; Mecalux’s published dock-occupancy formula; Cincinnati Dock Door Repair Research operational definitions. Verified September 12, 2026. 1 20 23

The dock-to-stock definition and clock-comparison checks show why supplier receipt, truck departure and completed put-away must not be treated as the same event.

Inventory accuracy, turnover and storage

Source: Oracle, historical 11g Release 7 inventory-performance definitions; Modula’s turnover formula; APQC Measure 100784; Cincinnati Dock Door Repair Research operational definitions. Verified September 12, 2026. 19 21 5 23

Fulfilment, picking and order reliability

Source: APQC Measures 101445 and 101741; Cincinnati Dock Door Repair Research operational definitions. “Component index” is our distinguishing label for APQC’s multiplication method; APQC calls its measure “perfect order performance.” Verified September 12, 2026. 2 3 23

Internal cycle time and warehousing cost

Source: Mecalux’s warehousing-cost-per-order core formula; Cincinnati Dock Door Repair Research operational boundaries. Verified September 12, 2026. 20 23

Exposure-adjusted safety rates

Source: BLS, “How To Compute Your Firm’s Incidence Rate for Safety Management,” for the nonfatal H + I + J and DART H + I numerators; OSHA’s August 23, 2016 interpretation for the common rate base. Verified September 12, 2026. 12 9

Both rates are educational references here. Fatal injuries are not erased by this nonfatal comparison: they are outside K23’s numerator and must not be described as absent because that rate is low. Hazard identification, abatement and recordkeeping decisions belong with a qualified safety professional and the issuing agency. 12 9

Dataset files and attribution reference

Warehouse KPI dictionary — CSV: 24 rows, including metric ID, group, name, formula, unit, numerator, denominator, clock or cohort, inclusion and exclusion policy, source IDs and URLs, formula origin, comparability warning, reporting cadence, source vintage, verification date, evidence tier and permanent anchor.

Warehouse KPI reference — JSON: the full structured release, including the dictionary, benchmark reference, definition crosswalk, safety table, county-to-authority crosswalk, worked-example results, source register and stated limitations. Version 2026-09-12-v2 is a reference compilation and test-data release, not an original warehouse survey.

Benchmark reference — CSV: 16 rows—11 numerical reference records and five documented exclusions. Every numerical row has a statistic type, unit, source population, source period, source URL and verification date; excluded WERC thresholds have empty numeric-value fields and an explicit exclusion reason, not zeros or invented substitutes.

Definition crosswalk — CSV: nine rows, linking each comparison to the metrics, changed denominator or clock, source URLs and the reporting mistake the check prevents. The safety table CSV contains eight industry rows, and the county crosswalk CSV contains all 15 metro counties with the relevant state and authority sources.

The complete synthetic inputs are fill rates, perfect orders, inventory counts, labour hours and dock time. The reproduction script recalculates all five examples from those files without network access or third-party packages; its checked results are also included. It checks arithmetic, not the continuing accuracy of external benchmark pages.

How to cite this page

Publication: Cincinnati Dock Door Repair Research
Page: Warehouse KPIs: Formulas, Worked Examples, and Which Benchmarks Actually Compare
URL: https://cincinnatidockdoorrepair.com/research/warehouse-kpis/
Version: 2026-09-12-v2
Last verified: September 12, 2026
Source data vintages: listed with each figure and in the source register.

Benchmark measurements remain attributed to their named publishers. Cincinnati Dock Door Repair Research’s contribution is the compilation, explicit operational definitions, comparison checks and constructed example inputs; the source register distinguishes those contributions from outside measurements and from sources reviewed but not adopted.

Which warehouse KPI benchmarks are actually comparable?

Published warehouse benchmarks are comparable only after their definitions, statistics, populations and periods have been matched. APQC’s order-fill median is 94.0% on a displayed sample of 3,218 labelled “All Companies,” but that does not establish an on-time-shipment target or a local warehouse standard. This table preserves the five WERC lookup subjects while excluding their numerical thresholds, because those values were not checked in the original report. 2 13

Table 6. Published warehouse benchmark reference, with population, statistic type and documented exclusions

Source: APQC’s six primary measure pages; OSHA Instruction CPL-03-00-026, Table 1; BLS NAICS 493 industry rates. WERC’s report page supports the report/access description, not a numeric threshold in this table. ★ means the numerical value was read directly in the named primary document on September 12, 2026. “Excluded” is a documented absence, not a provisional benchmark. 1 2 3 4 5 6 7 11 13

Why the two federal warehouse injury rates disagree — and why neither is wrong

The two federal nonfatal recordable-case figures shown here differ: 5.2 and 4.8. The larger figure is OSHA’s 2020–2024 five-year average for NAICS 493110, general warehousing and storage only. The smaller is BLS’s single-year 2024 rate for NAICS 493, the broader warehousing and storage subsector, which includes refrigerated, farm-product and other warehousing. Different periods and industry scopes make these different statistics—not conflicting estimates of the same population in the same year. Neither comparison alone identifies the cause of the difference. 7 11

Table 7. Recordable and DART rates in OSHA’s warehousing emphasis program, 2020–2024 five-year averages

Source: OSHA Instruction CPL-03-00-026, Table 1, effective July 31, 2026. Units: cases per 100 full-time-equivalent workers. OSHA’s footnote identifies the Postal Service row as ITA-based; the other rows are calculated from BLS data. The postal administrative submissions must not be described as part of the BLS probability sample. Verified September 12, 2026. 7 8

Why this page assigns no universal targets

This source set does not establish a warehouse performance target valid across product mix, order profile, automation level, unit of measure, labour model and service promise. A pallet-based output rate and a piece-based output rate are not even the same unit, before differences in operations are considered. Four categories are worth keeping apart: an observed benchmark drawn from a sample, expert advice about what to aim for, a contractual target you owe a customer, and an internal goal you chose. Only the first, as defined here, describes measured performance in a named sample; advice and goals are not substitute observations.

What does this data show—and what doesn’t it show?

The tables above establish what each metric counts and which numerical references were verified, with their populations attached. They do not establish what any individual warehouse should achieve, and they do not describe Ohio’s operating performance. APQC’s medians carry displayed sample sizes but no observation periods on the public pages; the label “All Companies” does not establish that a sample contains only U.S. warehouses. 1 2 3 4 5 6

How was this warehouse KPI reference compiled?

We checked the six APQC measure pages, the federal safety sources and the original documents used for definitions, then separated those measurements from this publication’s constructed examples. Every published benchmark retains its source period and population; every example retains its complete input file. The compilation was verified on September 12, 2026, and the calculations were rerun from the supplied CSVs.

We opened all six APQC Open Standards Benchmarking measure pages: dock-to-stock cycle time (100677), order fill rate (101445), perfect order performance (101741), inventory accuracy (100781), inventory carrying cost (100784) and operate-warehousing cost (103784). We read the measure ID, displayed total sample size, median, units and available calculation terms from the public pages. All six displayed their median and sample size while the 25th- and 75th-percentile cells showed dashes. Their observation years were not disclosed, and we do not infer survey dates or unique-company counts beyond the displayed sample labels. 1 2 3 4 5 6

For the safety layer we retrieved OSHA’s 13-page directive CPL-03-00-026 from its own PDF URL and read the effective date, expiration provision, background and Table 1, including its footnotes. The final values do not rely on search-result snippets. The Postal Service row is explicitly identified as ITA-based; the other Table 1 rows are BLS-derived. We also read BLS’s own NAICS 493 page for the separate 2024 rates and its incidence-rate guidance for the nonfatal numerator and employee-hours denominator. We did not independently rebuild OSHA’s five-year averages or its historical employment series from individual BLS releases. 7 8 11 12

For the historical figures we inspected the original de Koster, Le-Duc and Roodbergen review in the Erasmus University repository, including its abstract, introduction, put-system passage and Figure 4. The abstract calls the cost share an estimate; the figure attributes its time distribution to Tompkins and colleagues. We checked that attribution in the review, not in a separate copy of the 2003 textbook. The review’s introductory “picks” wording and its later “put handlings” wording are preserved as a terminology distinction rather than silently merged. 17 18

For WERC we read its public report page, its 2025 release announcement and MHI’s 2026 summary. We also inspected the two dated Yale documents identified in the source list. Those documents are public reproductions, not the original WERC benchmark tables: their numerical thresholds and year-over-year movements are therefore not adopted as verified measurements here. WERC’s primary page lists 36 metrics for the 2026 edition and a quintile framework; at verification, it listed the report at $200 for individual members and $550 for non-members, with different access for eligible organizational members. These are access facts, not operational benchmark values. 13 14 15 16 24

The five worked examples were checked against their synthetic CSV inputs and recalculated with an executable script using exact decimal arithmetic. An independent calculation using rational arithmetic checked the headline results and the two alternative scenarios discussed in the text. The perfect-order component-index percentage is exactly 130,321 ÷ 1,600 = 81.450625%; the underlying fraction before multiplication by 100 is 0.81450625. All examples are constructed test inputs disclosed as such, never observations or estimates of real warehouse performance. 23

Each dictionary row was checked for dimensional consistency: the numerator and denominator must produce the stated unit. Completed-receipt and completed-order cycle times are arithmetic means of included elapsed durations, not medians of warehouse-level results. A percentage pooled across teams should be calculated from the summed numerator and denominator, not by averaging team percentages without their weights. An operation with no eligible observations has an undefined rate, not an automatic 0% or 100%.

The declared cohort rules are part of the definition. Record cancellations, split shipments, frozen promised dates, open work and unknown delivery outcomes separately rather than silently changing the denominator. Report how average inventory value was computed, which employees contribute person-hours, which doors are serviceable and scheduled, and which cost categories were included. These are explicit implementation choices in this reference, not hidden claims that every source mandates an identical system.

For the local layer, we joined the 15 counties in OMB’s Cincinnati metropolitan delineation to OSHA’s Ohio, Kentucky and Indiana coverage descriptions. That creates a county-to-authority reference, not a local performance sample. The metropolitan boundary is statistical geography; it does not itself determine regulatory jurisdiction, and the state-plan exceptions remain visible. 22 25 26 27

Verification notation: ★ means the numerical value was read in its named primary document, or the calculation was reproduced from the disclosed inputs. Excluded means no numerical claim is published because primary numerical verification was not completed; it is not a pending or provisional data point. S and E describe formula origin, not the strength of a performance benchmark. The source register records document vintages, verification dates and access limits separately.

How can the same warehouse data produce different KPI answers?

Every example below is a constructed illustration, not an observation of a real warehouse. Each one is small enough to check by hand and is published with its complete input file. The point of all five is the same: holding the inputs fixed does not hold the result fixed when the definition changes.

When 99% line fill means only 90% complete orders

Setup. 100 orders. 10 lines per order, one requested unit per line — 1,000 lines in total. Ten lines are unfilled, and each one sits on a different order.

Table 8. Fill-rate results from the same 1,000 lines

Source: Cincinnati Dock Door Repair Research, constructed illustration, recomputed from example-01-fill-input.csv on September 12, 2026. Synthetic data, not observed warehouse performance.

Place those same ten misses on one order instead of ten and order fill goes to 99% with line fill unchanged. The total number of filled lines stays the same; the number of incomplete orders changes from ten to one. The order-level measure detects that distribution of misses, while the aggregate line-level measure does not. This alternative is also checked in the reproduction script.

Why four 95% components do not determine perfect-order success

APQC publishes its perfect-order method in plain terms: multiply the on-time delivery, complete, damage-free and accurate-documentation rates expressed as decimal fractions, then multiply by 100. 3 We built two 100-order scenarios in which all four component rates are exactly 95%, and varied only which orders fail.

Table 9. Two scenarios, identical components, different truth

Source: Cincinnati Dock Door Repair Research, constructed illustration, recomputed from example-02-perfect-order-input.csv on September 12, 2026. Component index method as published by APQC, Measure 101741. Synthetic data.

The index cannot tell those two operations apart; a direct count can. This does not make APQC’s measure wrong—we call it a component index here to distinguish its published multiplication method from a direct joint count. It does mean that an 88.0% median perfect-order figure and a warehouse's own count of flawless orders are two different quantities, and swapping one for the other is a reporting error, not a rounding difference.

Balanced inventory totals can hide zero exact matches

Setup. Two item-location records, both expressed in the same counting unit for this illustration. The system says 100 units in each. The count finds 90 in one and 110 in the other.

Table 10. Two records, two verdicts

Aggregate quantity ratio: 100%. Exact-match record accuracy (K06): 0%.

Source: Cincinnati Dock Door Repair Research, constructed illustration, recomputed from example-03-inventory-input.csv on September 12, 2026. Synthetic data.

Offsetting errors cancel in a total and survive intact in a picking aisle. Any inventory accuracy figure that does not say whether it counts records or units is unusable for comparison, which is why APQC's 95.0% inventory accuracy median is not mapped to K06 anywhere on this page.

150 lines per hour is not 150 lines per labour hour

Setup. Three pickers, one eight-hour shift, 400 completed lines each.

Table 11. One shift, two productivity numbers

Source: Cincinnati Dock Door Repair Research, constructed illustration, recomputed from example-04-labor-input.csv on September 12, 2026. Synthetic data.

Both are correct. In this example, the elapsed-hour throughput figure describes the system and the person-hour figure describes the included labour. If a fourth picker also completes 400 lines in the same eight hours, throughput rises to 200 lines per elapsed hour and labour productivity remains 50 lines per person-hour; adding a worker alone does not guarantee that output. Comparing a system throughput number with a person-hour benchmark substitutes the team’s shared shift clock for its summed labour hours.

Full docks at one instant can still mean 50% productive time

Setup. Two dock doors, scheduled eight hours each. Both stand idle for the first four hours and work for the last four.

Table 12. Snapshot against scheduled time

Source: Cincinnati Dock Door Repair Research, constructed illustration, recomputed from example-05-dock-input.csv on September 12, 2026. Synthetic data.

Walk the dock only at the end of this constructed shift and every door is occupied, despite half of the scheduled door-hours being unproductive. The time-based measure exposes that difference; it does not identify its cause. Nothing in these inputs establishes whether the idle period came from arrival timing, available work, staffing, equipment or another constraint, so neither a building fault nor a scheduling fault follows from this table alone.

What nine definition checks should precede a warehouse comparison?

Nine checks cover substitutions that can make two warehouse numbers look comparable when they are not. The five examples above demonstrate the first five checks; the other four follow the stated clock, endpoint, cost and exposure definitions. The order-versus-line check alone moves the same 1,000 lines from 99.00% line fill to 90.00% order fill.

Table 13. Definition crosswalk

Source: Cincinnati Dock Door Repair Research definition crosswalk, compiled and checked September 12, 2026, from the definitions in Tables 1–5 and their named sources. 1 2 3 6 12 19 20 23

Where did the older warehouse cost and picking figures come from?

The three figures below appear in a 2007 literature review, and they are older than a current-year label makes them look. We checked the review’s own wording and Figure 4, rather than treating later repetitions as new research. The table records what those passages actually say; it does not establish that the figures describe today’s warehouses or that the review was the first appearance of every number. 17 18

Table 14. Provenance and scope of three historical warehouse figures

Source: de Koster, Le-Duc and Roodbergen (2007), European Journal of Operational Research 182(2), 481–501, especially the abstract, pp. 482 and 484, and Figure 4 on p. 486. Original pages and chart checked September 12, 2026 in the Erasmus University repository. 17 18

Three further claims are not adopted as industry-wide facts here: labour at 50–70% of a warehouse operating budget, 99% inventory accuracy as an industry standard, and carrying cost at 20–30% of inventory value. The inspected source set does not establish those ranges as universal measurements or requirements. That is a limit on this reference’s evidence, not a claim that no supporting study exists anywhere. APQC’s own carrying-cost median is reported separately with its actual denominator and undisclosed observation period, rather than used to certify or refute every other percentage. 5

How do you choose which warehouse KPIs to track?

Start from the operating question, not from a target percentage. Each question below maps to specific definitions in Tables 1–5; the ID is what makes the measurement reproducible six months later when someone asks how the number was built.

Table 15. Operating question to metric definition

Source: Cincinnati Dock Door Repair Research’s operating-question-to-definition crosswalk, compiled September 12, 2026. These selections are editorial guidance, not a mandatory scorecard. 23

Two habits keep a dashboard honest. Keep speed, accuracy and safety visible together, so a reported gain in one does not hide a deterioration in another. Write down the cohort rules—cutoffs, cancellations, split shipments, open orders—before the first report runs: changing a denominator can create an apparent improvement with no improvement in the underlying work. The dashboard should display each metric’s actual, period, numerator, denominator and target source; an internally chosen target stays labelled as an internal goal.

What can these benchmarks tell Ohio and Hamilton County warehouses?

None of the performance benchmarks on this page establishes an Ohio or Hamilton County operating average. APQC’s public samples are labelled “All Companies” without a local breakdown, and the federal safety rates are national industry figures. The local table below is instead a county-to-authority crosswalk: it answers a jurisdiction question, not a productivity question. 1 2 3 4 5 6 7 11

No locally representative picking-rate, fill-rate or dock-to-stock dataset was verified in the sources inspected for this reference. BLS’s Quarterly Census of Employment and Wages provides industry and area data, but establishment, employment and wage counts do not establish picks per hour. We have not filled the operational-data gap with labour-market statistics or treated the absence of a verified value here as proof that no local dataset exists. 28

One thing can change locally: the responsible enforcement authority, not the incidence-rate formula. The Cincinnati metro area spans three states. Federal OSHA covers most private-sector workplaces in Ohio, while Kentucky and Indiana operate approved State Plans covering most private-sector workplaces and state and local government employment. Coverage exceptions matter—federal employers, USPS and certain other operations remain outside those State Plans—so a county name alone does not settle every establishment’s jurisdiction or imply that an inspection will occur. 25 26 27

Table 16. Cincinnati OH-KY-IN metro area (CBSA 17140), by principal private-sector workplace-safety enforcement authority

Source: OMB Bulletin 23-01, List 2, CBSA 17140 (Appendix p. 48), joined to OSHA’s Ohio coverage page and Kentucky and Indiana State Plan descriptions. County delineation: July 21, 2023; jurisdiction descriptions verified September 12, 2026. State Plans must be at least as effective as federal OSHA; this crosswalk does not claim that the 2026 federal warehousing directive has automatically been adopted by either state. 22 10 25 26 27

Why do the source dates matter in 2026?

Two 2026 developments affect how this reference is dated: OSHA issued a revised warehousing directive effective July 31, and WERC released a new DC Measures edition in May. The directive has a five-year expiration provision, while WERC’s public summary describes 36 metrics and a new most-used-measures list. Neither publication date turns an older observation into a new measurement. 7 13 16

The APQC medians here are dated by access because their public observation years are undisclosed. BLS’s broad warehousing safety rates describe 2024; OSHA’s narrower industry comparisons describe 2020–2024. The WERC 2025 announcement and 2026 summary identify different most-used-metric lists, but a change in what respondents track is not proof that the same facilities improved or deteriorated. Numerical year-over-year performance claims from the public reproductions are not presented here without the original benchmark tables. Old formulas are not automatically obsolete, and a current-year headline is not a substitute for a source vintage. 1 7 11 16 24

What are this reference’s limitations and excluded figures?

This compilation is a definition-and-source reference, not an original survey of warehouses or a representative distribution of facility performance. Its five worked examples are synthetic inputs built to expose definitional differences, not observations of real facilities. A formula, a benchmark and an internal target remain different things even when they share the same unit.

What this compilation is not. It contains no causal estimate, no representative Ohio or Hamilton County operating distribution, and no repair, service or equipment pricing. The examples establish arithmetic consequences of declared inputs; they do not establish real-world productivity, the cause of idle time or the performance of any named warehouse. The safety-rate comparisons are educational references, not a finding that a facility is safe or compliant. OSHA itself warns against relying on one safety indicator. 9

Benchmarks and source limits.

• APQC’s six public measure pages do not disclose observation years. We can report their displayed medians, sample sizes and definitions as checked on September 12, 2026, but not an unstated survey period, geography or deduplication rule. All six numerical rows were directly rechecked in this pass; none remains a provisional carry-forward.

• APQC’s inventory-accuracy description does not establish enough calculation detail to equate its 95.0% median with exact-match record accuracy. It is therefore not mapped to K06 or presented as an exact-match target.

• The original WERC numerical benchmark tables were not inspected. Five lookup subjects remain in Table 6 as documented exclusions, without numeric values. Yale’s dated public reproductions were examined for provenance, not substituted for primary numerical verification; their threshold and year-over-year movement figures are not published as verified benchmarks here.

• OSHA’s five-year averages were transcribed from its primary directive, not recalculated from underlying annual records. The Postal Service row is ITA-based administrative data; the other rows are BLS-derived. BLS’s separate 2024 NAICS 493 rates were read directly on its own page.

• No Ohio or Hamilton County operational KPI value is published because no locally representative value was verified in the inspected source set. The 15-county crosswalk describes statistical geography and principal enforcement coverage, not a local performance sample.

• The labour-budget range, universal inventory-accuracy claim and carrying-cost range discussed above are excluded as industry-wide rules. This reference does not claim that no research on those subjects exists; it does not have the evidence needed to adopt those particular assertions universally.

• Oracle’s exact-match and tolerance definitions come from a historical product release. The 2007 review’s cost, travel-time and put-system passages are also historical; the Tompkins textbook attribution was checked in the review rather than independently in the textbook.

Dates and missing values. The visible verification date records this source-checking pass, not a promise that every publisher’s underlying data is equally recent. Blank numeric fields in the five excluded benchmark records mean “no verified value published”; they never mean zero performance. Changes in source values or definitions require a new version with a stated revision, rather than a silent replacement of the archived release.

What else should you know about warehouse KPIs?

The questions below distinguish warehouse metric definitions from performance targets and explain the most consequential comparison choices. Each answer refers back to the published formulas or source evidence, not to an assumed universal benchmark.

What are the five main warehouse KPIs?

There is no universal five. WERC’s 2026 public summary names on-time shipments, dock-to-stock cycle time, average warehouse capacity used, peak warehouse capacity used and backorders as a percentage of total lines as its five most-used measures. That is a finding about what respondents track, not a recommendation that every facility should use the same five. 16

What is a good dock-to-stock time?

APQC’s public median is 15.0 hours, with a displayed sample of 3,560 labelled “All Companies” and no observation years disclosed. It is a reference median, not a universal target; a verified numerical WERC threshold is not supplied here. Before using a comparison, confirm the same receipt-to-recorded-and-put-away endpoints, elapsed-time rules, receipt unit and operating population. 1 13

What is a good order-picking accuracy percentage?

Settle four things first: whether you are counting lines, units or orders; whether the sample is audited or comprehensive; whether a defect caught and fixed before shipment still counts as a first-pass failure; and who is in the comparison group. K15 uses first-pass audited lines, while other definitions can count whole orders or corrected results. This page does not prescribe a universal percentage or repeat a WERC threshold that was not checked in the original numerical table. 23 13

What is the difference between order fill rate and OTIF?

Order fill rate (K11) asks whether the order was filled completely within the declared cohort and cutoff. Delivered OTIF (K18) asks whether the customer received it complete and by the delivery deadline. In an illustrative order shipped complete but delivered two days late, the order can pass the first measure and fail the second. 2 23

Is dock-to-stock time the same as truck turnaround time?

No. Dock-to-stock ends when received goods are recorded and put away inside the warehouse, whereas an arrival-to-departure truck-turnaround measure ends when the vehicle leaves. A quickly emptied trailer and freight waiting for put-away can therefore produce different results on those two clocks; neither endpoint can be silently substituted for the other. 1 23

How do I calculate warehouse productivity per hour?

Declare the output unit and the clock. Three pickers completing 400 lines each in one eight-hour shift produce 150 lines per elapsed hour and 50 lines per person-hour in the constructed example. A source labelled per person-hour must be compared with summed person-hours, not with the shared shift duration. 23

What is a good warehouse injury rate?

OSHA’s directive reports 2020–2024 averages of 5.2 recordable nonfatal cases and 4.4 DART cases per 100 full-time-equivalent workers for general warehousing, against 2.6 and 1.6 for all private industry. These are industry reference rates, not safety targets or proof that a workplace is safe. Compare the same industry scope, case type, time period and hours-worked basis, and do not rely on a single indicator. 7 9 12

Can I compare my warehouse’s number directly against the figures on this page?

Only after matching the definition and population. The APQC component-product measure is not an observed perfect-order count, revenue-normalised warehousing cost is not cost per order, and a national industry safety rate is not an Ohio or Hamilton County average. The nine checks in Table 13 identify the denominator, clock, endpoint and exposure substitutions that this reference guards against. 3 6 11 23

How often do these figures change?

APQC’s public benchmark displays, source definitions and report editions require separate checks; a page-access date does not reveal an undisclosed observation year. The review schedule for this reference is quarterly for public benchmark pages and links, with a source check when new BLS safety data or a new WERC report appears. The visible verification date and version change only after an actual check; the reproduction script does not monitor external sources.

Which sources support this warehouse KPI reference?

The register below identifies the issuing agencies, original measure publishers, definition documents and the literature-review pages used in this reference. It also identifies the public reproductions examined but not adopted as verified WERC numbers, and distinguishes the review’s textbook attribution from a separately inspected textbook.

1. APQC. Dock-to-stock cycle time in hours for supplier deliveries; Measure 100677. Observation period not disclosed; public-page snapshot 2026-09-12. Primary measure page directly checked; All Companies is not an established U.S.-warehouse-only sample. Verification date: September 12, 2026.

2. APQC. Order fill rate; Measure 101445. Observation period not disclosed; public-page snapshot 2026-09-12. Primary measure page directly checked. Verification date: September 12, 2026.

3. APQC. Perfect order performance; Measure 101741. Observation period not disclosed; public-page snapshot 2026-09-12. Component index is this reference’s distinguishing label, not APQC’s published metric name. Verification date: September 12, 2026.

4. APQC. Inventory accuracy; Measure 100781. Observation period not disclosed; public-page snapshot 2026-09-12. Public calculation detail does not establish equivalence to exact-match record accuracy. Verification date: September 12, 2026.

5. APQC. Inventory carrying cost as a percentage of inventory value; Measure 100784. Observation period not disclosed; public-page snapshot 2026-09-12. Reporting-period detail is not established; do not silently label the value annual. Verification date: September 12, 2026.

6. APQC. Total cost to perform the process operate warehousing per $1,000 revenue; Measure 103784. Observation period not disclosed; public-page snapshot 2026-09-12. Not a cost-per-order benchmark. Verification date: September 12, 2026.

7. OSHA. Instruction CPL-03-00-026: National Emphasis Program on Warehousing and Distribution Center Operations. Signed 2026-07-06; effective 2026-07-31; Table 1 covers 2020–2024. Primary 13-page PDF retrieved and its text/table read. Postal Service row uses OSHA ITA data; other Table 1 rows are calculated from BLS data. Underlying historical series were not independently reconstructed. Verification date: September 12, 2026.

8. OSHA. Injury Tracking Application (ITA) Data. Current page checked 2026-09-12. Not the BLS SOII probability sample. The older Establishment-Specific-Injury-and-Illness-Data URL redirects here. Verification date: September 12, 2026.

9. OSHA. Clarification on how the formula is used by OSHA to calculate incident rates. 2016-08-23 interpretation, checked 2026-09-12. Interpretation is explanatory guidance, not a new safety requirement. Verification date: September 12, 2026.

10. OSHA. State Plans. Current page checked 2026-09-12. Specific jurisdiction exceptions require the state-specific pages. Verification date: September 12, 2026.

11. U.S. Bureau of Labor Statistics. Industries at a Glance: Warehousing and Storage, NAICS 493. 2024 nonfatal case rates, checked 2026-09-12. National NAICS 493 figures, not NAICS 493110 alone or local values. Verification date: September 12, 2026.

12. U.S. Bureau of Labor Statistics. How To Compute Your Firm’s Incidence Rate for Safety Management. Current official guidance checked 2026-09-12. K23 is explicitly the nonfatal comparison basis; fatal cases are not in its numerator. Verification date: September 12, 2026.

13. Warehousing Education and Research Council. DC Measures report and benchmarking resources. 2026 report listing and historical report access, checked 2026-09-12. Original 2025 and 2026 numerical benchmark tables were not inspected. No unverified WERC threshold is published in this release. Verification date: September 12, 2026.

14. Yale Lift Truck Technologies / Hyster-Yale Materials Handling. Top 12 distribution center metrics; document 0000YBC0IG003. Copyright 2025; checked 2026-09-12. A public reproduction attributing data to WERC; not used to verify the original WERC numeric thresholds. Verification date: September 12, 2026.

15. Yale Lift Truck Technologies / Hyster-Yale Materials Handling. Benchmarking and improving distribution center metrics; document 0000YBC0WP008. Copyright 2025; checked 2026-09-12. A public reproduction, not the original WERC report; threshold and year-over-year numerical claims excluded. Verification date: September 12, 2026.

16. MHI. WERC’s 2026 DC Measures Report Addresses Strategic Paradox. 2026 report, released in May; checked 2026-09-12. MHI is WERC’s parent organization. Public summary does not expose the complete numerical benchmark tables. Verification date: September 12, 2026.

17. de Koster, R.; Le-Duc, T.; Roodbergen, K.J.. Design and control of warehouse order picking: a literature review; European Journal of Operational Research 182(2), 481–501. 2007 journal publication; received 2005-08-29; accepted 2006-07-17; online 2006-10-25. Reviewed relevant original pages and Figure 4. These are historical statements, not measurements of 2026 warehouses. Verification date: September 12, 2026.

18. Tompkins et al., as cited by de Koster et al.. Facilities Planning (2003), cited in Figure 4 and bibliography of de Koster et al. (2007). 2003 textbook attribution in the 2007 review; checked 2026-09-12. Textbook itself was not inspected. Bibliographic attribution and chart were verified in the cited review; no claim of independent textbook verification. Verification date: September 12, 2026.

19. Oracle. Materials Management and Logistics, Inventory Guide: Analyze Performance. Historical 11g Release 7 (11.1.7), checked 2026-09-12. Historical product documentation, not a current warehouse benchmark study. Verification date: September 12, 2026.

20. Mecalux. 9 warehouse KPIs to keep an eye on. Page checked 2026-09-12. Additional serviceable-door, unique-order and cost-boundary choices are stated by this reference. Verification date: September 12, 2026.

21. Modula. 30 Warehouse KPIs to Track & Measure Performance. Page checked 2026-09-12. Used for formula definition, not as primary empirical benchmark evidence. Verification date: September 12, 2026.

22. Office of Management and Budget. OMB Bulletin No. 23-01; List 2, Cincinnati OH-KY-IN, CBSA 17140. 2023-07-21 delineation; Appendix printed page 48 / PDF page 51; checked 2026-09-12. Statistical geography only; the metro boundary does not establish workplace-safety jurisdiction. Verification date: September 12, 2026.

23. Cincinnati Dock Door Repair Research. Warehouse KPI dictionary, benchmark reference, definition crosswalk, county crosswalk and constructed worked-example inputs. 2026-09-12-v2. Not a survey or observed warehouse-performance dataset. Verification date: September 12, 2026.

24. Warehousing Education and Research Council. WERC Releases 2025 DC Measures Report with a Focus on Combining Vision with Vigilance. 2025-06-05 release; checked 2026-09-12. Public release does not supply the numerical performance thresholds. Verification date: September 12, 2026.

25. OSHA. Kentucky State Plan. Current jurisdiction description checked 2026-09-12. Most private workplaces and state/local government; federal and specified operations remain outside the plan. Verification date: September 12, 2026.

26. OSHA. Indiana State Plan. Current jurisdiction description checked 2026-09-12. Most private workplaces and state/local government; federal and specified operations remain outside the plan. Verification date: September 12, 2026.

27. OSHA. Ohio: State Plan and OSHA Area Offices. Current jurisdiction description checked 2026-09-12. County list is a location crosswalk, not a guarantee of inspection. Verification date: September 12, 2026.

28. U.S. Bureau of Labor Statistics. QCEW Open Data Access. Current documentation checked 2026-09-12. No Ohio or Hamilton County operational performance figure is inferred from these data. Verification date: September 12, 2026.

The machine-readable source register carries the same source IDs, URLs, document vintages, verification date and use limitations as the structured dataset.

Verification and version record

2026-09-12-v2 — September 12, 2026. All six APQC values and the two BLS NAICS 493 rates were directly checked; OSHA’s Table 1 source footnotes were carried through; unverified original-report WERC numbers were excluded; and every example was recalculated. The source register, 24-row dictionary, 16-row benchmark-and-exclusions file, nine-row crosswalk and complete synthetic inputs belong to this same release.

Cincinnati Dock Door Repair Research is the independent research and reference section of cincinnatidockdoorrepair.com.

Last verified: September 12, 2026.