Research · Nesting Utilization Report 2026 · numbers as of 13 September 2026
The median nesting job uses 69% of the plate.
This report covers every completed nesting job run on Lapas between April 2026 and : over 6,900 layouts from 1,100+ shops. Median layout utilization is 69%. The middle half of jobs falls between 53% and 80%. One job in four reaches 80% or better, and one in five finishes under 50%.
- Median utilization
- 69%
- Middle half
- 53% to 80%
- At 80% or better
- 24.5%
- Under 50%
- 22.3%
6,903 completed jobs
first and third quartile
1,690 jobs, about one in four
1,541 jobs, about one in five
Published by Lapas CC BY 4.0 refreshed monthly
Summary
6,903 completed nesting jobs, run by 1,186 shops between April 2026 and 13 September 2026, measured as placed part area over total plate area and area weighted across every plate of the job. No job was selected, rated or filtered for how well it nested.
- Distribution: median 69%, quartiles 53% and 80%, mean 64%. 24.5% of jobs reach 80% or better, 22.3% finish under 50%.
- Job size is the biggest lever: median 52% for jobs of 1 to 10 parts against 75% for 101 to 300 parts. Batching more parts onto a plate moves utilization more than any setting.
- Stock is not standard: 73% of jobs run on a plate that is not one of the five common sizes, across 1,205 distinct plate dimensions.
- Shops iterate: 49% of jobs are a re-nest of a project already nested, across 2,917 projects.
- Shop to shop spread is wide: among the 379 shops with 5 or more nests, the median shop sits at 68% and the middle half between 54% and 78%.
The distribution: median 69%, quartiles 53% and 80%
Across all 6,903 jobs the median layout uses 69% of the plate. The spread around it is what matters: the first quartile sits at 53% and the third at 80%, so the middle half of real nesting work spans 27 points. 1,690 jobs (24.5%) reached 80% or better and 1,541 (22.3%) came in under 50%. The mean, 64%, sits below the median because the low tail is long.
6,894 of 6,903 jobs are drawn. The other 9 carry an engine-reported density outside 0 to 100, which is a measurement artefact rather than a layout, so they have no bin. They stay in the row data and in every percentile on this page rather than being quietly dropped: at that count they cannot move an order statistic across 6,903 jobs.
Job size is the biggest lever in the data
Grouping jobs by how many parts were requested separates the distribution more cleanly than any other field in the set. Jobs of 1 to 10 parts have a median of 52%. Every bucket above that sits in the low seventies or better, peaking at 75% for 101 to 300 parts. The plain implication: batching more parts onto a plate moves utilization more than any setting.
The gain flattens rather than continuing. Past 100 parts the medians move by a couple of points, not by twenty, so the large step is from a handful of parts to a full plate of them.
| Parts per job | Jobs | p25 | Median | p75 | Mean |
|---|---|---|---|---|---|
| 1 to 10 | 1,411 | 26% | 52% | 70% | 48% |
| 11 to 30 | 2,013 | 58% | 71% | 81% | 67% |
| 31 to 100 | 2,171 | 59% | 72% | 81% | 68% |
| 101 to 300 | 844 | 60% | 75% | 83% | 70% |
| 300 or more | 464 | 58% | 72% | 81% | 68% |
73% of jobs run on non-standard stock
1,205 distinct plate sizes appear across the 6,903 jobs, and 73% of jobs use a plate that matches none of the five common classes. The single most common size is 1,000 x 800 mm, on 475 jobs, which is a small plate rather than a mill sheet. Whatever the catalogue says, most nesting work is done on offcuts and odd stock.
Among the standard classes the medians run 1250 x 2500 at 77%, 1000 x 2000 at 77%, 1500 x 3000 at 74%, 4 x 8 ft at 67% and 5 x 10 ft at 64%. Jobs on inch-native stock, which is inferred from the plate dimensions rather than recorded, are 16.6% of the set and run a median of 65% against 70% for metric stock.
Read this as an observation, not a cause
Plate size and job size are not independent here. Large plates attract different work from small ones, and the inch and metric groups differ in job size and stock mix as well as in units. Nothing in this data separates those effects, so the gap between 65% and 70% is a description of what the two groups did, not evidence that a unit system or a plate size causes a yield difference.
Shops nest the same project again, and again
49% of all jobs are a re-nest of a project that had already been nested at least once. The 6,323 jobs that carry a project identifier cover 2,917 distinct projects, and the tail is long: 292 jobs are the eleventh or later nest of their project. Whatever a shop does with the result, nesting is not a one-shot step in it.
Jobs by nest number within their project
- 1st 2,917
- 2nd 1,192
- 3rd 646
- 4th 418
- 5th 273
- 6th 191
- 7th 137
- 8th 111
- 9th 85
- 10th 61
- 11th+ 292
- Parts per job
- 31
- Single-plate jobs
- 52.5%
- Jobs with a part left off
- 14.9%
- Nest time
- 33 s
median, against a mean of 107
p90 is 6 plates
1,029 jobs left at least one part unplaced
median; tracks the requested time limit
The engine runs to a time limit the shop asks for, so nest duration measures the budget at least as much as the work. Read the 33 second median as what shops chose to wait, not as how long the problem took.
Shop to shop, the spread is nearly as wide as job to job
Taking each shop's own median rather than each job removes the noise of one-off small jobs. Among the 379 shops that ran 5 or more nests, the median shop sits at 68% and the middle half falls between 54% and 78%. That is a 24 point interquartile range between shops, against 27 points between individual jobs. The variation is not only in the jobs, it is in the shops.
| Percentile of shops | p10 | p25 | p50 | p75 | p90 |
|---|---|---|---|---|---|
| Shop median utilization | 40% | 54% | 68% | 78% | 84% |
1,186 shops ran the 6,903 jobs. The median shop ran 3 of them, so most shops contribute very few rows, which is why the spread table is limited to shops with 5 or more.
Method, and what this data cannot tell you
What utilization means here
Placed part area divided by total plate area, area weighted across every plate of the job. A job that fills three plates and leaves a fourth nearly empty is scored on all four, so a partly filled last plate pulls the figure down. That is the honest whole-job number, and it is lower than a per-plate or best-plate figure would be.
What is in the set
Every completed job run between April 2026 and 13 September 2026: 6,903 jobs from 1,186 shops. Excluded before any figure was computed: administrative and internal accounts, and jobs that failed rather than producing a layout. Nothing was excluded for nesting badly.
Privacy
The published rows carry no shop, account, project or file identifier and no part geometry. Each shop is a stable integer assigned by hashing, ordered so the number carries no signal about size or recency, and the mapping was never saved. Dates are truncated to the month.
Reproducing it
Every aggregate on this page is recomputed from the published CSV by scripts/build-utilization-study.mjs, and the page prints only what that script writes. Percentiles use linear interpolation between order statistics. Plate sizes are counted orientation insensitive, so a 2000 x 1000 plate and a 1000 x 2000 plate are one size.
Limits, stated plainly
- This median is not a vendor utilization figure. Marketing numbers are usually measured on a full plate of a chosen job. This one includes 10-part jobs and partly filled last plates, so it is lower by construction and the two are not comparable.
- No comparison to manual nesting is made or implied. Nothing in this data records what a job would have yielded laid out by hand or in another tool.
- No material or kerf. Material is named on 22 of the 6,903 jobs and thickness on 22, so neither supports any cut of this data. Kerf is not stored at all, so no kerf correction is applied to any figure here.
- Units are inferred. Dimensions are stored in millimetres. A job is labelled inch-native when both plate dimensions are exact inch values, which is a guess about the stock a shop bought, not a recorded setting.
- The user base skews free. These are jobs from a population that is overwhelmingly on the free plan, including first-time and trial runs, which is part of why the low tail is as long as it is.
- Development traffic cannot be fully identified. Internal accounts were removed by account list. There is no environment field, so an unflagged internal account would not have been caught.
- One tool, not the industry. Every job was run on one nesting engine. This is a description of what these shops nested, not of nesting everywhere.
The data, published in full
Both files are released under CC BY 4.0. Reuse them, including commercially, with attribution. Figures on this page are recomputable from the CSV alone, which is the point of publishing it.
- Row data, CSV
- nesting-utilization-2026-jobs.csv
One row per job, 6,903 rows. Utilization, plates, plate size and class, inferred units, parts requested and placed, spacing, nest duration and the nest number within its project.
- Aggregates, JSON
- nesting-utilization-2026-summary.json
Every table and chart on this page, in the exact form the page reads them.
- Attribution line
- Lapas, Nesting Utilization Report 2026, CC BY 4.0, https://lapas.io/research/nesting-utilization-2026/
Figures are as of and are refreshed monthly. If a number goes down at the next refresh, the lower number is what gets published. Found an error in the method or the arithmetic? Write to [email protected] and the correction will be published here.
This data comes from Lapas, browser-based nesting software for cutting shops, which has a free plan. See pricing. Related: the engine benchmark against Deepnest, which measures a different thing on published test instances.