69.5%
of bill headers were below 500, but they represented only 7.9% of billed amount at the approximate median source ID. This is an automation workload.
Level proprietary research, 2024 cohort
Nearly seven in ten bills carried less than one-tenth of the dollars
These operating cost benchmarks analyze 1,192,695 eligible 2024 bill headers from 652 source company IDs. At the approximate median source ID, 69.5% of bill headers fell below the source-record 500 threshold but represented only 7.9% of eligible billed amount.
That changes the CEO agenda. Automate the high-volume, low-dollar workflow. Negotiate the five-vendor core that held 72.5% of billed amount. Govern the one-time-vendor tail without pretending it is the main savings pool.
These accounts payable benchmarks support contractor vendor spend reviews without pretending the cohort is a verified contractor census. The page combines a vendor concentration benchmark, major-vendor bill footprints, public-threshold tests, and a separate observation about software audit false positives.
1.19M
eligible 2024 bill headers
652
eligible source company IDs
72.5%
billed amount in the top five vendor names
2.2x
peak month versus median active month
Read the denominator literally: a source company ID is not asserted to be an independent company or a contractor. A bill header is not cash, material consumption, job cost, overpayment, or verified savings. Dollar-looking values inherit the source-record currency assumption.
Level 2024 historical AP research
The data splits cost control into three different jobs
72.5%
of billed amount sat with the five largest vendor names at the approximate median source ID. This is the negotiation and service-level agenda.
33.3%
of vendor names appeared once, but represented only 3.8% of billed amount. This is an onboarding and control problem.
Small bill share of header volume69.5%
Small bill share of billed amount7.9%
One-time vendors as share of names33.3%
One-time vendors as share of billed amount3.8%
Approximate medians across 652 eligible source company IDs and 1,192,695 eligible 2024 bill headers. Source IDs were not independently deduplicated or verified as contractors. Bills are not cash, consumption, overpayment, or savings. Open the full study →
The operating portfolio
681 bills, 48 vendor names, and three different control jobs
The approximate median source ID had 681 eligible bill headers and 48 normalized vendor names in 2024. The portfolio is large enough to overwhelm manual review, but concentrated enough to manage with different lanes.
Five largest vendor names
72.5%Commercial terms, repeat baskets, freight, availability, rebates, credit, and service levels belong here.
All remaining vendor names
27.5%The remaining dollars still matter, but a different control method is needed for routine, emergency, specialty, and one-time vendors.
Vendors used once
33.3%One-time vendors represented 33.3% of names, creating onboarding, coding, tax-document, approval, and fraud-review work.
Billed amount at one-time vendors
3.8%Only 3.8% of billed amount sat in the one-time tail at the approximate median source ID. This weakens a broad long-tail negotiation story.
Major vendor bill footprint
Common vendor names appeared differently across the cohort
This table reports source IDs with at least one eligible bill header for each reviewed major vendor family. It is not market share, contractor adoption, cash paid, or proof that one supplier is better.
| Vendor family | Source IDs | Share of 652 IDs | Approx. median bill headers among users | Approx. median annual billed amount* |
|---|---|---|---|---|
| Home Depot | 370 | 56.7% | 22 | $3,142 |
| Grainger | 361 | 55.4% | 11 | $3,819 |
| Trane | 305 | 46.8% | 28 | $63,511 |
| Ferguson | 295 | 45.2% | 19 | $10,122 |
| Johnstone Supply | 263 | 40.3% | 20 | $8,004 |
| United Refrigeration | 261 | 40% | 35 | $19,811 |
| Sunbelt Rentals | 243 | 37.3% | 6 | $6,063 |
| United Rentals | 206 | 31.6% | 5 | $5,678 |
| Amazon | 197 | 30.2% | 5 | $1,244 |
| Lowe's | 195 | 29.9% | 8 | $891 |
*Source-record currency is assumed USD but was not independently proven. Billed amount is not cash paid, expense, consumption, or savings. Approximate medians are among source IDs using that vendor family.
Selection sensitivity is part of the result
Home Depot prevalence moved from 51.3% when the cohort required 10 valid bill headers to 61.1% when it required 50. Grainger prevalence moved from 46.7% when the cohort required 10 valid bill headers to 63.7% when it required 50. The amount cap barely moved prevalence, but the activity rule did. A vendor-prevalence statistic without its minimum-activity rule is incomplete.
Discount eligibility reality check
Threshold-sized bill headers were uncommon, and still did not prove eligibility
Home Depot
12.2%
of 370 vendor-using source IDs had at least one bill header at or above the compared $2,500 public trigger.
Lowe's
4.1%
of 195 vendor-using source IDs had at least one bill header at or above the compared $2,000 public trigger.
Amazon
0%
of 197 vendor-using source IDs had at least one bill header at or above the compared $10,000 public trigger.
The rebate upper bound is not a savings estimate
For the separate AGC and Home Depot semiannual program, 74 of 370 Home Depot-using source IDs, or 20%, had more than $12,500 of bill headers in at least one calendar half. Applying 2% to all $6.03 million above-threshold cell dollars produces $120,642 of arithmetic, but membership, enrollment, registered tender, exclusions, pre-tax net purchases, returns, timing, and rebate receipt were not observed. The number is an investigation ceiling, not achievable savings.
Level field observation
A cadence-only software audit overstated the review pool by 54.2 times
79.6%
of 206 apparent recurring-cost findings did not survive account, structural-spend, and person/vendor controls. Only 42 remained software-subject review candidates.
98.2%
of the apparent annualized dollars did not survive those controls. The raw detector produced $10.68 million, while the controlled review population was $197,066. Neither number is savings.
One read-only detector-control run, published as a Level field observation rather than a market benchmark. Payroll, tax, benefits, pass-through cash, people, inventory, insurance, and transfers can all look recurring. Read the complete study and method →
Monthly volatility
The peak month was 2.20 times the median active month
2.2x
approximate median peak-to-median active-month billed amount
This is a triage signal for seasonality, project starts, equipment purchases, accounting timing, or source-data problems. It is not evidence that costs rose or that the peak is avoidable. Cohort members also had unequal active-month coverage, so compare each entity with its own trailing baseline.
The approximate median source ID interacted with 18.6 vendor names in a median active month. That makes a monthly exception review more practical than attempting to inspect every annual vendor relationship with equal intensity.
Method and reproducibility
Every headline is derived from company-level aggregates, not pooled dollars
- Retain the latest bill revision per source-company-plus-bill key by update time, exclude deleted rows, require a parseable 2024 issue date and named vendor, and keep positive header totals no greater than the source-record 1,000,000 ceiling.
- Require at least 20 eligible bill headers per source company ID, compute each metric within source ID first, then calculate approximate company-level medians so a high-volume source ID does not dominate the result.
- Normalize reviewed major vendor-family aliases for the published vendor table and report minimum-activity sensitivity where the cohort rule materially changes prevalence.
- Publish software-detector control ratios as a separate Level field observation. Do not merge them with the historical AP cohort or call the surviving review amount savings.
Query receipts
d35d06f3-df04-4695-b50f-a7fac54cc92a
SHA-256 f7e0c6975daa102649a0416f4121bb435426f78c521611e3a0a83dc068f6edc2453e47ee-9d5f-43b3-8f48-7ab224d77817
SHA-256 140402af5408f9ea2cf4325aa109507b7bacfb165f39ea4046f1a07f404ad40605a30667-0f6d-48c7-a568-7e713a73f702
SHA-256 5a23c210c1d5c0475095b0ec818d7baff0f2b2100722e16244a30ef33ca10d4f0070ec92-f10d-4eda-84dd-99702df893d4
SHA-256 f9b6b347acb5d7b3aecdfa58ff075a2aee0e0c9705726e996f18c753c5763ee485ebe645-4958-4824-bcbf-3b66fedffeb7
SHA-256 2463ac0748c0d77dcda904cb3c03b357f24b39116c4ffc10bae74a3f5c79f721
Deliberately withheld
- customer and source-company identifiers
- raw bill or transaction rows
- uncommon merchant names
- software-vendor cells that fail minimum-size, product-scope, or dominance review
- PO, payment-term, job-linkage, freight, and purchaser-field benchmarks whose semantics remain unverified
- claims of contractor market prevalence, cash paid, realized savings, or avoidable expense
Questions about the study
Is this a contractor industry benchmark?
No. It is Level historical AP research across 652 eligible source company IDs. Those IDs were not independently deduplicated and were not verified as contractors. The findings are useful operating diagnostics, not contractor-market percentiles.
Does billed amount mean cash paid or expense?
No. A bill header is an obligation record. It is not proof of payment, material consumption, job cost, negotiated price, overpayment, or realized savings.
Why publish the vendor rows?
The ten major, commonly known vendor families have large cells and help show the difference between big-box, trade-supply, and rental bill footprints. Uncommon merchants and small or dominated software cells remain withheld.
Can I reproduce the calculations?
The downloadable JSON carries the metric definitions, cohort envelope, query IDs, aggregate result hashes, and derived values. The CSV exposes the findings and rows. Raw customer and source-company records are not public.
Grow without losing control of the numbers
Drop your info and we’ll show your real margin after callbacks and rework, which crews and jobs are worth the time, and how much cash is sitting in work you have already finished.
2,200+ service businesses benchmarked·$13.25B in revenue analyzed·Weekly action cadence
No commitment. Real numbers, not generic advice.