Certificate of Insurance Statistics for Construction

The claim as repeated. Versions in use: "9 out of 10 insurance certificates contain material misrepresentations," "IRMI audits found 9 out of 10 certificates failed compliance checks," "an ARMR survey in 2005 found..." and "research shows 90%."

Where it came from. David Dybdahl, American Risk Management Resources Network (ARMR). First write-up: R&R Magazine, July 8, 2015. Then IRMI Expert Commentary, July 17, 2015 (fetched in November 2026). Dybdahl used a roughly 90% figure for restoration general liability policies from 2012 on, and restated it in 2017, April 2021 and August 2021 ("more than 80% of the time").

What it shows. One client's vendor pool: artisan and restoration contractors working under one insurer's master services contract. Phase 1 compared certificates with the contract, and all certificates looked 100% compliant. Phase 2 compared the actual policies with the contract, and more than 90% did not meet the requirements in a material way. The IRMI text reads: "Reading between the lines, 9 out of 10 of the insurance certificates ... contained material misrepresentations."

What it does not show. It is not an industry rate. It is not "90% of certificates fail review." Certificates passed review in phase 1. It is not construction-wide, and not current. Later repeaters moved the date to 2005, credited IRMI as the auditor (IRMI published it), and in one case inverted the finding. A separate "90%" in a 2025 BOMA session write-up is a speaker's estimate with no study.

Verdict. The audit is real, specific and sourced, and it makes a good point: a certificate can look perfect while the policy behind it does not match the contract. It supports a narrow sentence about one client's vendors in 2015. It does not support any claim about all certificates. One careful repeater, an insurance agency's 2025 blog, quotes the "artisan contractor" qualifier and links the source.

Safe wording. "One 2015 audit of a single client's restoration contractors found more than 90% carried policies that did not match the contract, even though every certificate looked compliant."

The claim as repeated. "On average, 70% of certificates are non-compliant when they're first received." Variants: "seven out of every ten COIs have been shown to be non-compliant" (Vertikal, 2019), "70% of building contractors are noncompliant" (a 2017 Forbes contributor post), "over 70% ... on construction projects" (2025, credited to "industry data"), and "according to Jones Insurance, 7 out of 10" (2026).

Where it came from. The earliest dated appearance found is a myCOI blog post dated February 10, 2017 (read in November 2026 in an archive.org copy). It gives no source, sample or period. It also says 30% of revised certificates are still not compliant. Two other pages credit the figure to myCOI's founder. It is likely myCOI's own processing experience, but that is an inference.

What it shows. That a COI-tracking vendor has said for nine years that most certificates come in with a problem. Nothing more can be read from it, because no data is published.

What it does not show. No sample, no definition of "non-compliant," no count of certificates, no industry, no year. The unit also changes across repeaters (certificates, contractors, COIs "on construction projects"). One repeater's link to Jones for the same figure does not contain it. The 2015 IRMI article has its own 70% (shown in section A), about policy gaps, with no link found between the two.

Verdict. The number may reflect real experience. It cannot be checked, so it cannot be cited as a finding.

Safe wording. "Vendors often say about 7 in 10 certificates fail on first submission. None has published the data."

The claim as repeated. "75% of third-party vendors do not meet the insurance requirements of the companies that hire them," often as a construction or COI error rate.

Where it came from. Evident's "State of Third-Party Insurance Verification" report, released June 23, 2021, and repeated in its 2022 edition. Read in November 2026 through the Metro Atlanta CEO reprint (the Business Wire release returned 403). Evident's own wording says the report draws on "tens of thousands of verification records."

What it shows. In the records of Evident's enterprise customers, 75% of the third parties did not meet every requirement. The reprint also gives a 25% average rate of verified third parties, 23% of vendors that never responded and at least 1 in 10 that fall out of compliance without telling the customer. The 2022 coverage adds that of the 75% that failed, 25% had no appropriate coverage at all.

What it does not show. It does not say which industries, which requirements or how "meet" is defined. The full report is gated. The 2023 release lists construction among many industries but gives no construction rate. The 75% includes vendors who did not respond, so it mixes "did not answer" with "answered with a gap." It is not a share of certificates containing errors.

Verdict. This is a real dataset from a verification vendor, with a stated source. It supports a sentence about Evident's customer data. It does not support a sentence about construction or about certificate error rates.

Safe wording. "Evident, a verification vendor, reported in 2021 that 75% of its customers' third parties did not meet every insurance requirement."

The claim as repeated. "On average, 70% of certificates are non-compliant when they're first received."

What we found. The earliest dated source we found is a myCOI blog post from February 10, 2017, now hosted at illumend.ai. No sample, method or period is given anywhere.

Do not cite this as a measured finding.

The claim as repeated. "90% of COIs are initially incorrect."

What we found. The article gives the speaker's own estimate ("he estimates") and cites no study. The setting is property management.

Do not cite this as a measured finding.

The claim as repeated. "Insurance fraud steals at least $308.6B every year from American consumers."

What we found. The page gives no year, producer or method. The NAIC page that repeats the figure does not mention construction, premium fraud or certificates. The figure covers all insurance lines.

Do not cite this as a measured finding.

The claim as repeated. Insurance fraud, including falsified COIs, costs the US construction industry billions, credited to the NAIC.

What we found. The NAIC page that Constrafor cites says nothing about construction or certificates.

Do not cite this as a measured finding.

The claim as repeated. "An estimated 10% of COIs in the construction industry are either fake or contain inaccurate information," credited to IRMI, 2018.

What we found. The IRMI link on that page returns a 404 error.

Do not cite this as a measured finding.

The claim as repeated. "Nearly 40% of organizations experienced insurance-related gaps ... a 2023 survey by RIMS."

What we found. The sentence is on the TrustLayer page. We did not find the RIMS survey.

Do not cite this as a measured finding.

The claim as repeated. "The actual response rate is more like 30% to 40%."

What we found. The article is written in the first person by an Evident executive. It gives no sample or method.

Do not cite this as a measured finding.

The claim as repeated. "Up to 20.5 percent" of construction workers who should be employees are misclassified or paid off the books.

What we found. The report cites the figure with an in-text "(5)" that the report's reference list does not resolve. We could not trace it to an original source.

Do not cite this as a measured finding.