About 60% of the RFIs on our own construction projects were caused by the construction documents themselves. Not field conditions, not owner changes, not contractors asking how to build something — a defect in the drawings and specifications: information that was missing, two documents that disagreed, or an instruction that could not be acted on as written.
That number gets asserted a lot in this industry, usually without a source. So in August 2026 we measured it, on data we own: the complete RFI logs from three multifamily projects our founder’s firm developed — 827 RFIs with full question text, written between 2022 and 2025 by three different general contractors to three different design teams. This page is the study: the definition, the numbers, the method, and the limits.
What we measured
The dataset is the complete project-management RFI log from each project — every formal RFI asked over the life of the job, not a sample of interesting ones. Project A is a garden-style multifamily development in the Mountain West (357 RFIs). Projects B and C are mid-rise multifamily projects in the Southeast (255 and 215 RFIs). Duplicate log entries were removed and page-break artifacts stripped before classification; 827 of 828 unique RFIs had enough question text to classify.
These are our founder’s own projects. The documents whose defects we are counting were the documents his teams built from, which is exactly why we are comfortable publishing the result: this is not a study that blames anyone we would not blame ourselves.
The definition, stated before the number
An RFI counts as caused by a document defect only when the question exists because the construction documents were deficient in one of three ways:
- Omission — required information is absent: a detail, dimension, mounting height, finish, or schedule entry that should exist and does not.
- Conflict — two documents disagree: sheet against sheet, plan against schedule, drawing against specification.
- Ambiguity — the information is present but genuinely cannot be acted on as written.
Everything else counts against the claim. Substitution requests, field and existing conditions, means-and-methods questions, owner-directed changes, product availability problems, and administrative transmittals were all classified as not document-caused — including cases where a document review might plausibly have helped. Polite phrasing was ignored: nearly every RFI says “please confirm” or “please clarify,” so classification was on the substance of what was asked, never the opener.
Results: 60.2% traced to the documents
| Root cause | Share of 827 RFIs |
|---|---|
| Missing information (omission) | 32.2% |
| Documents conflict with each other | 18.5% |
| Genuinely ambiguous as written | 9.3% |
| Document defect, total | 60.2% |
| Field or existing conditions | 14.5% |
| Owner-directed changes | 12.3% |
| Means and methods | 4.5% |
| Substitution requests | 2.8% |
| Product availability | 1.3% |
| Administrative and other | 4.4% |
The result held independently on each project: 62.2%, 62.8%, and 55.3%. Three different general contractors, three different design teams, two regions, one consistent answer. The single largest category of RFI on all three projects was not a question about how to build the building — it was a request for information the documents should already have contained.
What a document-caused RFI looks like
Paraphrased from the logs, with identifying details removed:
- Conflict. The architectural floor plan shows the balconies at 12′-11″ wide. The structural foundation plan shows the same balconies at 13′-2″. The framer asks which discipline to lay out from — for the entire building.
- Omission. Light fixtures appear throughout the electrical sheets. No mounting heights are given anywhere for any of them. The electrician asks for all of them at once.
- Conflict. The plumber asks whether the corridor floor sheathing must be fire-retardant-treated. The architectural sheet defers to structural; the structural sheet defers back to architectural. No document answers.
- Ambiguity. The landscape plan says existing mature plantings are to be “relocated as needed.” Nobody can price or schedule “as needed.”
None of these are exotic. They are the ordinary texture of a set produced by multiple consultants on different timelines, and each one stopped work in the field until someone answered.
How this compares to published benchmarks
The most defensible published figures measure narrower things. The Navigant Construction Forum’s 2013 study — 1,362 projects, about 1.1 million RFIs — found 13.2% of RFIs “not justifiable”: questions that should never have been asked because the answer was already in the documents.¹ That is close to the inverse of our construct: Navigant counted RFIs the contractor should not have written; we counted RFIs the documents forced the contractor to write. The Construction Industry Institute is widely credited with an estimate that 30–50% of RFIs stem from errors, omissions, or ambiguities in the contract documents² — the same construct as ours, on which our first-party measurement lands higher, at 60%. We publish ours as our number on our projects, not as an industry constant.
What the preventable share costs
Navigant’s same study put the average processing cost of a single RFI — writing it, researching it, answering it, documenting it — at about $1,080.¹ At that rate, the 498 document-caused RFIs across these three projects represent roughly $540,000 in processing labor alone, about $180,000 per project — before counting a single change order, schedule day, or piece of rework that the underlying conflicts produced. Processing cost is the floor of what a document defect costs; it is just the most measurable layer.
Method and limits
Each RFI’s full question text was classified into the categories above by a large-language-model classifier run at temperature zero, with the complete category definitions in the prompt and a strict instruction to classify on substance rather than phrasing. A 100-RFI random sample was then human-reviewed against the source text: roughly 86 were clearly correct, the disagreements split evenly in both directions, and the single unambiguous misclassification ran against the headline — an RFI caused by a code-clearance defect in the design that was filed as an owner change because the contractor proposed the fix. The classifier systematically labels an RFI by how it is framed rather than what caused it, and contractors frame politely. For that reason we treat 60.2% as a floor.
The limits are real: three projects, one property type (multifamily), one owner’s ecosystem, and RFIs measure only the defects the field noticed and formally asked about. This is not a claim about all construction. It is a complete, defined measurement of every RFI on three real projects, published with its method — which is more than can be said for most numbers in this industry, including the unsourced “60%” we set out to check.
What to do with this
If 60% of RFIs on a project trace to the documents, then RFI volume is mostly a document-quality problem wearing a workflow costume — and the highest-leverage moment to attack it is before issue, not after mobilization. The three defect classes this study counted — missing information, conflicting documents, ambiguous instructions — are exactly what a cross-discipline document review looks for, whether a person runs it with a highlighter or software runs it sheet against sheet. Flikt.AI performs that review on the 2D PDF set you already have, same day, and every finding cites the sheet it came from so your team can judge it. How to reduce RFIs before construction starts covers the working sequence; what an RFI actually costs covers the economics.
Frequently asked questions
What percentage of RFIs are caused by the construction documents?
On the three multifamily projects in this study — 827 RFIs classified, 2022–2025 — 60.2% traced to a defect in the construction documents: 32.2% missing information, 18.5% conflicting documents, 9.3% genuinely ambiguous. The Construction Industry Institute’s widely cited estimate for the same construct is 30–50%; our first-party measurement came in above it.
Does that mean 60% of RFIs are preventable?
Preventable in the specific sense that the cause was in the documents before construction started, so a pre-issue review could in principle have caught it. It does not mean a review catches every one, and it does not count the other 40% — field conditions, owner changes, means and methods — which no document review prevents.
How were the RFIs classified?
By a language-model classifier at temperature zero using the category definitions published on this page, on the full question text of each RFI. A 100-RFI random sample was human-reviewed; the error profile was balanced, and the one clear misclassification undercounted document defects rather than overcounting them, which is why the figure is treated as a floor.
Whose projects were these?
Three multifamily projects developed by our founder’s own firm, built by three different general contractors from three different design teams’ documents. The projects are anonymized because the point is the pattern, not the parties — and the pattern was consistent across all three: 62.2%, 62.8%, and 55.3%.
Is 60% an industry-wide number?
No, and we do not claim it is. It is a complete measurement of three real projects with the definition and method disclosed. The honest industry range for the same construct is the Construction Industry Institute’s 30–50% estimate; our data suggests the true share on ordinary multifamily work may sit at the top of that range or above it.
Sources
1. Navigant Construction Forum, Impact and Control of RFIs on Construction Projects, 2013 — 1,362 projects, ~1.1M RFIs; 13.2% “not justifiable”; ~$1,080 average processing cost per RFI.
2. Construction Industry Institute — the widely attributed 30–50% estimate for RFIs stemming from errors, omissions, or ambiguities in contract documents; we cite it as an attributed estimate, as CII’s underlying research is member-gated.
3. This study’s dataset: 827 RFIs from the complete logs of three multifamily projects, 2022–2025, classified August 2026. Definitions and method as stated above.
The pre-IFC coordination checklist
50 checks before you issue for construction, ordered by what actually goes wrong — built from 1,516 findings across 34 real plan sets. Comes as an assignable tracker (Excel) with status, owner and due date, plus a printable PDF. No call, no card.