The index separates three different numbers: broader platform activity, project-level case-study totals, and the individual issue records available for reproducible public analysis. This prevents a large headline number from being mistaken for a fully downloadable dataset.
- published projects
- 521
- project-level findings
- 90,768
- structured public records
- 40,401
- broader platform total
- 187,000+
Key finding: 67.6% were Critical or High severity
Across the 40,401 structured public records, 6,282 were tagged Critical and 21,056 were tagged High. Severity indicates review priority; it does not mean every potential finding was accepted unchanged by the project team.
Severity distribution
The public issue records retain the severity assigned in the published case-study data. Critical and High findings account for more than two-thirds of the structured sample.
Recurring issue-title patterns
These groups use disclosed, case-insensitive term matching on issue titles. They are indicators of recurring documentation topics, not a machine-learned taxonomy. One title may match several groups, so the percentages must not be added together.
- Specifications, schedules & referencesTitles mentioning specifications, schedules, callouts, drawing indexes, title blocks, or broken sheet/detail references.See the related review guide
- Dimensions & geometryTitles mentioning dimensions, elevations, grid lines, clear width, clearance, height, width, length, or slope.See the related review guide
- Missing informationTitles explicitly identifying missing, omitted, absent, incomplete, unshown, or insufficient information.See the related review guide
- MEP systemsTitles naming mechanical, electrical, plumbing, HVAC, duct, piping, panelboard, conduit, airflow, ventilation, or water-heater conditions.See the related review guide
- Codes & standardsTitles explicitly naming code, a model-code acronym, or a commonly referenced design standard.See the related review guide
- Fire & life safetyTitles mentioning fire, life safety, egress, exits, smoke, sprinklers, firestopping, or occupant load.See the related review guide
- ADA & accessibilityTitles mentioning ADA, accessibility, wheelchairs, maneuvering clearances, grab bars, reach ranges, or curb ramps.See the related review guide
Source-category mix
Source categories are normalized into eight reporting groups. The largest group is General and coordination, which is intentionally preserved instead of guessing a discipline for cross-sheet findings.
Project-type context
The table shows project types with at least eight published records. These are descriptive project totals, not cross-type error rates: set size, review scope, design stage, and the commercial-heavy sample differ substantially.
| Project type | Projects | Median findings | Average findings |
|---|---|---|---|
| Commercial | 235 | 76 | 134 |
| Civil | 84 | 74.5 | 136 |
| MEP Commercial | 79 | 163 | 308 |
| Residential | 40 | 39 | 82 |
| Multifamily | 26 | 163 | 411 |
| Education | 17 | 114 | 143 |
| Tenant Improvement | 11 | 91 | 264 |
| Healthcare | 10 | 45.5 | 132 |
| Infrastructure | 9 | 82 | 176 |
How the index was built
- The project sample contains 521 public case-study records. Records are not necessarily unique buildings. Project-level issue totals are summed from each record; selected published excerpts may be smaller.
- The structured issue sample contains every uniquely identified issue record published in the matching case-study data modules. Some case studies publish a subset of the complete report, so this total is lower than the sum of project-level totals.
- Severity percentages use the Critical, High, Medium, and Low labels stored on each published issue record.
- Discipline groups normalize the source category field into eight reporting groups. General and coordination findings remain separate rather than being guessed into a discipline.
- Topic clusters use disclosed, case-insensitive term matching on issue titles. Clusters overlap, so percentages should not be added together.
Why project totals and published issue records differ
The project-level total sums the finding count reported for every indexed case study. The structured-record total counts the individual issue objects published with those case studies. Some public pages provide selected findings rather than every item in the full customer report, so the structured total is lower.
Limitations
This is a descriptive sample of projects reviewed with InspectMind, not a random sample of the construction industry. Project mix, set size, design stage, review scope, and publishing depth vary. Potential findings require professional validation.
Download and cite the aggregate data
The aggregate files contain no customer-identifying project data. Cite as: “InspectMind, 2026 Construction Drawing Error Index,” accessed with the date you viewed this page.
Explore the issue library
- Browse 120 anonymized topic-by-discipline benchmarks
- Construction Drawing Dimension Conflicts
- Missing Details in Construction Drawings
- Common MEP Coordination Issues in Construction Drawings
- Fire and Life Safety Drawing Errors
- ADA and Accessibility Drawing Errors
Frequently asked questions
- How large is the Construction Drawing Error Index dataset?
- The 2026 edition analyzes 521 published case-study projects with 90,768 project-level findings and 40,401 structured public issue records.
- What percentage of published issues were Critical or High severity?
- Critical findings represented 15.5% of structured public issue records and High findings represented 52.1%, for 67.6% combined.
- What recurring drawing-error patterns appeared most often?
- In disclosed title-term analysis, specifications, schedules, and references appeared in 6,918 issue titles; dimension and geometry terms appeared in 6,888; and explicit missing-information terms appeared in 3,530. Topic groups overlap.
- Can the Construction Drawing Error Index data be downloaded?
- Yes. InspectMind publishes the aggregate report as both CSV and JSON. The files contain summary, severity, discipline, topic-cluster, and project-type statistics without customer-identifying project data.