Standish CHAOS Report: The Source Behind "66% of IT Projects Fail"
The CHAOS Report is the single most-cited source on IT project failure. This page gives the methodology, sample, definitions, and the canonical citation so you can use the number with confidence.
The classification: success, challenged, failed
CHAOS uses three buckets, defined consistently since 1994. A project is "successful" only if it delivered on time, on the original budget, and with the agreed feature set. Anything that slipped on cost, schedule, or scope is "challenged". Anything that was cancelled before completion or never used post-delivery is "failed".
| Outcome | Definition | 2020 share (large projects) |
|---|---|---|
| Successful | On time, on budget, agreed features delivered, user-satisfaction met | About 31% |
| Challenged | Completed but late, over budget, or with reduced scope | About 50% |
| Failed | Cancelled, never used, or written off | About 19% |
Shares vary slightly by report year and project-size band. Small projects fare considerably better than large ones.
Sample, methodology and limitations
Standish surveys IT executives directly. The CHAOS database is reported to hold more than 50,000 project records, with samples drawn for each annual report. The exact sample frame is not published, and the raw data is proprietary, which has drawn academic criticism (Eveleens and Verhoef 2010 in IEEE Software argued the report's definitions inflate failure rates).
For citation purposes you should treat CHAOS as an industry survey, not an audited dataset. It is reliable for direction (large projects struggle more than small ones, agile delivery outperforms waterfall) but the absolute percentages should be paired with methodology context.
How to cite
Authoritative source: standishgroup.com
Direct CHAOS report PDFs are paywalled. The PMI library and IEEE Software contain peer reviews and secondary citations that are openly accessible.