Where our numbers come from.
We sell a product whose argument is that a claim without provenance is not evidence. It would be difficult to make that argument on a page of unsourced statistics, so here is every figure we use, and the ones we have decided not to.
Figures we use
Gartner · 2025 · 506 CIOs
McKinsey, with the University of Oxford · 2012 · 5,400+ projects
McKinsey, with the University of Oxford · 2012 · 5,400+ projects
McKinsey · 2019
BCG · 2020
Some of these are old. The McKinsey overrun study is from 2012 and remains the largest of its kind; we show the year rather than hiding it, because a figure whose age is concealed is being used to imply a currency it does not have.
Figures we refuse to use
Every one of these is more dramatic than the numbers above, and every one of them appears routinely in this market's marketing.
Standish CHAOS project failure rates
Discredited in the peer-reviewed literature (Jørgensen; Eveleens & Verhoef in IEEE Software; Glass in CACM) and unverifiable because the underlying data has never been public. A technically literate CIO may well know this, and using it would cost more credibility than the number buys.
Any "% of projects fail due to unvalidated assumptions" statistic
It was searched for specifically and does not exist. Publishing one would mean inventing a statistic — on a site whose argument is that assumptions must be labelled as assumptions.
"Decision debt" presented as a researched phenomenon
A useful phrase, not a finding. It can be used as an idea in our own words; it cannot be cited.
"70% of VMware customers will have left by 2028"
A vendor-blog figure that contradicts Gartner’s own 35%. Using the larger number because it suits us is the exact behaviour this product exists to catch.
If you find a figure on this site without a source and a date next to it, it is a bug.Tell us and we will fix it or remove it.