Why IT Projects Blow Their Budgets: The One in Six
The short version: the average overrun is the wrong number to plan with. In a study of 1,471 IT projects, costs overran by 27% on average, but one project in six overran by an average of 200%, with schedules nearly 70% late. Those few projects are where the damage is. If each project has a one-in-six chance of becoming one of them, a portfolio of six projects has a 67% chance of containing at least one, and a portfolio of twelve 89%. A distribution with rare, huge outcomes has fat tails, and an average-based contingency doesn't cover them.
The new customer system has a budget of $2 million and a 10% contingency, "in line with the industry average". The project manager has a slide: IT projects overrun by about a quarter on average, so 10% is a bit tight but fine with good management.
Kevin wonders what "on average" is hiding. It turns out to be most of the risk.
The study
Researchers at Oxford collected 1,471 IT projects, from public bodies and companies, with their planned and actual costs and schedules. Two numbers came out:
| Cost overrun | |
|---|---|
| Average across all projects | 27% |
| The worst one in six | about 200% on average, and nearly 70% late |
The average makes IT projects sound like a modest, predictable overspend. The one in six makes them sound like a lottery with a nasty prize. Both are true; only the second one sinks companies.
Fat tails
In a normal, bell-shaped world, outcomes far from the average are vanishingly rare. In a world with fat tails, extreme outcomes are rare but not that rare, and when they happen they're huge. Project costs, market crashes, insurance losses and city floods all look like this. The average still exists, but it describes almost no actual project: most come in near budget, a few explode.
That's also why the planning fallacy is so expensive here: optimistic plans meet a distribution where the downside is long and the upside is short.
A portfolio of projects
If each project, independently, has a one-in-six chance of becoming one of the extreme ones, the chance that a portfolio contains at least one:
| Projects in the portfolio | At least one with an extreme overrun |
|---|---|
| 1 | 17% |
| 3 | 42% |
| 6 | 67% |
| 12 | 89% |
A company running a dozen IT projects should expect at least one blow-up, and plan for it, rather than treat it as bad luck when it arrives.
Chance a portfolio contains at least one extreme overrun
Statistics for noobs: ask about the tail, not the middle
When outcomes are lopsided, the average and the typical case can both look calm while the risk sits in the extremes. Ask: what does the worst 10% or 20% look like, and how often does it happen? A contingency that covers the average overrun protects against the cases that don't need protecting.
Where it costs you money
Investment returns have fat tails too. "The market's average year is +8%" says nothing about the one year in ten that falls 20% or more. Plan your savings for the tail: how much could you lose in a bad year, and could you hold on?
The moral
Kevin's project came in at 18% over budget, which the project manager presented as a success against the industry average. The other project the company started that year, a "simple" data migration, ran 230% over and two years late. The lessons-learned review recommended raising contingency to 15%.
What this study can't tell you
- A collection, not a random sample. The projects were the ones the researchers could get data on; the true shares may differ.
- The one in six is a summary. The study describes the extreme group by its average; the overruns inside it vary widely.
- Our portfolio table assumes independence. Projects sharing staff, suppliers or technology may fail together, which makes the risk lumpier still.
Kevin's project and the data migration are an illustration. The 1,471 projects, the 27% and the one-in-six figures come from a study published in the Harvard Business Review in 2011. The portfolio table is our own calculation. Educational content, not business or financial advice.