How many ERP projects really fail?

There is no reliable failure rate for ERP projects. The figures circulating in presentations range from 50 to over 70 percent and come from surveys that each define „failure“ differently or not at all. Two other values are documented: 65 to 80 percent of large technology programmes exceed their budget or timeline, and only 25 to 35 percent achieve the intended impact on earnings and cash flow. Both come from McKinsey and describe large technology programmes, not ERP rollouts specifically. For your own planning they are therefore an indication, not a forecast.

Why there is no reliable rate

The problem is not the data, it is the definition. Ask five people when an ERP project has failed and you get five answers. Has it failed when it was cancelled? When it went live, but half a year later? When the budget was exceeded by forty percent but the system runs? When everything went to plan but the company sees none of the hoped-for effects?

Depending on where you draw that line, the same set of projects sits at 20 percent failures or at 70. That is exactly what explains the spread in the figures quoted. A percentage without a supplied definition is therefore an opinion with a decimal place, and it carries no decision.

On top of that comes a selection effect: nobody likes talking about cancelled projects, and published accounts come predominantly from those where things went well. Anyone estimating the rate from public sources is estimating from a skewed sample.

The two figures that are documented

65 to 80 percent exceed budget or timeline. This is the most frequently quoted figure, and it is almost always reported wrongly. It does not say that this many projects fail. It says that this many become more expensive or longer than planned. A project that takes three months longer is over plan and still a success.

25 to 35 percent achieve the intended impact. This is the more uncomfortable figure and the one too little is said about. It does not describe the course of the project but the outcome: whether the hoped-for effects on earnings and cash flow actually materialise. This is where the real loss sits, and it is mostly never measured, because after go-live nobody does the arithmetic again.

Both values come from the same McKinsey research and apply to large technology programmes, not to ERP rollouts in particular. Anyone quoting them as „ERP figures“ narrows them inadmissibly. We quote them anyway, because they are the only ones with a traceable origin, and we state the qualification along with them.

What this means for the mid-market

The mechanisms are the same, the order of magnitude is not. A programme with a budget in the hundreds of millions fails on coordination between dozens of sub-projects. A Business Central rollout with three hundred employees fails on something else, and mostly on the same three things.

• Processes that only get settled during the project. What nobody decided beforehand gets decided during the project, under time pressure and by the wrong people.
• Master data nobody looked at beforehand. Duplicates, inconsistent units of measure and a chart of accounts grown over time travel along and turn up during testing as a surprise.
• Decisions that take too long. It is not the wrong decision that costs the project, it is the one that never comes.

None of these three points is a software problem, and all three can be worked on before the project starts. That is why we consider the rate question the less useful one.

The reasons in detail: Why ERP projects fail

The more useful question than the rate

An industry rate tells you nothing about your project. Two numbers of your own do, and both can be gathered in a week.

First, the decision time. How many days pass at your company between a decision proposal and the approval? If the value runs into weeks, your project has a steering problem, regardless of how good the partner is.

Second, the growth in requirements. How many items have been added since the start that were not in the original scope? Steady growth without anything struck off in return is the most reliable early indicator there is.

These two values replace any rate. They do not describe what happened to others, they describe what is happening at your company right now.

The concrete signs: How do I recognise that my ERP project is failing?

Wenn es Sie bereits getroffen hat: The ERP project has failed. Can it be rescued?