The cost of finding out late

A bank was a few months from go-live when its finance department finally got pulled in. They brought an accounting change the new performance-management system had created, and nobody had built for it. The modules had to be reworked. McKinsey, which documented the case, puts the result plainly: the launch slipped more than three months, at a cost of more than $8 million.

Nobody in finance learned about that change in month eleven. They knew in month one. The project was not set up to hear them.

That is the pattern behind most late requirements. It is worth understanding because it is the most fixable expensive thing on a project.

What late costs

McKinsey and the University of Oxford looked at more than 5,400 IT projects above $15 million. On average they ran 45 percent over budget and delivered 56 percent less value than planned. Every additional year on the schedule added 15 percent to the overrun.

The Standish Group sees the same thing across more than 25,000 projects. Sixty-one percent of small projects succeed. Six percent of the largest do. Standish has said since 1994 that size, which in practice means time, is the single most important factor in how a project ends.

So the schedule is the cost. Anything that puts weeks back on the calendar is money. Anything that pulls a surprise forward from month eleven to week one is more.

Why the surprise is always a person

Ask what causes the slippage and the answer is requirements. The Project Management Institute found 47 percent of unsuccessful projects fail because of inaccurate requirements management. But a requirement is not a document. It is something a person knows.

And people do not say what they know. In an interview study of 40 employees, Milliken, Morrison and Hewlin found 85 percent had held back an issue from a supervisor that they believed was important. Of those, 74 percent said colleagues who knew the same thing stayed quiet too. The reasons were fear of being labelled a problem, and the belief that saying it would change nothing.

Gallup found only 8 percent of employees strongly agree their organisation acts on surveys. Boston Consulting Group found three in four executives believed they had good engagement on their transformation, while only one in three had it from middle management.

Leaders think the project is aligned. The people who know it is not have decided not to say so. That is where the month-eleven requirement lives.

Why your current tools cannot find it

You have two ways to ask. A survey reaches everyone, but it cannot ask a follow-up, and only 8 percent of the people filling it in strongly believe anything will come of it. An interview gets the follow-up, but a consultant can schedule and synthesise ten or fifteen of them, it takes months, and the person asking works for the boss.

One tool cannot hear. The other cannot reach. The requirement that matters falls between them.

What changes when everyone is heard

Candidly AI runs one on one voice conversations at scale. Participants call a dedicated number whenever it suits them. The conversation adapts, asks the follow-up, and can be fully anonymous. Hundreds of people in days, not fifteen in four months. Every conversation is read by a person, and the Ground Truth Report gives you the themes with the verbatim quotes behind them, every participant quoted at least once.

Two things happen. The listening window collapses from months to days, and that time comes straight off the front of the critical path. And the person in finance who knew in month one gets asked, in a setting where saying it costs them nothing.

The research supports the mechanism. It does not put a number on how much risk that removes, and we are not going to invent one. What it does price, clearly, is the schedule.

You buy back the schedule, and you buy back the surprise.

Hear from a hundred people this month, or find out in month eleven what the other ninety thought.

How much do large IT projects typically go over budget?

McKinsey and the University of Oxford analysed more than 5,400 IT projects with budgets above $15 million. On average those projects ran 45 percent over budget and 7 percent over time while delivering 56 percent less value than predicted. Seventeen percent overran by more than 200 percent, badly enough to threaten the company running them.

Does a longer project schedule increase cost overruns?

Yes. In the McKinsey and Oxford dataset, every additional year a project is scheduled to last increases cost overruns by an average of 15 percent. The Standish Group's CHAOS 2015 analysis of more than 25,000 projects found 61 percent of small projects succeeded against 6 percent of the largest, with 43 percent of the largest failing outright.

What is the most common cause of project failure?

Requirements. The Project Management Institute found 47 percent of unsuccessful projects fail to meet their goals because of inaccurate requirements management, and 39 percent of organisations name inadequate requirements gathering as the primary cause of failure. McKinsey found that failure to manage strategy and stakeholders, together with mastering technology and content, typically causes about half of all cost overruns.

Why do stakeholders raise concerns late instead of early?

Because they choose not to speak. In an exploratory interview study of 40 employees, Milliken, Morrison and Hewlin found that 85 percent had been in a situation where they felt unable to raise an important issue with a supervisor, and 74 percent of those said colleagues who knew about the same issue also stayed quiet. The two leading reasons were fear of being labelled negatively and a belief that speaking would change nothing.

What is Candidly AI?

Candidly AI is a Calgary-based Canadian platform that holds hundreds of one on one voice conversations with the people whose perspective a decision depends on. Participants call a dedicated number on their own schedule, the conversation adapts with follow-up questions, and the organisation receives the Ground Truth Report: themed findings backed by verbatim quotes, with every participant quoted at least once. Candidly AI is ISO 27001 Certified and SOC 2 Audited, runs on fully Canadian infrastructure with self-hosted models, and makes no third-party model API calls.