Analysis Paralysis in Business Is Not a Leadership Failure

A CTO I was asked to help had been evaluating trading-platform replacements for fourteen months. His team had produced three vendor comparison matrices, two proof-of-concept builds, a 47-page migration risk assessment, and a board paper that recommended "further analysis." The platform they were trying to replace was eleven years old. Every month it stayed in production, the maintenance cost grew. Every month, the evaluation committee met, reviewed new data, and asked for more. Nobody could identify a single new fact that had emerged in the preceding six months. The same information circulated through the same committee and produced the same recommendation: more analysis. This is what analysis paralysis in business looks like from the inside.

His CIO told me, privately, that the committee had reached an informal consensus six months earlier. The process would not let them act on it. It is not a failure of nerve or leadership. It is analysis paralysis: a failure of process, not people. Analysis paralysis in business is what happens when governance machinery generates analysis indefinitely because it has no test for sufficient certainty: no mechanism for anyone in the room to say, we know enough.

How analysis paralysis in business destroyed BlackBerry

BlackBerry's co-CEO Mike Lazaridis saw the iPhone in January 2007 and recognised the threat immediately. "How did they do that?" he asked. His co-CEO Jim Balsillie dismissed it: "It's OK. We'll be fine."

What followed was not indecision by two weak leaders. It was nine years of strategic debate channelled through product councils, executive committees, and strategy reviews that produced competing directions without resolving any of them. Physical keyboard or touchscreen. Enterprise or consumer. Their first touchscreen device, the Storm, launched in 2008 with an unresponsive screen and broken software. The PlayBook tablet shipped in 2010 without native email. Each product reflected an organisation that could not close its own internal debate.

At peak, BlackBerry held more than 50% of the US smartphone market and a market capitalisation of $75 billion. By 2016, their share was below 0.1%. The company wrote off $485 million on the PlayBook alone and exited hardware manufacturing entirely.

Click to expand

I have seen this pattern in mining companies and aviation authorities. Two factions generating analysis, neither authorised to close the debate. BlackBerry had the data and the strategy reviews. What it lacked was a mechanism to say: we know enough to commit. Had their leadership surfaced the assumption underlying their strategy, one would have dominated: that business users would always prefer physical keyboards. By 2009, confidence in that assumption was collapsing. That signal alone would have forced a decision years earlier. Instead, the strategy consultants who supervised the reviews kept billing. The longer the internal debate ran, the larger the invoices.

Twenty-two years of analysis at Citibank

Citibank deployed a loan processing system called Flexcube in 1997. For twenty-two years, risk committees and technology steering groups reviewed whether to replace it. Compliance flagged it. Feasibility studies were commissioned. The system stayed.

In August 2020, a contractor using the antiquated interface checked the wrong box on a payment form. Instead of sending a $7.8 million interest payment to Revlon's lenders, Citibank accidentally wired $900 million. Three people verified the transaction. All three missed the error. Creditors refused to return roughly $500 million, and a federal judge ruled they could keep it.

I investigated a case with similar characteristics in the mining sector. Years of documented concern about a system everyone knew was failing, no mechanism that would let anyone say: enough. The documentation was immaculate. The decision was absent. Citibank's governance machinery produced the same result: twenty-two years of reports about the risk of keeping an obsolete system, and no verdict. Nobody asked the question that would have broken the cycle: given what we know about this system's limitations, do we have sufficient certainty that leaving it in place will not produce an unacceptable outcome?

The question was never asked because the process was not built to ask it. It was built to produce assessments. Nobody seems to have noticed that an assessment is not a verdict. Twenty-two years of governance produced an admirable compliance record. It also produced a $500 million loss. The compliance record is presumably still on file somewhere.

The pattern runs across every industry

A 2025 survey by CloudBees of more than 300 enterprise IT leaders found that 61% of organisations reported migration fatigue delaying projects by six months or longer. Seventy-six percent reported a drop in developer morale. Of course it dropped. The engineers who knew the answer were forced to watch a committee not reach it for six months. The consulting firms who recommended the governance processes that caused the delays are not, I notice, counted among those who missed revenue.

The mechanism is the same across every sector and every scale. A process generates vendor comparisons, proof-of-concept builds, committee reviews, re-evaluations. Each cycle produces new information. None of the cycles includes a test for whether the existing information is enough to decide. The process was built to produce documentation, not verdicts. It cycles because cycling is what it does.

Analysis paralysis in business is not a personality trait. It is an engineering defect in how organisations structure decisions. The cost is not abstract. It is the CTO watching his maintenance budget climb by $200,000 a month while the committee meets to discuss whether they have enough information. They had enough information six months ago.

What stops analysis paralysis in business

The Universal Decision-Making Method has a built-in gate that these governance processes lack. At Step 4, the Decider asks one question: do I have sufficient certainty that this decision will produce the intended outcome?

Sufficient certainty is not maximum certainty. As Roger Estall and I wrote in Deciding, "sufficient certainty does not necessarily mean the greatest amount of certainty possible because that could be wasteful of resources." There is no formula for sufficiency. The test is whether the assumptions underlying the decision have been surfaced and their significance assessed. If the Decider is satisfied that the remaining uncertainty is acceptable given the Purpose, the analysis stops.

If the answer is yes, stop analysing and implement with monitoring. If the answer is no, the method gives the Decider three bounded moves. Each is concrete. Each terminates. There is no fourth option labelled "continue reviewing."

The CTO I mentioned at the start had never been asked that question. His evaluation committee had no mechanism for it. When we surfaced the assumptions underlying his platform choice and classified them by significance, one assumption stood out: that the current platform could continue to meet regulatory requirements for another twelve months. Confidence in that assumption was low and falling. Influence on the desired outcome was high. That moved the decision from "needs more analysis" to "needs to be made."

The committee approved the migration within three weeks. The apparatus had generated fourteen months of analysis. The method produced a decision in an afternoon.

Grant Purdy is the co-author, with Roger Estall, of Deciding (2020), and the architect of the Universal Decision-Making Method.