The base rate is the answer before you know anything about this particular case, and it is usually closer to the truth than the answer after.
Why it matters when the plan changes
Every plan is described by the people making it in terms of its own specifics, which are genuinely distinctive, and that distinctiveness is what makes the base rate feel irrelevant and what makes it necessary. Ignoring the rate at which something occurs in a comparable population, in favour of case-specific detail, is a recognised reasoning error, not a rare lapse. How often do integrations of this size deliver on schedule is a question with an answer, and it is rarely the question anyone asks.
The tension is that finding the right reference class is genuinely hard. Too narrow and there are no cases; too broad and the rate is meaningless. That difficulty is real, and it is usually used as a reason to skip the exercise rather than to do it approximately, which is almost always better than not doing it at all. Forecasting practice that scores predictions against outcomes treats reference to prior cases as part of how judgement improves, not as an optional refinement.
In practice
A team estimates an integration at nine months from the bottom up, task by task. Nobody asks how long comparable integrations in the group have actually taken. The answer, available in the organisation's own history, is fourteen to twenty months, and the plan is committed against the estimate rather than the record.
Evidence
Ignoring the rate at which something occurs in a reference population, in favour of case-specific information, is a recognised reasoning error.
Base rate fallacy, Wikipedia (2026)Scored, feedback-driven forecasting improves judgement, and reference to prior outcomes is part of how that improvement happens.
Superforecasting and the Good Judgment Project (2026)
What it cannot tell you
A base rate describes how a class of cases has gone on average; it does not identify which features of this particular case will move it above or below that average. Chosen from too broad a class, it produces a number that is technically true and practically useless. It says nothing about mechanism, only frequency.
Questions
Find cases that resemble this one on the dimensions that drive the outcome, not on the ones that are easiest to match. For an integration that usually means size, number of units and whether systems are being merged, rather than industry or geography.
Build an approximate one and say it is approximate. A rate drawn from twelve loosely comparable cases with the limits stated is more useful than an estimate built entirely from the specifics of this case, which carries an optimism that nothing corrects.
No. It anchors the judgement, and case-specific information adjusts from there. Work associated with the Good Judgment Project treats scored, feedback-driven forecasting, which includes reference to prior outcomes, as central to improving judgement, not a bolt-on. Starting from specifics alone produces an estimate no one checks against comparable cases.
In their own history, usually unexamined. How long previous integrations took, how many prior transformations delivered their stated benefits, how often appointments at this level worked out. The data almost always exists somewhere and nobody compiles it, because compiling it is nobody's job.
Because it says this plan will probably go like the others, which reads as a judgement on the people making it. Documentation of the base rate fallacy in 2026 describes exactly this discomfort: substituting specific, vivid detail for a comparable class of outcomes because the class feels less generous.