How does "Anchor on the base rate before adding inside-view details" work?

Anchoring research shows that adjustments from an initial number are insufficient — people consistently end up closer to their starting point than the evidence warrants. If you start from the inside view (which is optimistic) and try to adjust downward toward the base rate, you will not adjust far enough. Starting from the base rate and adjusting upward for genuine distinguishing features of your project produces much more accurate final estimates.

How does "Forecast a distribution, not a point estimate" work?

Point estimates communicate false precision and suppress useful information about variance. A project that takes six months on average might have a 10th-percentile outcome of three months and a 90th-percentile outcome of eighteen months — a distribution that should change planning decisions, resource allocation, and contingency budgeting in ways a point estimate conceals. Communicating distributions makes uncertainty legible and improves downstream decisions.

How does "Distinguish cognitive optimism bias from strategic misrepresentation" work?

Flyvbjerg identified two sources of forecast error: optimism bias (genuine cognitive distortion) and strategic misrepresentation (deliberate underestimation to win approval or funding). Reference class forecasting corrects optimism bias. Strategic misrepresentation requires institutional accountability mechanisms, not cognitive tools. Conflating the two leads to applying the wrong intervention.

How does "Update your forecast incrementally as new evidence arrives" work?

Forecasts that are not updated in response to evidence are not forecasts — they are commitments in disguise. The updating mechanism is Bayesian: new evidence shifts the probability, with the magnitude of the shift proportional to the evidence’s diagnostic value. The failure mode is updating too slowly (anchoring on the original estimate) or updating too violently (overweighting recent salient events).

How does "Conduct post-project debriefs to contribute to the class data" work?

Reference class forecasting is only as good as the class data behind it. If practitioners never record honest outcomes, the class data available to future forecasters remains thin or systematically biased (projects that went badly are often not published). Disciplined post-project recording creates the base rate infrastructure that corrects the next round of planning.

Tetlock & Kahneman — Superforecasting

How do you make more accurate forecasts by using base rates instead of inside-view thinking?

Using base rates and outside-view anchors to forecast more honestly

Short answer

Reference class forecasting, developed by Daniel Kahneman and Amos Tversky and formalized by Bent Flyvbjerg, improves forecast accuracy by anchoring on the statistical distribution of outcomes for similar past projects rather than on the details of the current one. The method reliably corrects the optimism bias that inflates cost, time, and benefit estimates in planning — the evidence base here is real and specific.

When people plan projects, they build a detailed narrative about their specific situation and forecast from it — the "inside view." Kahneman recognized that this process is reliably optimistic, because the narrative focuses on what could go right and ignores the distribution of outcomes for comparable projects. Reference class forecasting forces an anchor on that distribution before the inside-view details are added back. The practices below operationalize this outside-view discipline and are useful far beyond project management.

The practices (6)

Why it works

Inside-view planning treats your situation as unique, so historical failure rates feel irrelevant. Identifying the reference class forces acknowledgment that your project belongs to a category with a known distribution of outcomes — and that distribution is the best prior for your forecast before any project-specific information is added. The mechanism is Bayesian: the class distribution is the prior; inside-view details are updating evidence, not the starting point.

How to do it

1Define what type of project or decision you are making (e.g., "software project over six months with a new team," not "my project").
2Search for historical data on completed projects of that type: duration, cost overruns, success rates.
3Choose the reference class that is specific enough to be meaningful and broad enough to have statistical power.
4Record the median and the range (10th–90th percentile) of outcomes for this class before opening your project plan.

Evidence

Observational

Flyvbjerg and colleagues analyzed hundreds of large infrastructure projects and found systematic cost overruns averaging 28% for roads and much higher for rail and tunnels. Reference class forecasting has since been formally recommended by the UK Treasury and the American Planning Association as a planning tool.

Honest caveat: The method requires a usable reference class; for genuinely novel situations (first-ever technology, unprecedented events) the class is too thin or nonexistent, and base-rate anchoring provides false precision.

— Flyvbjerg, Holm & Buhl (2002), "Underestimating costs in public works projects," Journal of the American Planning Association
— Kahneman & Tversky (1979), "Intuitive prediction: Biases and corrective procedures"

Common mistake: Defining the reference class too narrowly ("projects exactly like mine") to preserve optimism — if you can only find two past examples, you do not have a reference class, you have anecdotes.

Go deeper on this practice →

Practice this with IX Coach

Reading about a practice changes nothing on its own. IX Coach turns these into a guided, adaptive routine — discerning where you are in real time and walking the practice with you, session after session.

Start with IX Coach →