Five errors account for most bad decisions: incomplete framing (the real options were never considered), anchoring (the first number or option seen became the baseline for everything else), confirmation bias (information was sought to confirm rather than test), sunk cost thinking (past investment was weighted over future value), and overconfidence in forecasts (uncertainty in timelines and outcomes was systematically underestimated). Each of these feels like good thinking from the inside, which is why they persist.
The five errors and why they repeat
Incomplete framing happens at the very beginning of a decision process, which is why it is so costly. The decision-maker defines the choice too narrowly, or accepts the first framing they encounter, and never considers options outside that initial frame. A person choosing between two job offers may never ask whether keeping their current job or creating a freelance path are also options. The options not named cannot be chosen.
Anchoring occurs when the first piece of information encountered — a salary number, a price, an initial estimate — becomes the reference point against which everything else is measured. Negotiations, valuations, and forecasts are all vulnerable to anchoring. The anchor sticks even when the person knows it is arbitrary.
Confirmation bias is the tendency to seek and weight information that supports an existing belief over information that challenges it. It does not require intent. It is a default cognitive pattern. The person conducting research on a decision often unconsciously selects the sources, questions, and evidence that confirm what they already think. The bias is strongest for decisions the person cares about, which makes it most damaging precisely where careful thinking matters most.
Sunk cost thinking is weighting past investment — time, money, effort — in a decision about future action. The past investment cannot be recovered regardless of which option is chosen next, so it is logically irrelevant to a rational forward-looking decision. In practice, most people find it very difficult to write off what they have already put in, which leads them to continue courses of action that a fresh analysis would not support.
Overconfidence in forecasts is the systematic underestimation of uncertainty. People predict narrower ranges of outcomes than the evidence supports, set timelines that are too optimistic, and underestimate how much circumstances can change. This error underlies the other four in important ways: an overconfident person is less likely to seek disconfirming information, less likely to question their initial frame, and more likely to dismiss warning signs.
What makes these errors hard to catch
Each of these mistakes feels like good thinking in the moment. Incomplete framing feels like focus. Anchoring feels like having a reference point. Confirmation bias feels like thorough research. Sunk cost thinking feels like commitment and follow-through. Overconfidence feels like clarity and conviction. The signal that distinguishes good thinking from these errors is subtle, and the errors often co-occur, with each reinforcing the others.
The most practical defence is to build in a structured challenge at the decision point. Before committing, ask: what options am I not considering? What would have to be true for my preferred choice to be wrong? What information have I not looked for? These questions do not guarantee catching every error, but they create enough friction to surface the most obvious ones.
Overconfidence as the meta-bias
Overconfidence is the bias that underlies and enables the other four. An overconfident decision-maker sees no need to question their initial frame, no reason to seek disconfirming evidence, and no cause to doubt their forecasts. The errors they make feel like sound judgment precisely because overconfidence removes the doubt that would otherwise prompt a second look. Calibrating your confidence to your actual track record, using a decision journal to track predictions against outcomes, is the most direct intervention.
A pre-decision error check
Before finalising any significant decision, run through these five questions. One: what options am I not considering? Two: what was the first number or benchmark I used, and am I still anchored to it? Three: what evidence would change my mind, and have I genuinely looked for it? Four: if I had not already invested time, money, or effort in this direction, would I still choose it? Five: what is my range of expected outcomes, and am I being honest about the uncertainty?
This check takes less than five minutes and consistently surfaces at least one error or blind spot. It does not need to be comprehensive to be useful. Catching one of the five errors before committing is worth the time of the full exercise.
Don't just read about it — run your actual decision through our AI Decision Assistant.
DecisionsMatter.ai is an AI decision assistant that walks you through a structured 5-step analysis: framing, bias check, pre-mortem, and decision record. Your first analysis is free.
One decision insight a week.
Mental models, frameworks, and decision science — no noise.