Decisions and steering

Continue, stop, choose: the biases that distort project trade-offs.

Continue or stop, migrate or keep, ship or wait: steering decisions are made under pressure and with incomplete information. This family gathers the biases that make us keep going with a losing project (costs already committed, escalation), fear change (loss aversion, status quo) or fail to see the incident coming (normalcy, omission).

They often combine: the “sniffer planes” case is a good example of a chain. Countermeasures: stopping criteria set in advance, a pre-mortem, a decision-maker outside the project.

Sunk costsWe keep a project, a contract, or a tool because we've already invested a lot in it, not because what comes next is profitable. Where the bias hits (drifting projects, go / no-go, technical debt), how to spot it and protect against it.The law of the instrumentWhen we master a tool, we see it as the answer to everything. Where it strikes (choice of technology, architecture, method), how to spot it and how to start from the problem rather than the tool.Normalcy biasWe underestimate the possibility of an unusual incident because it has never happened. Where it strikes (backups, recovery plans, security, operations), how to spot it and counter it with real testing.Escalation of commitmentWhen faced with poor results, people reinforce a commitment (money, team, support) instead of revisiting it, to justify the previous decision. Where it strikes (project steering, committees, sponsors), how to spot it and get out.Loss aversionA loss weighs about twice as much as an equivalent gain, and we avoid changing what works well enough. Where it strikes (migrations, rebuilds, tool changes), how to spot it and how to counterbalance it.The status quo biasWe prefer what already exists, the default option, or the current situation, even when another option is better. Where it strikes (architecture evolution, method changes, tool choices), how to spot it and counterbalance it.The omission biasWhen outcomes are equal, harm caused by an action is judged more severely than the same harm caused by inaction; so people prefer to let things happen. Where it strikes (release decisions, fixes, maintenance), how to spot it, and how to defend against it.