Pre-registration
Committing publicly to your hypotheses and analysis plan before seeing the data.
What it means
Pre-registration is the practice of publicly specifying a study's hypotheses, design, sample size, and analysis plan in a time-stamped record before the data are collected or examined. Its core purpose is to draw a clear line between confirmatory tests — those genuinely predicted in advance — and exploratory analyses generated after looking at the data, a distinction that ordinary publication erases. The mechanism by which it improves science is constraint: by committing ahead of time to specific outcomes and analyses, researchers remove most of the undisclosed flexibility that enables p-hacking and HARKing, so a result that survives a pre-registered test is far more credible. A stronger variant, the registered report, submits the plan for peer review and in-principle acceptance before results exist, which also neutralizes publication bias because acceptance does not depend on how the results turn out. Pre-registration does not forbid exploratory analysis; it simply requires that exploration be labeled honestly as such, and it has become a central reform to emerge from the replication crisis. It matters because it restores the logic of hypothesis testing and makes the published claim that a finding was 'predicted' actually verifiable.
Examples
Lodging a detailed analysis plan on a public registry before collecting any data; a registered report goes further and is peer-reviewed and accepted before the results are known.
A marketing team writes down, before the A/B test runs, that success means completed checkouts — so they cannot later declare victory on whichever of ten metrics happened to move.
A school trialling a reading scheme states in advance which year group and which test it will use, closing off the option of hunting through classes until one looks improved.
First described in Open-science reforms (2010s).