The value of a datum – how little data do we need for a quantitative risk analysis?

Aim Conservation managers are typically faced with limited resources, time and information. The philosophy underlying risk assessment should be robust to these limitations. While there is a broad support for the concept of risk assessments, there is a tendency to rely on expert opinion and exclude formal data analysis, possibly because available information is often scarce. When data analyses are conducted, often much simplified models are advocated, even though this means excluding processes believed by experts to be important. In this manuscript, we ask: should statistical analyses be conducted and decisions modified based on a single datum? How many data points are needed before predictions are meaningful? Given limited data, how complex should models be? Location World‐wide. Methods We use simulation approaches with known ‘true’ values to assess which inferences are possible, given different amounts of information. We use two metrics of performance: the magnitude of uncertainty (using posterior mean squared error) and bias (using P – P plots). We assess six models of relevance to conservation ecologists. Results We show that the greatest reduction in uncertainty occurred at the smallest sample sizes for models examined, and much of parameter space could be excluded. Thus, analyses based on even a single datum potentially can be useful. Further, with only a few observations, the predicted distribution of outcomes matched the probabilities of actual occurrences, even for relatively complex state‐space models with multiple sources of stochasticity. Main conclusions We highlight the utility of quantitative analyses even with severely limited data, given existing practices and arguments in the conservation literature. The purpose of our manuscript is in part a philosophical discourse, as modifications are needed to how conservation ecologists are often trained to think about problems and data, and in part a demonstration via simulation analysis.