EN
Three clear cylinders holding puzzle fragments; the last is frosted and conceals its contents.

What to do when evidence is insufficient to decide

Missing evidence, conflicting evidence, and uncertainty are not the same problem. When they are collapsed into a generic category, an organization may manufacture certainty, defer a decision indefinitely, or treat something it has not yet observed as a gap.

PRYSMAP4 min read

Decisions will inevitably have to be made with incomplete information. The question is not how to eliminate all uncertainty, but how to identify its source and respond in proportion to the stakes.

A useful framework begins by separating three states.

Absence: no relevant observation exists yet

The organization has not seen the person perform a representative task, has no work product to review, or relied on a source that had no access to the behavior in question.

Nor does it justify assigning the lowest level. The appropriate response is to seek a targeted observation: a real task, a work sample, a limited simulation, an interview that reconstructs past decisions, or a carefully staged experience.

Before gathering more information, define what evidence would change the decision. Otherwise, opinions will accumulate without strengthening the basis for the claim.

Contradiction: sources or observations over time tell different stories

A high self-assessment and a low manager rating may look like a disagreement over who is right. They may instead reflect different expectations, access to different situations, changes over time, or an ambiguous scale.

The contradiction must be traceable: what each source observed, in which context, under which rule, and during which period. Sometimes one piece of evidence is more relevant. In other cases, both are valid under different conditions.

Adding another assessor is not always the answer. The discrepancy must first be located. If the problem lies in the definition, a third opinion will only amplify the noise.

Uncertainty: evidence exists but supports more than one reasonable interpretation

Several relevant observations may still fall short of the required level of certainty. The sample may be small, the context may change, performance may sit near a threshold, or the link between the evidence and the decision may be indirect.

Uncertainty does not mean the process has failed. A defensible result can acknowledge its own margin. The National Academies note that uncertainty in science-informed decisions can arise from missing evidence, variability, models, or profound lack of knowledge. In talent decisions, these forms of uncertainty are often harder to quantify, but the obligation to state the limits remains useful.

The appropriate response may be a provisional decision, another observation, or a narrower conclusion.

Two variables decide the next step

Once the state has been identified, the next step depends on the cost of error and the value of obtaining new information.

When a decision is reversible and the cost of error is low, it may be reasonable to act on an explicit hypothesis and monitor the outcome. A temporary assignment, a limited project, or supervised practice can reduce uncertainty while generating learning.

When the consequences are significant or difficult to reverse, the evidence threshold should rise. This applies to promotions into critical roles, exclusion from opportunities, succession decisions, or conclusions that will follow a person for years.

But requesting more evidence also has a cost. It can delay an opportunity, increase the burden, or create unnecessary surveillance. Gathering more information is justified only when it has a realistic chance of changing the decision or reducing a material risk.

Mountain landscape jigsaw puzzle with one central piece missing and several loose pieces.

Documenting the limits is part of the decision

An incomplete conclusion may be better than a fictitious certainty if it is clear:

  • which question it is intended to answer;
  • what evidence is available;
  • what is missing or contradicted;
  • which interpretation is being adopted, and with what degree of certainty;
  • which provisional decision is being made;
  • which signal would require a review.

This record improves defensibility and prevents a hypothesis from turning, over time, into a fact attributed to the person. It also distinguishes a confirmed gap from a need for further observation.

Evidence-based practice does not require waiting for a perfect test. It requires using the best available evidence explicitly and critically, considering multiple sources, and preserving the limits of the conclusions they support.

When evidence is missing, the organization can still make a decision. What it should not do is hide the uncertainty inside a score.