Why the three layers matter
Research discussions often move quickly from what someone said to what a team should change. A participant struggles to find delivery information, for example, and the team proposes redesigning the entire checkout. That might eventually be sensible, but the observation alone does not establish the cause of the difficulty or the best remedy. Separating observations, interpretations, and decisions makes those steps visible.
An observation is something recorded in the study, such as a participant opening several pages before finding delivery details. An interpretation is an explanation of what that behavior might mean, such as the possibility that delivery information is hard to locate. A decision is a proposed action, such as testing a clearer delivery link. Each layer serves a different purpose; treating them as interchangeable can give a tentative idea more certainty than it deserves.
Record observations with enough context
Start with a short description of what happened, using language close to what the study captured. Note the task or question, the participant’s circumstances if relevant, and any conditions that may have shaped the response. In a survey, an observation might be the pattern of answers to a particular question. In an interview, it might be a participant describing a recent purchase or encountering a difficulty during a task. Avoid adding a motive to the observation unless the participant actually stated one.
Context matters because the same behavior can have different meanings. Someone who leaves a product page may be confused, may have found the answer they needed, or may simply be interrupted. Notes do not have to include every detail, but another reader should be able to tell what was seen and what was not. Where privacy permits, retain a path back to the relevant anonymized note or response so the team can check its interpretation later.
Treat interpretations as explanations to examine
Once observations are documented, write down plausible explanations separately. If several participants overlook delivery details, one interpretation is that the information is difficult to spot. Another is that participants expect it in a different place. A third is that the study task encouraged them to look elsewhere. Listing alternatives does not make the findings useless; it helps the team choose what to investigate or change next.
Ask what evidence supports each explanation and what would weaken it. Does the pattern appear across different tasks or only one? Did participants describe the same difficulty in their own words? Could the study setting have influenced their behavior? The aim is not to eliminate every uncertainty. It is to identify which explanations are well supported, which remain plausible, and which depend mainly on an assumption.
Make decisions at the right scale
A decision should state what the team will do and why the available evidence warrants that level of action. A recurring, clearly observed problem may justify a targeted change. A single ambiguous incident may be better handled with a small test or an additional research question. The distinction is not a rigid rule about sample size; it is a prompt to match the scope and reversibility of an action to the strength of the evidence.
For the delivery-information example, the team might decide to test a more prominent link before reorganizing checkout. The decision record can name the observation, the favored interpretation, reasonable alternatives, and the result the team hopes to see. That leaves room to learn from the change. If it does not help, the team can revisit the explanation rather than assuming the original observation was wrong.
Use a simple format in the final handoff
A three-column working document can keep the layers distinct: observation, interpretation, and proposed decision. Give each row one specific finding rather than combining several issues into a broad theme. Add a brief note on uncertainty where it would affect the choice of action. If a decision depends on input outside the study, such as implementation constraints, label that input separately instead of presenting it as a research finding.
Before sharing results, read each row from left to right. Can someone trace the proposed action back to what was observed? Does the interpretation go beyond what participants actually did or said? Are alternative explanations still visible when they matter? This check will not turn uncertain research into certainty. It will give the team a clearer account of why it is acting—and what it may need to learn next.
