Treat disagreement as a question, not a vote
Imagine a retailer asking why shoppers leave before checkout. Survey respondents often mention price. In interviews, shoppers talk at length about confusing delivery options. A checkout report shows many exits near the account-creation step. It is tempting to pick the largest number, the most vivid story, or the chart closest to the purchase. Each choice would overlook something.
Triangulation means examining how different sources relate to the same decision. It does not mean averaging their answers or requiring them to agree. A survey can describe what respondents say matters; an interview can reveal how they interpret a question or experience; behavior data can show where actions occur. None necessarily establishes why a particular shopper left. Start by writing the decision plainly, such as “Which checkout issue should we investigate or change first?” Then ask what each source contributes to that decision.
Check whether the sources describe the same people and moment
Before comparing findings, put each source on a simple evidence sheet. Record who was included, when the information was collected, the question or behavior measured, and any important exclusions. A survey of recent purchasers may say little about people who abandoned checkout. Interviews with frequent customers may surface difficulties that first-time visitors would describe differently. An exit report may include people who were only checking delivery costs.
Look closely at the wording behind apparent contradictions. “Price is important when choosing a retailer” is not the same claim as “Price caused me to leave this checkout.” Likewise, an exit near account creation shows a point in a journey, not necessarily the obstacle that prompted the exit. The shopper might have stopped earlier, become distracted, or gone elsewhere to compare options.
Timing and definitions matter too. If a survey asks about a typical purchase while the behavior report covers one particular checkout flow, the sources answer related but distinct questions. Mark those differences before deciding that the findings conflict. Sometimes a disagreement disappears once the population, task, or measure is aligned.
Use the gap to form testable explanations
When findings still seem to clash, list several explanations rather than defending a preferred source. In the checkout example, price could be a broad concern while delivery choices create difficulty at the point of purchase. Account creation might be the final inconvenience after those concerns have built up. Interviewees might also find delivery easier to discuss in detail than price, while survey wording might make price the most obvious answer.
For each explanation, note what you would expect to see if it were true. If delivery confusion is central, shoppers may hesitate or seek help when delivery choices appear. If account creation is the main obstacle, difficulty may persist even when delivery information is made clearer. This turns a disagreement into a short list of questions that further observation, interviews, or a focused usability exercise could address.
Keep contrary evidence visible. A neat story that accommodates every finding can still be wrong, especially if it cannot be challenged. Ask what observation would make you reconsider it. This habit helps distinguish a useful working explanation from an explanation that merely sounds plausible.
Judge evidence by fit, not by a fixed ranking
No source is automatically best. A well-designed survey may be useful for comparing stated concerns across groups, but its value depends on whom it reached and what it asked. Interviews can expose language and context that preset answer choices miss, but a few participants cannot establish how common a view is. Behavioral records may capture actions at scale, yet their labels, missing events, and lack of context can limit interpretation.
Match the evidence to the claim you need to make. If the decision concerns how to explain delivery options, observed confusion and customers’ own descriptions may be particularly useful. If it concerns how widely a concern is shared across audience segments, a suitable survey may help more. State conclusions at the level the evidence supports: “Delivery choices appear worth investigating” is different from “Delivery choices are the main cause of abandonment.”
It can help to label each finding as strong, tentative, or not applicable to the decision, with a brief reason. The purpose is not to create a precise score. It is to make assumptions visible so another person can question them.
Make a proportionate decision and leave a trail
Research rarely removes every uncertainty. Decide what can be done responsibly with the evidence available. A low-cost clarification of delivery wording may be reasonable to try while investigating account creation. A major checkout redesign deserves closer scrutiny if its rationale depends on a single ambiguous signal. Define in advance what change you expect to observe and what result would prompt you to revisit the decision.
In a short decision note, include the question, the sources used, the main points of agreement and disagreement, the leading explanations, and the remaining unknowns. Separate observations from interpretations: “Exits occur near account creation” is an observation; “Account creation drives exits” is an interpretation. Record why the proposed next step fits the evidence and its limitations.
This approach makes disagreement productive. Instead of forcing a confident answer from imperfect sources, the team gains a clearer question, a more defensible action, and a way to learn from what happens next.
