Product research · Data minimization · Workplace products
How can product teams learn from usage while protecting private conversations?
Choose the least intrusive evidence that can answer the research question, set purpose-specific access, and review any move from usage events to message content separately.
Choose evidence for a specific product question
Write the question before collecting data. To investigate workflow completion, you might need an approved completion event and a time window. To understand why a step is difficult, an invited research session may explain more than a message archive.
Separate event metadata, user-supplied feedback, research recordings and conversation content in the plan. Give each field a purpose, recipients and retention rule. Descriptions, search terms and document titles can also contain sensitive information, so a field called metadata still needs review.
GOV.UK’s user-research data guidance recommends collecting what the research question needs and limiting access to people who need it. Apply that principle when designing instrumentation: test whether a less detailed event, a broader category or a voluntary question would support the decision.
Make access and participation specific to the intended use
Administrative access to a product should not settle whether researchers may read its conversations. Establish the authority and purpose for the proposed content access before it happens. If the question can be answered using approved events or newly collected research, keep those routes available.
For UK workplaces, the ICO’s monitoring guidance highlights purpose, proportionality and the difficulty of relying on employee consent where power is unequal. The guidance is under review following legal changes. Have appropriate privacy and legal specialists assess the actual workplace and jurisdiction; an invitation or checkbox alone cannot settle the requirements.
For voluntary research, explain what will be collected, who will analyze it and whether a third-party service will receive it. GOV.UK’s guidance treats later uses and additional recipients as matters that must fit the participation agreement. Keep those limits with the material.
Review what the findings reveal and how they will be used
Keep product-improvement findings separate from decisions about an individual’s performance. The ICO emphasizes defining monitoring purposes and limiting collection to what is needed. If a proposed audience or use changes, review the new purpose and exposure before sharing the output.
Use synthetic examples when explaining a product problem does not require a real conversation. Removing names from actual material can leave recognizable circumstances, unusual language or small groups. Have specialists assess disclosure risk for the actual output instead of treating aggregation or paraphrasing as an anonymity guarantee.
Prefer a finding that describes the obstacle, the situations in which it arose and the next test. Limit supporting detail to what that decision needs. Assign someone to handle access changes, retention and corrections, and verify that exports and research tools follow the same approved boundaries.