Plan research across stakeholders

A service research scope cannot reduce “the user” to the holder of a digital account. A decision may affect the applicant, someone assisting them, a front-counter worker, verification staff and the operational policy owner. Each sees different evidence and faces different incentives. Map each role to the decision it touches, evidence it can supply and the team’s access to it. An operations manager interview is not a substitute for observing an applicant’s attempt. The manager may describe an internal rule accurately without knowing how a visitor interprets its wording. Equally, a visitor may report a real obstacle without seeing a records-system constraint. Balance is not an equal count of interviews by role; it is a set of perspectives capable of explaining the full service journey.

Start with one decision question and derive rival hypotheses. For a mobile application service, the complaint might be “registration completion is low.” One explanation is that applicants misunderstand identity proof. Another is that they leave when asked for location permission. A third is failure on older devices. Each implies a different trace: document-related support contacts, abandonment at permission, or errors grouped by device. Specify the expected signal, source and inference limit before collecting data. If records are insufficient, do not reconstruct them only from staff recollection. Design a short observation or qualitative test that locates the failing step. In interview analysis, separate what the participant reported from the researcher’s interpretation and retain a case that contradicts the prevailing account.

Recruitment follows the risk in the decision. A convenience sample of available users can uncover a problem pattern but excludes people who never reached the service, needed assistance or abandoned it. Include successful, struggling and non-starting cases, assisted channels and meaningful differences in digital ability. A small sample cannot yield a population rate. Give each interview a purpose: understand the real task, the confusing step and the device or channel context. Avoid leading prompts such as “Wasn’t that button obvious?” Observe action before explaining it. Minimize personal data; an anonymous participant code is often enough for an analysis log, and sensitive details should not be collected merely because they are accessible.

An advisory research plan is a decision document, not just an interview calendar. It identifies the schedule, instrument, consent review, what would support or challenge each explanation, and how findings reach the decision owner. If counter staff fear being blamed, explain that the unit of analysis is the journey, handoffs and constraints, not the performance of an individual who did not consent to appraisal. If the technology team provides extensive analytics, ask for event definitions and coverage periods so measurement gaps are not mistaken for user abandonment. End with explicit risks: groups the sample may miss, data that cannot be linked and a decision that will be deferred if uncertainty persists. These become limits on the later recommendation, not defects to erase from a report.