SistemaHub
← The journal

Prepare · By SistemaHub · 2026-10-28

User Research Plan for a Custom Business System: Roles, Questions, Methods and a Decision Log

A research plan written before development starts gives a custom business system a defensible reason for each requirement. The plan does not need to be long. It needs four working parts: the roles you will research with, the questions you are trying to answer, the methods you will use, and a decision log that records how each finding changed the requirements. This article sets out those four parts and works through one hypothetical interview insight that changes a requirement.

Curved thread passing through a brass ring and ledger book into aligned geometric blocks on a cream surface.

Name the roles you will research with

Begin by identifying every position involved in the process, rather than only the individual who asked for the system. In a service-request workflow, this could encompass the coordinator who records requests, the technician who carries out the task, the supervisor who authorises exceptions, and the client who makes the request. According to UK service guidance, one should determine which people to conduct research with, which may mean every type of service user or only certain groups; during discovery for a new service, the likely users might not yet be clearly known. [1]

Write one line per role: what they do today, what they need to finish the work, and what they currently work around. Keep the list short enough to review in one sitting. If a role is unknown at the start, mark it as a research question rather than guessing.

Turn assumptions into prioritised research questions

A workshop is a practical way to collect the team's assumptions and convert them into questions. The guidance recommends starting with a workshop to capture and prioritise research questions, and being careful to turn unfounded assumptions and opinions into research questions. It adds that knowing the questions you need to answer and the assumptions you need to test helps the team focus on the most important things it needs to know. [1]

Order the questions according to how greatly their answers would influence what gets built. Any question whose response would leave a requirement unchanged can be deferred. Be prepared for the questions to change: the guidance observes that during discovery, research questions tend to be wide-ranging and shift rapidly, then grow more focused in subsequent development stages as attention narrows to specific user groups and service components. [1]

Choose methods the team can join and respond to

Method choice depends on what you want to learn and the phase you are in. The guidance advises choosing activities that provide strong evidence and reliable answers for the least time, effort and cost, and suggests that teams with little research experience prefer simple methods like interviews and usability testing that they can easily understand and join in with. [1]

For interviews, contextual research or usability testing, the guidance indicates you will usually need between 4 and 8 participants per round, and that if you need more participants for clear findings, you should run more rounds rather than one large round so you can adjust after each round. Surveys, A/B testing and benchmarking are described as needing hundreds of participants for clear findings. [1]

Plan the round before booking anything. Each round should answer specific research questions with specific user groups, and the guidance suggests aiming for at least one round every two weeks while staying in sync with the team's high-level roadmap, so research does not outpace the team's ability to respond. [1]

Before a session, run through the test tasks on the day before the research, and if you are visiting participants or using their devices, check firewall restrictions, screen resolutions and browser compatibility. A practice session with a team member is worth considering when the activities are unusual or complicated. [2]

Keep a decision log that links findings to requirement changes

A decision log is a simple table with one row per finding: date, source (which session or participant group), finding, affected requirement, decision, and owner. The point is traceability. When someone later asks why a requirement exists, the log shows which observation produced it and who approved the change.

The guidance recommends agreeing how to feed research findings into the team's agile tools and practices, such as planning and prioritisation meetings and story backlogs, and storing research findings and outputs in the team's collaboration tool or shared folders so everyone can use and share them. A decision log is one way to make that feeding-in visible. [1]

Plan analysis time alongside the sessions. The guidance suggests doing analysis on the same day or as soon as possible afterwards, and gives a working ratio of one hour of analysis for every two hours of research. [2]

Worked example: one interview insight becomes a requirement change

Hypothetical example, not a SistemaHub client story. A coordinator says in an interview that when a technician finishes a job, the completion message sometimes arrives after the coordinator has already closed the day's list, so the job is recorded as open until the next morning. The research question was: how does the coordinator know a job is complete? The finding is that completion depends on a message the coordinator may not see until the next working day.

The original requirement read: the system records job status. The revised requirement reads: the system shows a completion queue that the coordinator reviews at the end of each day, and marks any job with no completion update as awaiting confirmation. The decision log records the session date, the finding, the old requirement, the new requirement, and the coordinator as the person who confirmed the change. No numeric performance target is implied; the queue is a visibility change, not a promise about response times.

Before accepting the change, check whether the current tools could already surface the outstanding update through a configuration or procedure. If a small change closes the gap, prefer it. If the gap remains after that check, the revised requirement is the evidence for discussing a different system.

Involve the team and share what you learn

Research is easier to act on when the team sees it happen. The guidance suggests involving team members in observing sessions, helping with research activities such as note-taking and running workshops, making prototypes and materials, and analysing results, and it suggests everyone should observe at least two hours of research every six weeks. [1]

When sessions take place in a lab, the guidance outlines two supporting positions—an observation room manager and a participant host—and states that these can be filled by anyone who grasps the research objectives and is prepared to assist. It further suggests having a research assistant for visits to participants, tasked with taking notes, managing recordings and asking follow-up questions. [2]

SistemaHub builds custom web applications for Philippine businesses, and describes its process as understanding the real problem before opening a code editor, then designing structure and flow before building in modular pieces. A research plan and decision log fit naturally into that first stage. [5]

A useful next step

Write the roles, prioritised questions, method choices and decision log before development starts. Update the plan as findings arrive, and let the decision log show which observation changed which requirement. Revisit the plan at the start of each development phase rather than treating it as a one-off document.

Discuss your research plan and decision log with SistemaHub.

Sources & further reading

  1. Plan user research for your service - Service Manual - GOV.UK
  2. Plan a round of user research - Service Manual - GOV.UK
  3. Business Plan Custom Domain to 365 Personal/Family - Microsoft Q&A
  4. How to add custom domain in Microsoft 365 Family Plan? - Microsoft Q&A
  5. SistemaHub public project information

Written for SistemaHub with AI assistance and source-based editorial checks.