A stopping rule keeps curious and methodical research useful

A curious and a methodical person can work well together by agreeing before an investigation which decision it must support, what minimum information that decision requires and when new questions move into follow-up work. Curiosity protects against an explanation that is too narrow. A methodical approach makes visible what has already been examined and when the evidence is sufficient to act.

These preferences describe different topics. Curious belongs to ideas and change; methodical to planning and focus. One person can therefore have both preferences. The words do not assign fixed roles such as researcher and planner. Expertise, responsibility and available time determine who performs an analysis or makes a decision.

The distinctive tension appears when a relevant finding opens a new route during the investigation. Ignoring it may lead to a weak decision. Following every route can prevent the original decision from ever arriving. A useful stopping rule distinguishes information that may change the current choice from information that deserves later investigation.

What curious and methodical mean here

A curious person likes to look beyond the first explanation and connect new information with existing knowledge. Inquiring, open and interested are useful related words. This preference sits between conventional and progressive: there is room for another perspective without every discovery having to become radical innovation.

A methodical person likes to work prepared and in a recognisable sequence. Structured, planned and systematic may fit. In the guide, this preference sits between flexible and focused. Methodical work does not mean that a sequence can never change; a change simply helps more when it is clear which step, dependency or outcome it affects.

ConventionalCuriousProgressive

FlexibleMethodicalFocused

The foundation articles on solving problems with a curious person and handing over work to a methodical person discuss the preferences separately. This combination asks a different question: how can an investigation remain educational and finite when real work depends on its outcome?

Curiosity stops the plan from narrowing the question

An investigation often begins with a plausible explanation. A drop in conversion may appear to result from new copy, or a delay from one process step. A curious contribution asks what other explanations could create the same pattern and which assumption remains untested.

That widening is valuable when it changes the choice. If both a technical defect and unclear instructions can cause people to leave, rewriting the copy alone may waste effort. An extra data source or conversation can stop the team from treating a symptom.

The trap is allowing every finding to become a new research goal. An unexpected difference between teams raises questions about training, experience, scheduling and tools. Every question may be interesting, but not all are required for this week's decision. Without a boundary, the analysis grows faster than the insight.

A parking list with a reason helps. It records not only “look at this later”, but also which future decision the question could support. Curiosity is then preserved without forcing the current assignment to carry all of it.

Method protects findings from remaining loose impressions

A methodical contribution makes the route reviewable. What question is being answered, which source is used, what selection was made and which step comes next? This structure prevents two observations from becoming a pattern and helps the team avoid collecting only evidence that supports its first expectation.

A consistent method can also create calm when new information appears. A finding does not need to replace the whole plan immediately. The team first determines whether it disproves an assumption, requires an additional check or falls outside the current decision. The sequence protects the investigation from erratic shifts.

The trap is allowing a careful plan to become more important than the question. A checklist can be complete while the most important explanation never appeared in the original design. “This analysis was not planned” is not a substantive reason to ignore relevant information.

Methodical work stays strong when the plan includes a change rule. New information is put against the same question: can it change the choice, and if so, what is the smallest necessary adjustment? The route remains recognisable without forcing reality to fit the first design.

A research question is not a collection topic

“Investigate our onboarding” is broad enough to fill weeks. A decision question bounds the work: “Which single change will we test with the next group to reduce the largest demonstrable drop-off step?” The second wording does not determine the cause, but it does define what usable output must look like.

The stopping rule is not a ban on learning. It makes visible when learning still belongs to the current assignment. It also prevents time pressure from silently defining what counts as sufficient evidence.

Four parts of a research agreement

For an investigation with practical consequences, colleagues can record four parts in advance.

1. 1
   DECISIONName what will be chosen after the investigation

   Clarify which choice, test or next step the outcome needs to support.

2. 2
   EVIDENCEDefine the minimum information

   Record sources, comparison and threshold needed to assess the options fairly.

3. 3
   CHANGEGive new findings a route

   Decide when a finding changes the current plan and when it moves into separate follow-up work.

4. 4
   STOPClose with uncertainty in view

   Document what is sufficiently clear, what remains unknown and who decides the next step.

Not every investigation needs statistical thresholds or an extensive protocol. For a small, reversible process issue, the agreement may fit on one page. The greater the consequences and the harder recovery becomes, the more precise the sources and stopping criteria should be.

Example: why does security activation remain unfinished?

An organisation sees that some new employees do not activate the security app before their first working day. A new group starts in three weeks. The team must choose which change it will test with that group.

Many possible explanations appear in the first conversation. The invitation may enter a spam filter, the instructions may be too technical, employees may not yet have a company phone, or managers may discuss the step too late. Each explanation calls for a different action.

The team formulates its decision question: which single change will it test with the next group to reduce the largest visible drop-off step? As minimum evidence, the colleagues agree to distinguish four moments for the last two groups: invitation sent, link opened, installation started and activation completed. They also review ten helpdesk questions to recognise problems absent from the data.

The analysis shows that most invitations are opened. The largest drop occurs between opening and starting installation. Helpdesk questions repeatedly show that employees do not know whether they may use a private phone. A curious follow-up question is whether this differs by country, role or manager. That may matter, but today's data cannot support a reliable comparison.

The finding does change the current choice. The team therefore tests an invitation that explains before the installation block which devices are permitted and where to request an alternative. Differences between countries and roles move into a follow-up investigation, linked to a future decision about whether separate onboarding routes are necessary.

After the next group, the team does not only check whether activation rises. It also checks whether more people request another device and whether new helpdesk questions appear. The test remains reversible. The stopping rule has not solved every cause; it has enabled an evidence-based next step without losing the other signals.

Working agreements for research that can end

1. Start with a decision, not a topic. State which choice the investigation must make possible. A broad theme may remain as context.
2. Discuss evidence in advance. Record the source, comparison and minimum certainty needed. Change that agreement only for an explicit reason.
3. Separate plan changes from follow-up questions. A finding that can alter the current choice receives attention now. Other valuable questions receive an owner and a future decision context.
4. Document uncertainty at closure. A conclusion does not need to pretend everything is known. Record what remains unknown and why action is still appropriate.
5. Review the stopping rule after the action. Ask whether the investigation ended too early, continued unnecessarily or produced exactly enough information. Use that learning next time.

These agreements do not prescribe a division of work. A curious person can maintain the research log and a methodical person can introduce an unexpected explanation. Quality comes from connecting both widening and ordering to the same decision.

Research does not need to be complete to be useful

Curious and methodical collaboration becomes stronger when “learning more” and “following the plan” are not treated as competing goals. The shared measure is whether the investigation supplies enough relevant information for a responsible next step.

Discuss observable behaviour: which new question changed the choice, which step produced no information and what evidence was missing at closure? Judgements such as “keeps searching” or “only follows the process” reveal less about what the assignment requires.

The Personal User Guide comparison can show how preferences around ideas sit alongside preferences around planning. Personal User Guide for work helps turn that into a concrete research agreement. A useful opening sentence is: which decision must this investigation enable, and what is the minimum we need to know?