“One more tab” feels like diligence, but it is a decision you never made: the next result, or your fatigue, decides when the research ends. A stop rule makes that decision before you search. Write down which decision needs an answer, what finding could change it, and what event will end or redirect this pass. Research then ends on a reason you recorded instead of on exhaustion, deadline panic or a comforting source. Stopping on that reason still does not prove the evidence is sufficient.
The rule belongs to Target, the first step of TRACE in The Verified Research Loop. The book presents the stopping record as a control another person can inspect, not as a method shown to make decisions faster or more accurate. Another tab may still be necessary. The rule makes you say why.
What decision is the next tab serving?
“Research customer-support software” names a topic. “Should we start a limited trial next week, pick another tool, or defer?” names a decision and its alternatives. Chapter 3 asks you to write the decision first and the research question second: what evidence bears on whether this trial can do the job? Inside that question sits at least one proposition, a specific claim that a source or a test can support or weaken. Here: the trial tier exports tickets with their attachments. A tour of feature lists can be interesting without helping this choice.
Before searching, answer what the book calls the counterfactual test: if the answer came out materially different, which option, condition, timing or escalation would change? It is a guide, not a hard rule. A compliance check may be mandatory even when the team expects no change, and a monitoring question may update a later trigger rather than today’s choice. Record that role. Do not invent a decision link that is not there.
State the scope as well. Which plan, product version, region, customer group, date and comparison are in question? An exclusion does not make the excluded risk disappear. If an excluded issue could still overturn the action, give it an owner or keep the action on HOLD. A narrower search must not quietly shrink the decision.
What belongs in the rule before you search?
“Stop after 30 minutes” is a budget. “Find three sources” is another. Neither tells you whether you have read the relevant source body (the document itself, not a snippet or summary of it), resolved a critical claim, or may act. Set a budget, but write a separate condition for stopping or redirecting the work.
Copy this short brief into a note:
| Write before opening results | Fill in |
|---|---|
| Decision and owner | What action is being considered, by whom, against which alternatives? |
| Action-changing finding | Which result would change the choice, its conditions, its timing or its escalation? |
| Proposition and rival | What exact claim are you checking, and what credible result would weaken it? |
| Routes and budget | Which documents, tests or people will you check in this pass, and how much effort will you spend? |
| Stop or redirect | Which observable event ends this pass? What sends you back for more work? |
| Critical gap | What missing answer blocks the action or needs a specialist? Who owns it? |
| Reopen trigger | What changed fact, version or date would make you check again? |
Use the brief while you search, not only in the final memo. When a tempting result appears, hold it against the proposition and the rival. If a source seems to back the option you already prefer, open its body and check its scope before you let it end the search. If the budget runs out, record the unfinished work and make a separate decision about the action.
What are the five honest reasons to stop?
Chapter 3 distinguishes five events that end a search. The book’s rule is that they must not be collapsed into one another.
| Reason | What happened | What the record must keep |
|---|---|---|
| Deadline or resource | The budget ran out or the deadline arrived. | Planned sites not searched and bodies not obtained, plus who will extend, narrow, defer or accept the risk. |
| Coverage | A defined search set whose total is known has been covered. | The set, query, exclusions, failed access and duplicate handling. Covering the set is not covering the world. |
| Saturation observation | New searches keep returning the same bodies or source lines and no new claim that could change the choice. | Each added query, what new body or source line it found, and the blind spots that remain. |
| Decision | The owner judges that one more search is unlikely to change the action enough to justify its cost and delay. | The trade-off, answered through the six questions below. |
| Authority | The question reaches the limit of the researcher’s role. | The open issue, sent to specialist review or kept on HOLD. The legal, clinical, safety or employment decision stays outside the team’s scope. |
An hour passing is a resource stop. Familiar results describe your last few searches, not what is left. A deadline can end research; it cannot decide that the evidence is enough, and none of the five events proves that.
A decision stop is the easiest to misstate. The book calls it a value-of-information judgment, not a fact found in the search. You cannot read it off the number of open tabs, and you should not dress it up as a precise number. An authority stop is different again. Finding the rule or the clinical paper does not give the researcher the authority to apply it. Research can stop while the proposed action stays on HOLD or goes to specialist review.
When does one more tab earn its cost?
Before you open the next tab, put the six decision-stop questions from Chapter 3 to it:
- What unresolved proposition could still change the decision?
- How likely is this search route to resolve it, and why?
- What would it cost in time, money, access or delay?
- How reversible is the current action?
- Who bears the residual risk?
- What event will reopen the record?
For a stack of open tabs, start with the second question. Ask where a likely hit would come from. A new URL may be another account of a body you already have. Corroboration is not a source count explains why five reports of one survey are one evidence line, however many publishers carry it.
The first and third questions weigh what a search could change against what it costs. If a successful search would not change the option, a condition, the timing or the escalation, and no mandatory check remains, record why the remaining doubt does not matter for this bounded decision, who bears the residual risk and what event reopens it. That record is your decision stop. If the next body could close a critical gap, an expired budget does not close it for you. Someone must decide to extend, narrow or defer the work, or to accept the risk in a named, dated record; until then, an unresolved critical claim stays on HOLD.
Reversibility, the fourth question, changes which action you can permit on the evidence you have. It does not make that evidence stronger. A sandbox trial on synthetic data with a written exit condition is a narrower action than a live customer migration or an annual contract. Reversible vs irreversible decisions shows how a “trial” that moves real customer records can close the door while everyone treats it as open. Write the limit of the proposed action on the brief.
Worked example: eighteen tabs for one trial
Hypothetical example. A founder is weighing a two-week sandbox trial of a customer-support tool, run on synthetic tickets. Assume permission for the trial is already cleared. The alternatives are another tool or deferral. Importing real customer records and signing an annual contract are separate decisions. The founder writes one deciding condition: can a trial account export tickets and their attachments in a usable format? If it cannot, this trial does not go ahead.
Before searching, the founder sets a 45-minute budget for the first pass: read the current documentation for the trial plan, run one export with sample tickets and attachments, and log any failed access. The rival answer is that the advertised export drops attachments or is not available on the trial tier. The pass ends when the test gives a clear result or the 45 minutes run out. That duration is an illustrative planning choice, not a validated research time.
Eighteen tabs accumulate anyway. Reviews praise the tool. Several posts repeat its feature list. A comparison page says “export supported” but names no plan or version. None of them answers the trial-tier question. The founder holds each tab against the brief, closes the ones that cannot change the decision, and goes to the current help documentation and the trial account itself. An old result about a paid plan stays in the notes as context; it cannot settle what this trial account does.
The prewritten rule allows three outcomes:
- The sample export works, attachments included. The founder may approve the sandbox trial and nothing more. No customer data or annual contract follows from this check.
- The export fails the stated requirement. The founder picks another tool or defers this trial.
- The result is still unclear when the 45 minutes run out. The record reads “resource stop; export unresolved”, names who runs the next check, and keeps this trial on HOLD. The clock has not answered the question.
The example shows how a rule handles a result, a failure and an unknown without changing the question midstream. It does not show that the method saves time or leads to a better choice. Security, contract terms and a real-data migration still need their own owners before any broader commitment.
What can a recorded stop still miss?
A tidy brief can preserve a bad assumption. A search result can be mistaken for an inspected source body. Several reports can share one origin. And scope can fail in either direction. A run can end on a smaller, easier question than the one it started with; Chapter 11 calls this scope collapse and premature closure. Or the search can keep growing after the branches that matter to the decision already have enough support, which the book calls ungoverned scope, not rigor.
At the end, compare the final record with the current version of the brief. Did the decision and its scope stay the same? Did you read the body and version that bear on the critical claim? Is a material conflict still open? Could an excluded issue reverse the action? If so, go back to the brief, narrow the permitted use, or keep the decision on HOLD.
Keep the research route apart from the decision disposition. When research does not stop, the route is Continue under the current brief or Narrow through a new brief version. The disposition is a separate choice: Proceed, Conditional proceed, Pilot, Defer, Reject, Specialist review, Accepted risk or HOLD. Accepted risk is a dated decision by a named, authorized owner; it never turns an unverified claim into a verified one. What is TRACE? walks through all five steps, and its limit applies here too: the method shows what the evidence supports; it does not prove a conclusion true.
Try this today
Pick one question you are researching for a real decision. Before you open another result, write four lines: the decision and its owner; one finding that would change the action; a budget in time and named source routes; and the exact event that ends or redirects this pass. Add the critical gap that would keep the action on HOLD.
Now go through your open tabs. For each one, finish the sentence: “This could change the decision by showing ___.” Keep the tabs with a credible route to that finding. Close the rest. If a relevant tab holds only a snippet or repeats another report, note that limit instead of treating it as an inspected, independent source.
When your stop event comes, write one last line: which reason stopped the search, what remains unknown, and who owns the next check. You have tested whether you can end a search on a reason you wrote down. Whether the evidence permits action is a separate decision.
Cite this:Stop rule vs one more tab: when to stop researching.Len P. van der Hof. https://lenvanderhof.com/en/blog/stop-rule-vs-one-more-tab/ ·