What was submitted
What keeps happening. “I make a decision, I’m satisfied with it, and then I reopen it. I re-run the whole comparison. I usually land in the same place. Then I do it again a week later.”
The example. P. is 29, head of operations at a small e-commerce company. Six weeks ago she decided to hire a second logistics coordinator rather than outsource fulfilment to a third-party provider. The CEO approved it. The role was posted. Since then she has reopened the decision four times:
- After reading an article on third-party logistics pricing.
- After a colleague said, in passing, “I’d have just outsourced it.”
- After one weak candidate interview.
- On a quiet Sunday, with no prompt she can identify.
Each time she rebuilt the comparison from scratch, reached the same conclusion, and lost roughly a day. The same thing happens with smaller decisions (a vendor choice, a tooling change). One counter-example: a year ago she reopened a software vendor decision, reversed it, and is still glad she did.
The cost. “Four days on a decision that was already made. And I’m slower to commit to the next one because I know I’ll reopen it anyway.”
What she thinks is going on, and what she has tried. “I think I don’t trust myself. I’ve tried just forcing myself not to think about it. That lasts a couple of days.”
Sorting what we have
Reported: everything above, taken at face value. Established from the submission: the decision was made, approved and acted on, and she reports no underlying fact having changed.
Not known: whether a record of the original reasoning exists (rebuilding the comparison is compatible with no record, and with a record she does not consult); what the decision assumed and what it left out; what the article said; whether the reopens fell near steps that would make the hire real, such as an offer; what was different about the vendor decision she reversed.
Plausible explanations
1. At least one reopen carried something worth checking. She has reversed correctly before. Of the four triggers: the article may have contained pricing information that bears on the cost comparison. The interview is weak evidence about the candidate pool, one candidate from one posting, but it bears on how long the hire will take. The colleague’s remark is an opinion; it can prompt a question (“does he know something about our outsourcing options that I don’t?”) without justifying anything on its own. The Sunday reopen had no content. That none of the four changed her conclusion does not mean none should have prompted a check.
2. There is nothing to check a doubt against. Whether or not a record exists, she is not using one. A decision she can consult can be re-examined in minutes: read the reasons, check the trigger against them, close it or reopen it on purpose. A decision she can only re-derive costs a day every time. This reading fits the cost pattern directly. It explains why doubting is expensive, not why she doubts.
3. Reopening keeps the decision unfinished. The headcount is approved and the role is posted, so the commitment has been made in fact. What may still be open is her relation to it: while the comparison is running, she has not yet treated the decision as hers to stand behind, and the point where that becomes unavoidable is an offer. This would predict reopens intensifying near an offer, and it fits the Sunday reopen. The submission does not say when the reopens fell relative to the hiring steps, so it is untested.
4. Other people’s input is the dominant trigger. Two of four reopens followed something someone else said or wrote. If that is the pattern, the mechanism is specific and so is the fix. The Sunday reopen argues against this being the whole story.
The repeated reviews have not changed her conclusion. Consistency is not correctness: the same comparison re-run from the same inputs will reproduce the same error if there is one. What we still need to establish is whether those reviews added evidence, tested a different explanation, or simply repeated the original comparison. It says nothing between 2 and 3, both of which predict a day-long re-run; timing is what separates them.
What would distinguish them
Does a record exist? If not, 2 is confirmed as part of the picture. If it does, the question is why she re-runs rather than reads. Either the record lacks what she needs (the assumptions and what was left out, not just the verdict), or the re-run is doing something the record cannot, which points toward 3.
For each trigger, what did it touch? Not only the assumptions she can reconstruct, but whether the comparison left something out, whether a priority has shifted, whether an option exists now that did not then. The article and the interview go through this check. The remark yields one question to ask the colleague. The Sunday yields nothing to check, and that is its answer.
Timing. Did any reopen fall the day before an offer, a signature, a next-stage interview? Clustering there gives 3 weight.
The vendor reversal. What was different? If she learned something after deciding that changed a reason, that is the model for a legitimate reopen. If it was a feeling that happened to be right, her feeling-driven reopens are not always noise and should be turned into questions rather than dismissed.
How the answers change the recommendation
If a trigger touched a reason: reopen that reason deliberately, with a threshold written down. For the interview: “If no candidate meets the bar by week eight from posting, get a fresh outsourcing quote.” Eight weeks because the current coordinator has said overtime cannot continue past about three months and the posting is four weeks old; provisional, and it should move if candidate flow changes.
If nothing touched a reason: the decision stands, and the work is to make standing cheap: a record with the reasons, the assumptions, what was left out, and the revisit conditions. Reopens go through the record.
If reopens cluster near an offer: the record helps, but the pressure point is the offer itself. A proportionate step is to write the hiring bar now, before a candidate appears: must-haves, what the assessment consists of (references, a practical task), what counts as good enough. Then the question at offer time is “does this person meet the bar,” not “was the decision right.” Nothing irreversible is done early and the assessment stays intact.
If other people’s input is the driver: an opinion becomes a question: what does this person know that I don’t? Ask it. If the answer is nothing, close.
A proportionate next step
- Write the record.
- Run the four triggers through it.
- Write the hiring bar.
- Set the review point: week eight from posting, or the first candidate who meets the bar, whichever comes first. Until then, any urge to reopen goes to the record.
Stopping rule: the comparison is not rebuilt. A trigger is checked against the record; if it touches a reason, a factor that was left out, a changed priority, a new option or an error in the original reasoning, that point is examined and a threshold set; if not, the reopen is closed. Opinions and moods can prompt a written question, not a re-run.
Conditions that would trigger a deliberate reassessment, all provisional: no candidate meets the bar by week eight; an outsourcing quote close enough to in-house cost that the original price reason is mostly gone (she sets the number); a volume forecast beyond what two coordinators can handle; or a factor the original comparison left out that turns out to matter. Meeting one means re-evaluating on purpose, not reversing automatically.
Confidence and limits
Explanation 2 fits the cost pattern, but rebuilding does not prove no record exists; that is the first thing to check. Explanation 1 is partly supported by the article and the interview, thin as they are. Explanation 3 is plausible and untested; 4 is weakened by the Sunday reopen. The recommendation holds under 1 and 2 and does most of the work under 3 and 4, which is why it is proposed without settling the question.
“I don’t trust myself” may be true, but it is broad enough to explain anything. The narrower picture is that she reaches the same conclusion each time and lacks a way to hold it without re-deriving it. The submission does not tell us whether the original comparison has been independently challenged, which is worth finding out. If a record exists and she re-runs anyway, that picture is wrong and 3 needs more weight.
The recommendation to write revisit conditions in advance as if-then statements draws on the finding that plans specifying when and how to act improve follow-through compared with intentions alone: Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69–119. That research concerns goal pursuit generally; its application to decision records is an extension, not a direct finding.