Modern Business Leadership · Small experiments
Before the Trial Begins, Write the Rollback Note
“We can always change it back” is reassuring until somebody asks where the previous version is and who is allowed to restore it.
Reversible is an unfinished sentence
A team wants to try a new way of scheduling internal work. The change is small, the software is familiar, and the manager says that if it causes problems, they will go back to the old process.
That sounds reasonable. It is also missing most of the information needed to do it.
What exactly will go back? Who will decide? Where is the previous configuration? What happens to requests entered during the trial? How will staff know which process is now in use?
A rollback note answers those questions before the team needs the answers quickly. It turns the word “reversible” into a description of work someone can perform.
The example here is a fictional, low-risk internal scheduling trial. It is not a technical recovery plan for a critical system, an AI approval framework, or advice for changes affecting safety or regulated decisions. The point is to make an ordinary management experiment easier to undo responsibly.
Choose a change small enough to describe
Imagine a six-person creative team that schedules internal review requests through a shared spreadsheet. It wants to try a simple request form that writes to a separate trial queue. For two weeks, the team will use the new route for a defined class of non-urgent review work.
The experiment has a clear boundary. Existing client commitments remain in their current system. Urgent work follows the established process. The trial concerns how this team receives and allocates one kind of internal request.
That boundary makes a rollback note possible. “We are changing project management” would be too broad. “For these requests, use this form during this period” gives the team something specific to examine and restore.
Before the trial starts, record the old process as it actually works. Include the authoritative spreadsheet location, the current request fields, the responsible coordinator, and the ordinary notification method. Do not reconstruct the old process from memory after the experiment becomes inconvenient.
Modern Business Leadership
Name what would cause a return
Success criteria and rollback triggers are related but different. An experiment can fail to produce an improvement without creating a reason to interrupt it immediately. Conversely, a serious process problem may warrant stopping before the planned review.
For this fictional trial, a trigger might be that requests are repeatedly missing from the queue or that the coordinator cannot reliably identify the current assignment. The exact thresholds should be chosen by the team before the trial, based on the actual consequences of the work.
Avoid vague instructions such as “stop if people do not like it.” Dissatisfaction is worth investigating, but it can mean unfamiliarity, an unclear instruction, or a genuine defect. A useful trigger describes an observable condition and who assesses it.
Also identify who can pause intake while the question is reviewed. That person may not be the same person who approves a permanent process change. The note should make the distinction easy to understand.
The previous state needs an address
“Use the old spreadsheet” is only useful if the old spreadsheet still exists, can be accessed by the right people, and represents a known state. A link to a folder containing six similarly named versions leaves too much work for the moment when the team is already confused.
Identify the authoritative starting record and the person responsible for preserving it. For a simple process trial, that may be a dated copy of the instructions and a clearly named reference file. The appropriate preservation method depends on the tools and information involved.
Do not assume that keeping the old file solves the whole problem. New requests created during the trial must still be accounted for. The rollback note should explain how the team will reconcile that work before resuming the earlier process.
This is why restoration is more than a switch. The team is returning to a way of working while carrying forward events that occurred in the meantime.
A one-page note for the fictional trial
The following is a worked management example. It is intentionally small enough for the people doing the work to read before the trial begins.
| Field | Example entry |
|---|---|
| Change under trial | Route non-urgent internal review requests through the trial form for two weeks |
| Previous process | Shared review spreadsheet and the existing coordinator notification |
| Preserve before starting | Dated instructions, current field definitions, and the authoritative starting record |
| Trigger for review | Missing requests or assignment status that the coordinator cannot reliably establish |
| Pause authority | Team coordinator may pause new trial submissions and notify the manager |
| Restore owner | Named coordinator, with a named backup |
| Reconciliation | Account for every trial request, its owner, and its current state before transferring open work |
| Staff message | State which process is active, from what time, and where open requests can be found |
| Completion check | Coordinator and one other team member confirm the active queue and outstanding work |
The real note needs actual names, locations, conditions, and a restoration estimate appropriate to the tools. The table shows the decisions to make; it does not pretend to have made them for a real organization.
Walk through the return before the departure
Ask the restore owner to explain the sequence aloud. Where would they begin? Which information would they need? Who could answer a question about an ambiguous request? What would prevent them from completing the work?
For the fictional team, this walkthrough might reveal that only one person can export the trial queue. That is useful to learn before the person goes on leave. It might also reveal that the new form does not capture a field needed in the old spreadsheet. Now the team can address the mismatch while the trial remains easy to adjust.
A walkthrough is not proof that every failure has been anticipated. It is a practical way to expose missing dependencies in a low-risk change. More consequential systems need the appropriate technical, operational, and professional review.
The question is straightforward: could the named person use this note to restore the process without relying on the manager’s memory?
Modern Business Leadership
Tell people which version of work exists
When a trial stops, staff need a clear message about the active process. “We are going back” may leave them wondering whether to resubmit requests, keep using a bookmarked form, or wait for someone to move their work.
The message should say when the change takes effect, where new requests go, what happens to existing requests, and whom to contact about a discrepancy. Retire or clearly label the trial entry point so people do not continue submitting work into an unattended queue.
For the fictional trial, the coordinator might say that new requests now return to the shared spreadsheet, that open trial requests are being reconciled by the coordinator, and that staff should not submit duplicates. The exact message must match what the team is actually doing.
Communication is part of restoration. A technically restored tool with a team still following two processes has not restored a reliable way of working.
Keep the learning after the rollback
Returning to the old process does not make the experiment useless. Preserve the question tested, what happened, why the team stopped, and what remains uncertain. Avoid rewriting the record as if the outcome had been obvious from the beginning.
The form may have reduced interruptions while exposing an assignment problem. The trial may have been too broad. The team may need a smaller change before trying again. Those are different findings and should lead to different next decisions.
My article on what I examine before recommending a marketing change addresses the earlier question of what action the evidence supports. A rollback note belongs to the practical execution of a bounded trial once the team has chosen to try it.
Likewise, pilot decision design concerns what a successful pilot should authorize. The note here has a smaller job: make the return path usable if the team needs it.
Modern Business Leadership
Confidence can include an exit
A written return plan does not have to make a team timid. It can make a small experiment easier to authorize because the consequences and responsibilities are better understood.
The manager is no longer asking everyone to trust the phrase “we can change it back.” The team can see what has been preserved, who will act, how work will be reconciled, and how the active process will be communicated.
That is a useful kind of confidence. It leaves room for the experiment to teach something unexpected without turning an ordinary process change into an improvised rescue.
Before the trial begins, write down how the team gets home.
Rob Urban
A few useful distinctions
Questions about
the story.
Is this the same as a pilot success checklist?
No. Success criteria concern what the trial should demonstrate. A rollback note explains the conditions, responsibilities, and steps for returning to a previous way of working.
Does every small change need a large recovery plan?
No. The documentation should fit the risk and complexity. A short process note can be enough for a bounded, low-risk internal trial; consequential systems require more appropriate review.
What happens to work created during the trial?
It must be reconciled. The note should identify how open items, ownership, and current states are accounted for before the earlier process resumes.
Who should know the note exists?
The decision owner, restore owner, backup, and people whose work changes. They should know which version of the process is active and where to raise discrepancies.