A practical decision framework

  1. 1. Define the outcome

    Describe the user, task, and outcome that matter now. Capture the workflow from its trigger to completion, including manual workarounds and dependencies. Separate the desired result from a proposed implementation: “replace the application” is a solution, while “operators cannot complete the approval path reliably” is a problem statement. Confirm whether the product direction is still valid. A system can be technically repairable yet no longer serve the intended need. Ask the people who use and support it to review the description, and record disagreements as decisions to resolve rather than blending them into a vague requirement.

  2. 2. Map the critical path

    Trace the important workflow through screens, services, data, permissions, integrations, and operational handoffs. Mark where it fails, what evidence shows the failure, and who can reproduce or verify it. Include neighboring paths when they share a component or data contract. Avoid treating every code smell as a reason to replace the whole system. A recovery map should connect a user-visible consequence to a technical or operational constraint and a next investigation. This creates a comparable unit of analysis for repair and replacement.

  3. 3. Inspect repairability

    Review the code and architecture around the mapped path, along with tests, deployment practices, data quality, and ownership. Look for boundaries where a change can be isolated, whether critical behavior can be tested, and whether the team can safely release and observe a fix. State what is known and what remains a hypothesis. If the team cannot explain a failure, run a focused investigation before estimating a broad intervention. Repairability is about the path to a dependable outcome, not whether the code looks old or unfamiliar.

  4. 4. Compare the options

    For rescue, outline the smallest intervention that restores the path and the risks it leaves. For a rebuild, identify what must be recreated, migrated, integrated, operated, and supported while the new system is introduced. Include coexistence and rollback where relevant. Compare both routes against the same user outcome, data obligations, release needs, team capability, and product direction. A rewrite can remove a constraint but also create migration and continuity work; a repair can preserve useful behavior but leave structural limits. Make these tradeoffs explicit.

  5. 5. Choose evidence that can change the decision

    List the assumptions behind each option and identify evidence that would alter the choice. Examples include reproducing the failure, tracing a data contract, testing a narrow replacement boundary, or verifying whether a required workflow can be supported within the current architecture. Give each investigation an owner and a decision date tied to the work, not an arbitrary reporting ritual. If new evidence does not change the route, stop investigating and proceed. If it does, update the comparison and explain why.

  6. 6. Plan a safe sequence

    Sequence the work around continuity. A rescue may begin with observability, a targeted fix, regression coverage, and a controlled release. A replacement may begin with a boundary, migration strategy, data reconciliation, and a way to run or revert safely. Define ownership for production behavior and support before changing a critical path. Break work into checkpoints where the team can confirm the old and new behavior. Keep customer communication and internal operating procedures aligned with what is actually available.

  7. 7. Review the result

    Agree in advance what evidence will show that the chosen intervention addressed the problem. This could be a reproducible test, a completed workflow, verified data movement, or an agreed operational handoff. Review unresolved risks and the next product decision separately from the immediate fix. Do not describe a successful technical release as proof of broader product value unless that value was observed. The recovery map can then become a concise record of the decision, assumptions, owner, and follow-up.

Before committing, ask what the team would need to preserve: user habits, records, integrations, operational knowledge, or regulatory controls. These are not arguments for keeping the current implementation forever; they are migration requirements that belong in either plan. Also name the costs of delay and the risk of changing a live workflow. A decision record should state the chosen route, alternatives considered, evidence reviewed, unknowns, and the person who will revisit the choice if a stated condition changes.

Choose the next action

Use the table to make the current decision explicit. The categories are prompts for discussion; actual evidence and context should determine the path.

Signal Likely focus Next move
Core behavior is understood and constraint is bounded Focused intervention Assign an owner, define verification, and address the specific blocker.
Direction or user need remains uncertain Product learning Test the assumption that would change scope before committing to broad delivery.
Multiple paths depend on unclear foundations Technical or operating decision Map dependencies, compare options, and agree on sequencing and ownership.

For related work, see the Product Rescue and Product Rescue Starts With a Recovery Map. Use these pages as context; the framework here is general guidance and should be adapted to the team’s product, obligations, and constraints.

Questions teams ask

What are signs a product should be rebuilt instead of rescued?

A rebuild may be appropriate when core architecture or data constraints prevent required product behavior, and a bounded repair cannot remove that constraint. Confirm the limitation against a critical workflow and account for migration, coexistence, support, and release responsibilities before choosing that path.

Can a team replace one part of a product instead of rebuilding everything?

Yes. If the system has a boundary around the fragile capability, the team can compare a targeted replacement with a full rebuild. Check the data contract, integration behavior, and ownership needed to operate both sides during the transition.

What should a rebuild decision include?

Record the user outcome, evidence about the current constraint, alternatives considered, migration and continuity needs, and the owner of the decision. State what new evidence would cause the team to revisit the choice.