This checklist belongs to the failed automation review guide for workflow automation agencies. It invents no deadlines. Each stage is tied to the client's own event, and the retention figures quoted are documented platform defaults that the client may have changed.
Countdown to the rebuild
A failed automation review has its own rhythm: diagnose, report, rebuild, clean up. The repair history is most at risk at the end, when cleanup feels like good hygiene.
| Stage | What happens to the evidence | Who sets the date |
|---|---|---|
| Findings delivered | Failed runs and error queues are still in place; the repairs are scattered across business systems | Agency and client sponsor |
| Rebuild live | New or edited workflows replace old ones; some platforms keep history with the old workflow | Agency and operations lead |
| Old workflows cleared | Deleted workflows, emptied error queues and aged-out run history take the platform half with them | Client admin |
Retention keeps running underneath all three. Zapier's help says it can only promise up to 60 days of run data in Zap history. Microsoft documents that Power Automate stores flow run data for 28 days by default. Those clocks do not wait for the project plan.
Stage 1: before the review findings go to the client
Everything is still in place. The aim at this stage is to describe where the two halves of each incident live before anyone changes anything.
| Confirm | Who owns it | If skipped |
|---|---|---|
| Which platforms are in scope and what run history retention each is set to | Client admin for each platform | The 28-day or 60-day window can close before anyone asks |
| Whether incomplete executions are stored at all, where the platform makes it optional | Scenario owner | In Make, incomplete executions are stored only when the scenario setting is enabled, so some failures were never kept |
| Where human repairs are recorded: ERP or CRM change history, helpdesk tickets, email, or only as replays | Operations, finance and sales operations leads | The repair half stays invisible and looks unrelated to any run |
| What key, if any, ties a repair to a run: run ID, order number, ticket number | Same leads | Pairs cannot be matched later, even when both halves survive |
The repair half is usually outside the automation platform entirely. An accounts receivable clerk fixing a tax code in the ERP or a controller approving an exception by email leaves a record in their own system, not in the run log.
Stage 2: before rebuilt workflows go live
Rebuilding is where history can split from the workflow it belongs to. The questions here are about what the new version inherits.
| Confirm | Who owns it | If skipped |
|---|---|---|
| Whether rebuilt workflows are edits of the old ones or new workflows that replace them | Agency build lead | A replaced workflow's history may be deleted with it at cleanup |
| How failed runs were retried during the review: replayed unchanged, or retried after a change | Agency build lead and client admin | The difference between "pressed replay" and "fixed the cause" is lost |
| Whether replay is still possible for older failures | Client admin | Zapier notes that steps older than two months will not replay, so older failures can only be described, not retried |
| Whether the business-system repair records have their own retention rule | Finance or IT | The repair half may age out on a different clock from the run half |
n8n shows why the second row matters: it can retry a failed execution with the original workflow or with the currently saved one, so a retry after a change leaves part of the fix in the workflow's own history. Ask how the client's team retried, and note it.
Stage 3: before old workflows, error queues and run history are cleared
This is the real deadline, and it is the client's. Cleanup is often scheduled as the final task of the project.
| Confirm | Who owns it | If skipped |
|---|---|---|
| The date old workflows will be deleted or archived | Client admin | The platform half disappears on a day nobody chose for this reason |
| Whether error queues will be emptied, and whether anyone has noted counts and periods first | Client admin | The best evidence of what went wrong is removed for tidiness |
| Whether the owner has heard about the licensing option and who would decide | Partner, with the owner | The question is never asked |
| The owner's recorded choice: introduce, revisit later, or stop | Partner, with the owner | The decision is made by default when cleanup finishes |
Skipped items and what they cost
Illustrative. A 270-person industrial supplies distributor ran a review of its order-to-cash automations across two platforms.
| Item | What happened | Effect on run-to-repair pairs |
|---|---|---|
| Retention settings noted at stage 1 | Done; one platform at its default window, one extended by the client | Recent pairs intact on both platforms |
| Repair key identified | Done; order numbers appear in ERP comments for about half of repairs | Roughly half the pairs can be matched |
| Rebuild approach | Skipped; three workflows were replaced rather than edited | Their history went with the old workflows at cleanup |
| Owner decision | Recorded before cleanup: revisit after the next quarter | The choice stayed with the owner |
A partial result is normal. The checklist turned an accident into a decision. The guide on assessing human review records and correction workflows explains why matched error, correction and outcome sequences matter, and why AI agents fail covers the same pattern from the agent side.
Fit and authority check
The client company usually owns its business-system records; the agency only holds delegated access to the automation platform. Customer contracts and privacy rules may limit reuse, which the company's counsel decides. The person who can authorize a licensing discussion is the CEO, who in an owner-run company is usually the owner.
Check fit before any introduction: 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license the records, and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English.
Stay at metadata level: platforms, periods, rough counts, where repairs live, who owns them. Do not export run logs or Zap history to show SourceX what exists. What stays unknown until SourceX reviews it: demand for these records, the rights review result, de-identification needs and any offer. Sign the Channel Partner Agreement before submitting; attribution goes to the first valid introduction within the 90-day window, credited when the company submits a verified application. Illustrative only: a USD 100,000 Actual Supplier Payout to the company would mean a USD 55,000 reward under the rewards page conditions. When the owner opts in, the failed automation review introduction email is the next document. For accounting automations specifically, see MCP journal entry automation, and for the platform contrast, the MCP vs Zapier explainer.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If stage 3 is still ahead, the run-to-repair links hold, and the owner wants to talk before cleanup, register as a partner, confirm the company on who qualifies, and read the program terms first.