Are failed runs linked to the authorized actions that repaired them?
In most clients the honest answer is "the run is logged, the repair is somewhere else". The automation platform records the failure. The fix usually happens in another system: an accounts receivable clerk adds a missing tax code in the ERP, a sales operations lead merges a duplicate CRM contact, a manager approves an exception in a ticket. The link back to the failed run, if it exists, is a run ID pasted into a ticket, a Slack thread or nobody's memory.
Why it matters outside the project: a linked pair shows an error, the decision a qualified person took and whether the run then succeeded. That error, correction and outcome sequence is the kind of operational record AI developers describe needing, and it is rarely public. The guide to assessing human review records and correction workflows covers the same pattern in review queues, and integration error-handling records covers it on the systems side.
A worked run-to-repair ledger
Illustrative: a 230-person B2B equipment rental company whose order-to-invoice automations connect a CRM, an ERP and a shared mailbox. These are metadata a partner can note in a review, not records to collect.
| Failed run (platform log) | Error recorded | Human repair | Where the repair is recorded | Does the link survive? |
|---|---|---|---|---|
| Invoice creation step | Missing tax code on customer | AR clerk adds tax code, replays run | ERP customer change log, run replay | Yes: run ID noted in the ERP comment |
| Contact sync step | Duplicate contact match | Sales ops merges records, edits filter | CRM merge history, edited workflow | Yes: ticket links run and merge |
| Credit check step | Credit limit exceeded | Controller approves exception by email | Shared mailbox thread | Partly: email names the order, not the run |
| Delivery note step | Timeout from carrier system | Nobody; run left as errored | Not recorded | No: the run expires with no repair |
The fourth row is common and worth saying aloud. A run with no recorded repair is not a demonstration of anything, and it should not be described as one.
How long the platform half lasts
The run half of each pair is short-lived by design, which is why the link usually has to live in the business system. Zapier's help says it can only promise up to 60 days of run data in Zap history and suggests exporting history regularly for longer records; its replay help adds that steps older than two months will not replay. Microsoft documents that Power Automate stores flow run data for 28 days by default. Make stores incomplete executions only when the scenario setting is switched on, and notes that errors needing module or blueprint changes are resolved manually. n8n 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 version history. Check each client's own settings rather than assuming the defaults.
Why are human repair demonstrations easy to overlook?
Because a failed automation review is scoped to make automations stop failing. The agency reads error logs, fixes mappings, adds error handling and retries, and the success measure is a quieter error queue. Nobody is asked to preserve the record of how people fixed things in the meantime.
The people who did the fixing are also usually outside the project. They sit in finance, sales operations or customer service, and their repairs look like ordinary work in their own systems: a changed customer field, a merged contact, an approval email. Without a run ID or order number tying the repair back, the two halves look unrelated once the run history ages out.
Finally, the review itself can erase the evidence. Rebuilding or deleting a workflow can take its history with it, and cleaning up an error queue removes the failed runs that show what went wrong. Asking the question before cleanup starts costs nothing and keeps the option open for the owner.
Four metadata questions for the review
Stay at metadata level: what exists, how much, how long, which system and who owns it.
- By what key? Are failed runs linked to the authorized actions that repaired them, and by what key: run ID, order number, ticket number?
- Where are repairs recorded? In the ERP or CRM change history, a helpdesk, email, or only in the automation platform as a replay?
- How many, and how far back? Roughly how many failed runs per month, across which platforms, and over how many years of history in the business systems?
- Who owns and authorizes? Who owns those business systems, and who could authorize a discussion about licensing their records?
Record the answers in the ledger format above, with ranges rather than exact counts. Do not export run logs or Zap history to show what exists; SourceX handles any data review directly with the company if it proceeds. Platform choice is a detail, not a qualifier. The MCP vs Zapier explainer is a useful reminder that trigger-and-action tools log predictable runs; the repair history lives in the systems around them. The portfolio review handoff checklist shows a similar metadata-only handoff in a different setting.
How do you disclose the reward and get the owner's permission?
Ask before you mention SourceX, and disclose in the same breath. For example: "Those linked repairs may have value outside the company. May I mention one option? I should tell you I am a referral partner and may receive a reward if it goes ahead."
If the owner agrees, explain that SourceX licenses approved business data directly from companies for AI training and evaluation and pays the company, and that any reward to you is funded separately by SourceX and never deducted from the company's payout. If the owner declines, the review carries on unchanged.
State the reward exactly. The default is the greater of $50,000 plus 5% and 15% of the Actual Supplier Payout, with no cap across all agreements and renewals. Illustrative only: a $400,000 payout would mean a $70,000 reward. A reward becomes payable after SourceX has paid the referred company under its signed agreement and the payout conditions in the program terms are met, including finance approval and your tax and payment details.](/rewards) explains. If a client tried selling data before and it went nowhere, why past data monetization failed is worth sending.
Who owns automation and repair records, and who can authorize a licensing discussion?
The client company usually owns its business-system records, while the agency only has delegated access to the automation platform. Customer contracts and privacy obligations can limit reuse, and that is a question for the company's counsel, not the agency. The person who can open a licensing discussion is typically the owner, CEO or COO.
| Question | Usually answered by | What stays unknown until SourceX reviews it |
|---|---|---|
| Which repair records exist and for how long | Operations, finance or IT lead | Whether they match current dataset demand |
| Whether contracts or privacy rules restrict reuse | The company's counsel | The outcome of SourceX's rights review |
| Who can authorize a discussion | Owner or executive team | Any offer, its terms and timing |
| Company size and history | Owner or HR | Final eligibility confirmation |
Eligibility is on who qualifies: 50+ full-time employees at peak (contractors excluded), documented operating history, rights to license 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.
You sign the Channel Partner Agreement before submitting an introduction. Attribution goes to the first valid introduction within the 90-day window, credited when the company submits a verified application. Other roles are listed on the partner hub.
Go deeper on failed automation review records
- Failed automation review archive context checklist: a checklist of retention settings, rebuild scope and business-system policies to confirm before cleanup.
- Failed automation review introduction email: a ready-to-adapt email for introducing an opted-in owner to SourceX.
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 your client can link failed runs to the repairs its people made and the owner opts in, register as a partner, check the company against who qualifies, and read the program terms before you make the introduction.