The business task: keep a trusted number trusted
Finance, sales and the board read the same handful of figures every month: revenue, margin, headcount, cash. Behind each figure sits a pipeline that someone on the analytics team maintains. When an upstream feed changes, a figure can go wrong overnight, and the team has to notice, find the cause, correct it and tell the people who rely on it that the figure is safe again. How fast and how carefully that happens decides whether executives keep trusting the dashboards or go back to spreadsheets.
The record of those repairs is unusual material. It shows how an organization notices a problem, argues about what a metric should mean, and decides when a correction is good enough. That reasoning is spread across tools and almost never published.
How can a company connect data-break-to-repair records across systems?
By following the keys each tool already stores: the test's unique_id in the run results, the work item key in Jira, and the timestamps of the warehouse queries that rebuilt the model. Where a key is missing, the link depends on free text and a person's memory.
Three documented behaviors make those keys usable. dbt writes a run_results.json file for each invocation, listing executed nodes by unique_id with a status, and test results include a failures count. Jira links a commit to a work item when the commit message contains the work item key. Snowflake's ACCOUNT_USAGE.QUERY_HISTORY view keeps the last 365 days of queries.
Worked example: one joined timeline
Illustrative. A fictional 410-person specialty distributor runs the stack above. Its net revenue model breaks after a new returns feed is added.
| Step | When | System | Record | Key to the next step | Join status |
|---|---|---|---|---|---|
| 1. Test fails | Day 0, 06:10 | Model layer | not_null test on the order ID column of the net revenue model: status fail | Test unique_id | Supported if run files were kept |
| 2. Ticket opened | Day 0, 08:30 | Tracker | DATA-482, "Net revenue nulls after returns load" | Test name typed in the description | Unsupported: free text only |
| 3. Fix committed | Day 1, 10:05 | Git, linked to the tracker | Commit "DATA-482 coalesce order ID from returns feed" | Work item key in the commit message | Supported when the repository is connected |
| 4. Rerun passes | Day 1, 11:00 | Model layer | Same test unique_id: status pass | Run timestamp | Supported if run files were kept |
| 5. Model rebuilt | Day 1, 11:02 | Warehouse | Rebuild and dashboard refresh queries in query history | Timestamp and service user | Supported within the 365-day window |
| 6. Report verified | Day 2 | Tracker | Comment: "Finance checked April close" | None | Unknown: who checked, against what |
| 7. Meaning changes | Day 40 | Model layer | Returns netting logic edited; every test passes | None | No record ties it to the report |
Rows 1 to 5 can be joined with keys the systems already hold. Row 6 is a sentence. Row 7 is the case the next section is about.
What breaks when a passing pipeline still changes business meaning
Tests catch broken data, not changed definitions. A dbt data test is a query for failing rows, and it passes when the query returns zero rows. The four generic tests dbt ships (unique, not_null, accepted_values and relationships) check shape and integrity. None of them knows that finance expects net revenue to exclude returns from a prior quarter.
In the Illustrative timeline, the day 40 edit changes how returns are netted. Every test passes, so run_results.json records a clean run with nothing to investigate. Because that file lists only nodes that executed, with their status, it says nothing about what the SQL used to mean. The definition lives in the model code and in manifest.json, and the change is visible only in the commit history. If the commit message carries no Jira key, no ticket links the edit to the report that moved. Snowflake's query history keeps the rebuild queries, but only for 365 days.
So the join that matters most is the one no system makes for you: model change to report consumer. A partner can find out whether the company has it with three questions:
| Question | Strong answer | Weak answer |
|---|---|---|
| Do model changes require a Jira key in the commit message? | Enforced in the repository since a stated year | Encouraged, not enforced |
| Are dbt artifacts kept per run? | Archived to storage for every production run | Not kept after the next run |
| Does anyone record which reports depend on which models? | A maintained list of reports, their owners and their source models | Known by the analytics lead only |
The answers describe coverage. They do not require opening a model, a ticket or a single row of warehouse data.
Who can authorize the conversation, and what stays unknown
The analytics engineering lead can describe the chain. Only the CEO can authorize a licensing discussion, usually with the CFO, COO or general counsel advising, because warehouse data can include customer records under contract terms the data team does not control. The portfolio review handoff checklist shows how to route that.
Before introducing, check the baseline: 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.
What stays unknown until SourceX reviews it under the company's own agreement: whether the rights cover the records, how much of the chain survives, and whether any buyer wants it. Do not promise the company a price or a timeline. SourceX does not claim a connector for every tool named here; delivery format is agreed later.
Go deeper on data-break-to-repair records
- What a single-system data-break export leaves out: one Illustrative case read before and after its surrounding context is added.
- Data-break-to-repair metadata packet: the fields a partner can share without touching raw records, with an Illustrative filled packet.
- Data-break-to-repair completeness checklist: a pass, partial or fail scorecard for judging whether the records are complete enough to discuss.
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 a data team you advise can show its break-to-repair chain at the metadata level and a company officer will take the call, register as a partner, confirm the company against who qualifies, and read the program terms before you submit.