First 5 minutes — contain the blast
- Kill‑switches: immediately pause Zaps/Make scenarios, disable HubSpot/Salesforce workflows or stop scheduled jobs and imports; treat all running syncs as suspect.
- Pause jobs: stop any batch imports (CSV uploads), queued ETL jobs or middleware retries so failures don’t cascade and generate noise.
- Canary checks & evidence: create a single test record in the source and note timestamps; grab screenshots or task logs from the middleware and sink for the next 30 minutes.
Next 20 minutes — one‑page diagnostics and three smoke tests
Start a one‑page flow: check the Source (recent changes to forms, CSV headers, API keys, rate limits), Middleware (task history, auth errors, throttling, schema / mapping failures, dead‑letter queues) and Sink (permissions, recent writes, workflow rejections). Look for clear signs: schema mismatch errors in middleware, 401/403 from the API, or bulk rejected rows on CSV import.
Run three quick smoke tests to prove whether data flows: create a canary record in the source and watch for it in the middleware and sink; run an isolated middleware replay or single‑task re‑run (Zapier/Make let you run a single task); and check the sink for a recent timestamp or the canary field value (search HubSpot/CRM for the test email/ID). These three checks tell you which layer is failing.
Use platform examples to interpret results: a HubSpot issue often looks like permission errors, workflow changes or blocked imports; Zapier/Make will show task error messages and rate‑limit counts; a CSV import failure commonly comes from header mismatch, encoding or stray separators. Keep the focus on which layer (source/middleware/sink) showed the first failure, not the whole stack.
Final 5 minutes — quick fixes, owner messages and next steps
Common fast fixes you can do without engineers: pause + replay a small batch (limit to 50–100 records), toggle a mapping back to the previous value, apply a simple provenance filter (exclude records lacking a required ID), or quarantine failed rows to a dead‑letter sheet for manual review. For CSVs: correct headers/encoding, run a small dry import; for Zapier/Make: fix the field mapping and re‑run a single task; for HubSpot: pause the workflow, correct the trigger or field, then test with the canary.
Quick owner/escalation templates you can copy: "Subject: [30m] Integration failure — {integration name}\nBody: Short summary of what failed, timestamp of last good transaction, immediate action taken (paused jobs), owner assigned, and next check time (in 30m)." If escalating: "Subject: Escalation: integration {name} — needs engineering\nBody: include the one‑page diagnostics (source/middleware/sink), screenshots/log snippets, and a clear ask (restore access, check API quota)." Assign a named owner and set a 30‑minute follow‑up.
Clear next steps: quarantine dead‑letter records into a sheet with a simple exception reason column, plan a short backfill limited by size (sample, test, run in low‑traffic hours), and schedule a brief post‑mortem to capture root cause, immediate fix, owner and preventative action (add a canary, a kill‑switch or a simple retry/backoff). If you want local help stabilising or turning this into a repeatable routine for Fareham teams and across the South Coast, our marketing automation and CRM optimisation work can support the follow‑up — see marketing automation support in Hampshire. If you prefer, Optira can step in to run the triage and help stabilise the integration quickly.