Workflow: one trigger for new reviews from several platforms (Slack routing, template replies, Sheets export)

Disclosure first: I’m the developer of n8n-nodes-prisma-reviews, a community node I built for my own review workflows, and I’d like feedback from people who automate reviews in n8n.

The problem it solves
Each review platform has a different API, auth and JSON shape, so “new review → do something” ends up as one workflow per platform. The node normalises them: one trigger for new reviews on the locations you connect, and the same fields every time (rating, text, author, date, platform, existing reply).

How I use it

  1. New review → Slack, one channel per location, routed by rating with an IF node.
  2. 4-5 stars → reply from a template; 1-3 stars → a row in Notion or Sheets for a human to answer.
  3. A daily Schedule trigger that exports every review to Google Sheets for reporting.

Limits, honestly
Publishing a reply only works where the platform’s official API allows it (App Store, Google Play, Judge.me and a few more). Yelp, Tripadvisor and most others are read-only for everyone, because their APIs don’t accept replies. Test keys return sample reviews, so you can build and test a workflow without connecting a real account.

Install on self-hosted: Settings → Community Nodes → n8n-nodes-prisma-reviews (it’s in n8n’s verification queue for Cloud).

What I’d really like to know: how do you handle review alerts today, and which platform is the most painful for you?

Interesting work, Fran. Bringing review data from 32 platforms into a consistent structure opens up some useful automation possibilities.

I particularly like the human-review step for low-rated feedback. In addition to routing and notifications, reliable duplicate handling and failure recovery are important when processing review events at scale.

I’ve been working on webhook reliability and automation systems with concurrent-request handling, duplicate prevention, conflict detection, and operational monitoring.

Looking forward to seeing how the community node evolves, especially as more platforms approve reply publishing.

1 Like

Thanks, Shital. You picked the exact three things we spent the most time on, so here’s how the node and the API behave today, in case it’s useful to anyone building on it:

  • Duplicates: every webhook delivery carries an event id and retries keep the same id, so de-duplicating on it is enough. Writes take an Idempotency-Key: retry a reply with the same key and body and you get the first response back instead of a second reply.
  • Retries and failures: deliveries are retried with exponential backoff (30 s, doubling, up to an hour between attempts), and a reply that the platform refuses comes back as a reply.failed event with the platform’s reason.
  • The ugly case: platforms whose API can only create a reply, never read or delete it (Judge.me, eBay, Trusted Shops, eKomi). If the connection drops after the request went out, a blind retry could publish the reply twice and nothing could undo it. So the API never retries blindly there: on Google, the App Store and Google Play it reads the review back and reconciles; on the create-only ones it stops, marks the reply “unknown” and sends a reply.unknown event so a human checks. We added that after a developer pointed it out in a public thread last week, and it’s documented at Reply to Reviews by API: Create, Replace, Delete · Prisma.

Monitoring is the part I’d love your view on: today you get the events and the usage endpoint; if you’ve seen a good pattern for surfacing “this location hasn’t synced in X hours” to operators, I’m all ears. And yes, more platforms with replies are the roadmap: Google and Trustpilot as soon as they approve the integration.

That “last successful sync” visibility is an important operational signal.

I’d track the last successful sync timestamp per location, the expected sync interval, and the age of the latest successful sync. If the age exceeds the configured threshold, the system could flag the location as stale and notify the operator.

I’d also separate “no new reviews” from “sync failed” so a quiet location isn’t incorrectly flagged as an outage. A dashboard showing last success, last error, consecutive failures, and current sync status would make troubleshooting much easier.

One useful addition could be a heartbeat or periodic sync-status event, even when no new reviews are found.

1 Like

Thanks, this is really useful. Part of it is already there: every location in the API has status, last_synced_at and next_sync_at, so you can flag a location as stale when last_synced_at is older than its interval, and a connection.updated webhook fires when an account needs to be reconnected.

What’s missing is exactly your point about telling “no new reviews” apart from “sync failed”. I’m adding last_error and consecutive_failures to each location, plus a location.sync_failed webhook after repeated failures, so a dashboard can show last success, last error and failure streak without guessing. I’d rather not send a heartbeat event on every check (too many webhooks for most users), but last_synced_at gives you the same signal on demand.

I’ll post here when it ships. Thanks again for taking the time!

Thanks, Fran. That makes sense. Using last_synced_at with next_sync_at provides a clean basis for stale-sync detection, while last_error and consecutive_failures should help distinguish a quiet location from a failing integration.

One additional consideration: the alert itself should be deduplicated so operators receive one notification when a location becomes stale, rather than repeated alerts on every monitoring check. A recovery notification when syncing resumes would close the loop.

Looking forward to seeing the update ship.

1 Like

Following up, since this shipped after our exchange. After 3 failed checks in a row, a location sends location.sync_failed once per streak, not on every retry, and location.sync_recovered when a check works again. Each location also exposes sync.consecutive_failures, last_error and last_error_at, so a quiet location and a broken one look different. New reviews stay one alert each: review.created fires only the first time a review is stored, and edits come as review.updated. In n8n, pick the events in the Prisma Reviews Trigger. Thanks for pushing on this, Shital.

1 Like