Inventory alerts with n8n that people actually trust (and how to avoid alert spam)

Recording inventory and reviewing inventory are different jobs. A shop can have accurate stock figures and still depend on someone to inspect every product, compare quantities against thresholds, check incoming orders, and decide what needs attention. As the range grows, that review becomes the bottleneck — and the person doing it starts ignoring it.

Here’s an n8n alert pattern that fixes this. The key insight: an alert system people mute is worse than no alert system.

The pattern, step by step:

  1. Schedule the check. A Schedule Trigger runs it on a rhythm that matches the business — daily 9 AM works for most small shops. Not every hour; alert fatigue starts with frequency.
  2. Read the inventory. Pull records from wherever they live — Google Sheets, Airtable, or an API like Shopify’s. Keep the source behind one node so you can swap it later.
  3. Assess each product. Reorder point and target stock are business-defined per product. The formula: suggested order = target − current stock − quantity already incoming, floored at zero (never a negative suggestion).
  4. Update the full report. Write back the assessment for every product, including healthy ones. The report is your audit trail — when someone asks “why didn’t we get alerted?”, you have the answer.
  5. Alert — but dedupe. Low-stock products go down the alert branch only if no alert was already sent in the last 24 hours. Check against a recorded alert history. This is the single most important step: without it, a product sitting below threshold for a week generates seven identical emails, and the purchaser stops reading them.
  6. Send one clear email per product. Product, SKU, current stock, reorder point, target, incoming quantity, suggested order, supplier contact — everything needed to act, nothing more.
  7. Record the alert timestamp. This feeds step 5’s dedupe check.

The detail that builds trust: account for incoming stock.

In testing, I had headphones at 8 units with a target of 30 — but 25 already incoming. Naive math says “order 22.” The correct suggestion is zero, with the email noting the delivery should be checked rather than ordering more. If your alerts don’t account for stock already on order, purchasers learn to distrust the numbers and go back to spreadsheet-staring.

Honest scope note: I built this as a working prototype (8 nodes) tested with sample inventory in Google Sheets — intentionally inactive, no live deployment. A client version would read Shopify inventory directly instead of a spreadsheet; that integration isn’t built yet. No demand forecasting, no supplier lead times, no purchase orders — the email is guidance for a human purchaser, not an automated buying decision.

If you’ve done inventory automation, how did you handle the “already ordered” problem? Curious what approaches others took.

Great writeup — and an honest scope note, which I appreciate. The dedupe step (alert only if no alert in the last 24h) and the “incoming-aware” formula are exactly the two things that separate a trusted alert system from noise. Your headphones example is the classic case.

One upgrade I’d suggest from production work: track incoming stock as open purchase-order lines with expected delivery dates, not a single “incoming quantity” field per product.

In the workflow, that means a small Data Table (or a dedicated sheet) of open POs — product/SKU, supplier, qty ordered, qty received so far, expected date, status. When you assess each product, you sum only lines with status open and expected date >= today. When a receipt is logged, the stock count goes up and the PO line closes or decrements, so the math stays consistent and you never double-count.

Two failure modes this fixes that a flat “incoming” field doesn’t:

  1. Partial receipts. A PO of 50 where only 20 arrive. With PO lines you subtract what’s actually received; with a single field you’re hand-editing numbers and the audit trail dies.
  2. Stale incoming. A supplier ghosts you and the PO never arrives. Flag POs past their expected date as overdue in the report — “25 units incoming — overdue by 6 days.” That’s the real trust killer: your alert suppresses a reorder (“suggested: 0”) based on stock that isn’t actually coming. Overdue-PO surfacing turns a silent suppress into a visible escalation.

Bonus: that same Data Table doubles as the alert-history store for your step 5 dedupe check. One state store, two jobs, no extra infrastructure.

The step that usually breaks in production is receiving — suppliers confirm, then partial-ship, then go quiet. Plan for that ledger early and the rest of your pattern holds up nicely.

I’d be careful with the expected-date filter here. If those 25 headphones are six days late, dropping them from incoming stock could suggest another order while the supplier is still about to deliver the first one.

I’d keep the remaining quantity on the PO until there’s a receipt, cancellation or recorded adjustment. Separately, I’d mark an overdue delivery ‘check supplier’. Whether it still counts toward the reorder calculation should depend on the latest confirmation and the purchaser’s agreed rule. The ledger should show what’s outstanding even when the delivery is too uncertain to rely on.

There’s also an exception worth testing in the 24-hour rule: yesterday the email said ‘low stock, delivery due’; today the supplier confirms cancellation. That change needs an updated notification even if yesterday’s alert is still inside the suppression window.

I’d test ‘late but still confirmed’ alongside ‘cancelled’ before using the suggestion to buy anything. Same stock level, different purchasing decision. Would supplier confirmation be a manually updated status in your proposed PO ledger, or could it come from the supplier?

Agree on frequency. The other thing that made my low stock alerts trustworthy was dropping the global threshold. “Alert at 5” is fine for a product I sell 1 a week and useless for one I sell 8 a day with a 15 day supplier lead time.

I now compute the reorder point per variant: average daily sales over the last 30 days x lead time days + a small buffer. Daily sales comes from the orders API, lead time is one column in a Sheet the buyer maintains. Alerts dropped by about half and the ones left are ones I actually act on.

Do you send “back above threshold” messages too, or only the down direction? I went back and forth on that.

One way to handle the already-ordered problem is to treat open purchase order lines as their own data source instead of a single incoming number. Each line carries the SKU, remaining quantity, expected delivery date and status, and the alert logic only counts lines expected within the supplier’s lead time. Lines past their due date should be flagged separately rather than counted as stock, since an overdue delivery does not cover a shortage, and a partially received line should only count what is still outstanding. It also makes the check-the-delivery message more useful, because the email can name the specific PO that is late.