Signals
How MarginEngine detects and surfaces money-relevant patterns in your data.
What is a Signal?
A Signal is a detected money-relevant fact. Not a generic notification — a specific finding with dollars attached.
Bad example: "Refund rate is high." Good example: "$800 Leak — Refunds on SKU-RED (62% return rate this month)"
Every Signal starts with dollars because that's what matters to your business.
Signal types
| Type | Meaning | Example |
|---|---|---|
| Leak | Money is leaving your business now | "$1,200 Leak — Discounts on already-low-margin products" |
| Risk | A leak could start soon if nothing changes | "$340 Risk — Shipping costs up 15% vs last month" |
| Fog | We can't see something important | "Gross margin hidden — set Product Costs for top SKUs" |
| All Clear | No anomalies detected | "Revenue stable, margins holding, no new leaks" |
How Signals are generated
MarginEngine generates Signals from deterministic patterns in your data — not from AI guesses. The AI chooses which Signal is most important and explains it in plain language, but the underlying detection is math.
Examples of what triggers Signals:
- SKU refund rate exceeds 20%
- Product margin drops below a threshold
- Revenue concentration risk (one SKU = too much revenue)
- Discount erosion on already-thin margins
- Ad spend efficiency changes (Reported vs Real gap widens)
Acting on Signals
Every Signal includes a Next Move — the single action that would address it. You're never left with "something is wrong" without a path forward.
Next Moves open evidence in the Canvas, where you can see the full breakdown, drill into specific orders, and decide what to do.
MarginEngine suggests. You decide. The AI never takes action on your behalf without explicit approval.