Signals & AI
Ask your ecommerce data — in plain language
Sellrics exposes a suite of analytics tools over MCP (Model Context Protocol). Connect them to an AI assistant and ask questions against your real numbers — forecasts, root-cause analysis, what-if simulation.
The problem
The questions that matter — "why did margin drop last week", "what happens to cash if I cut Meta 20%" — rarely fit a fixed dashboard tile, and answering them by hand is slow.
How Sellrics does it
Sellrics runs an MCP server (backend/mcp/) that exposes its analytics tools over an OAuth connector: revenue and SKU forecasts (backend/api/forecast_revenue.py, forecast_sku.py), root-cause decomposition (backend/api/root_cause.py), and what-if scenario simulation, all reading the same fee-aware, completeness-flagged numbers the dashboard shows. The AI does not invent figures — it calls the tools and reports what they return.
What the numbers mean
A forecast projects a metric forward from its own history and seasonality, with a range. Root-cause analysis ranks the channels, SKUs, and campaigns behind a change. A what-if holds everything fixed except the levers you name and re-runs the model.
A worked example
Illustrative example- Scenario
- You ask: "Why did contribution margin drop last week?"
- What Sellrics surfaces
- The assistant calls the root-cause tool and answers: returns on one SKU rose from 18% to 31% and account for 2.7 of the 4.1-point drop; a fee-mix shift toward Amazon accounts for another 0.9. Then you ask "what if I cut that SKU’s ad spend to zero?" and get the cash and margin effect.
- The move
- Use it for the one-off questions between dashboard reviews — it keeps analysis at the speed of the decision.
You
Why did contribution margin drop last week?
Sellrics (via MCP)
Returns on SKU SE01 rose from 18% to 31% and account for 2.7 of the 4.1-point drop. A fee-mix shift toward Amazon FBA accounts for another 0.9. The rest is within normal weekly variation.
Answered by calling the root-cause tool against your real, fee-aware numbers — no figures invented.
Illustrative — a question answered from your real, fee-aware numbers via MCP.