Your systems detect.
Ward decides, executes, and proves it worked.
Six stages. One unbroken chain. The loop is the product. Below is one full cycle, told stage by stage. A real one fires every few seconds across an estate in production right now.
Detect
Ward watches every store, every KPI, every system of record. Anomalies surface against per-store baselines, not aggregate ones.
POS, inventory, labor, finance, weather, events, supplier feeds. Read into Ward continuously. Forecasts run on the same models your planners trust — ARIMA, Holt-Winters, Bayesian hierarchical, gradient-boosted residuals. The deviation isn't in a weekly report; it's on the page the moment it crosses threshold.
Attribute
Every signal is traced to a cause. Vendor short-ship. Promo conflict. Planogram drift. Schedule miss. Receiving dock anomaly. The cause isn't a guess — it's an attribution chain you can audit.
Ward assembles a reasoning graph: which sources it touched, which SQL it ran, which model fit, which competing causes it ruled out. The graph travels with the case. Compliance can re-derive the conclusion in one click.
Recommend
A case opens. The right play is already attached. Owner, deadline, target KPI, success metric — set on day one. No triage meeting. No "let's ask the analyst what to do."
Plays are the retail moves every operator should be running. Stockout escalation. Vendor invoice dispute. Promo conflict cancellation. Planogram correction. Markdown cadence reset. Each one knows its trigger condition, its procedure, its system of record write-back, and the metric it has to move to close.
Execute
The play runs against the system of record. Manhattan. Blue Yonder. SAP. Oracle Retail. Relex. The pricing engine. The promo engine. Read-only by default; any write-back gates on a named human.
Federated query for read paths. Approval-gated writes for mutations. Cedar policy enforces who can do what against which source. The loop never mutates state without an identifiable approver in the audit log.
Audit
Every step on the case has a signature. The detection. The attribution. The play match. The write-back. The approver. Streamed to your SIEM the moment it lands.
Audit-ready by construction. Every number on the page is one click from the SQL, the source tables, the model, and the parameters. Charters and policies live in your repo, versioned and signed. Roll back any change in one click. Diff two versions side by side.
Measure
The case closes only when the KPI moves. Fill rate up. Shrinkage down. Net promo lift positive. The outcome attaches to the play that fired it. The play's win rate updates. The loop sharpens.
A closed Ward case is not a checked box. It's a measured delta against a pre-agreed metric. Every cycle feeds the system: which plays close cleanly, which ones need recalibration, which ones the cause attribution misclassifies. The next signal arrives smarter than the last.
This loop runs 24/7
across every store.
Every cycle sharper than the last. Plays that close win on win-rate. Causes that get misclassified surface in the misclassification log. The loop, by construction, learns.
Walk a real case end to end.
See the loop fire on your data. 90-day pilot. Pre-agreed metrics.
Find out what your data has been hiding.
Tell us about your operation. We’ll show you the problems Ward catches — and the ones your current tools miss.