Supply Chain Visibility with Fabric and Dynamics 365 SCM

Stop chasing spreadsheets. See every order, item, and shipment in one live system—and act before issues explode.

What you’ll learn

• Why “more dashboards” ≠ visibility—and how to model the relationships that matter

• The reference architecture for Fabric + D365 SCM (ingest → model → alert → act)

• How to wire real-time signals (orders, WIP, ASN/ETA, telemetry) into OneLake

• Playbooks for proactive alerts, auto-reorder, and exception handling

• Scenario planning & continuous improvement with Fabric analytics and Power BI

• KPIs, governance, and a 14-day rollout plan to prove value fast

---

The problem in one sentence

Most systems show snapshots; operations need a storyline—item lineage from supplier to doorstep with every dependency visible.

---

Reference blueprint (Fabric + Dynamics 365 SCM)

Ingest & unify

• Data sources: D365 SCM (orders, inventory, WIP, production, shipments), TMS/WMS, carrier APIs, vendor portal, IoT/sensor feeds.

• Pipelines: Fabric Data Pipelines/Notebooks + Event Streams for near–real time (ASNs, GPS/telematics).

• Storage: OneLake → Lakehouse bronze (raw), silver (validated), gold (conformed supply model).

Model & serve

• Conformed entities: Supplier, PO, ASN, Receipt, Batch/WIP, InventoryPosition, Shipment, Delivery, Route, Exception.

• Lineage links: PO → ASN → Receipt → WIP Batch → Finished Good → Shipment → Delivery.

• Semantic layer: Power BI dataset over gold tables with row-level security (by region/site/customer).

Decide & act

• Rules & ML: stockout risk, lead-time variance, ETA deviation, supplier score drift.

• Automation: Power Automate / Logic Apps to trigger reorder, route re-plan, escalation, or customer comms.

• Surfaces: Power BI control-tower dashboards, Teams alerts, email/SMS for critical exceptions.

---

What “real visibility” looks like (capabilities)

1. Item-level traceability

   • Click from sales order → exact lot/serial → shipment → live map → proof of delivery.

2. Predictive exceptioning

   • Alert when ETA deviation > X%, safety stock breach in Y hours, or supplier OTIF dips.

3. Closed-loop actions

   • Auto-create purchase requisition, advance an alternate supplier, or split a shipment—logged and auditable.

4. Unified vendor performance

   • Composite score (OTIF, quality, cost variance, disruption risk) feeds sourcing & allocation automatically.

5. 360° IPO view

   • Inputs (POs), Process (WIP stages), Outputs (FG & shipments) on one timeline with bottleneck highlights.

---

Control-tower dashboards (must-have tiles)

• Network map: live shipments, heat by risk/severity, click-through to route & carrier.

• Order promise board: jeopardized lines with “time to miss” and recommended action.

• Inventory risk matrix: SKU × site with days cover, ATP, and reorder suggestions.

• Supplier scorecards: OTIF trend, lead-time variance, defect rate, auto-allocation % change.

• Throughput & bottlenecks: WIP dwell time by station; backlog aging; takt vs plan.

• Financial impact: revenue at risk, expedite cost forecast, stockout avoidance saved.

---

Automation playbooks (copy/paste)

A) Stockout prevention

• Trigger: DaysCover < threshold or promo spike detected

• Action: propose PO qty, route to buyer; if SLA-critical, auto-PO within guardrails; notify sales.

B) Late ASN / ETA slip

• Trigger: ETA variance > 2 hrs (high-value SKUs)

• Action: reroute shipment, alert DC, update customer promise, open alt-supply check.

C) Supplier slip pattern

• Trigger: 3-week OTIF decline or lead-time variance > 20%

• Action: reduce allocation by X%, elevate secondary supplier, schedule SQE review.

D) Production bottleneck

• Trigger: WIP dwell > control limit at station

• Action: rebalance labor, resequence jobs, notify maintenance if MTBF crossed.

---

Data model (gold layer) essentials

• Facts: FactOrderLine, FactWIP, FactInventorySnapshot, FactShipmentEvent, FactException.

• Dims: DimSKU, DimSite/DC, DimSupplier, DimCarrier, DimCustomer, DimCalendar.

• Keys: OrderLineId, Lot/Serial, ASNId, ShipmentId, RouteId.

• Lineage: maintain a link table to traverse PO→ASN→Receipt→Batch→FG→Shipment→Delivery.

---

KPIs that matter (with targets to tune)

• OTIF (on-time, in-full) ↑

• Lead-time variance ↓

• Stockout rate / backorders ↓

• Inventory turns & days cover ↑ / right-sized

• Exception MTTD/MTTR ↓ (detect/recover)

• Revenue at risk saved (from proactive actions) ↑

• Expedite/penalty costs ↓

---

14-day quick start (prove value fast)

Days 1–3 – Connect D365 SCM entities + one carrier feed to OneLake (bronze → silver).
Days 4–6 – Build gold tables (orders, inventory, shipments, exceptions) + lineage links.
Days 7–9 – Publish Power BI control-tower v1 with RLS; wire two alerts (stockout, ETA slip).
Days 10–12 – Automate one action (auto-reorder or reroute) via Power Automate.
Days 13–14 – Pilot with one product family & two sites; baseline KPIs; capture wins & gaps.

---

Common pitfalls (and how to avoid them)

• Siloed IDs (no lineage): create a canonical key map and link table across PO/ASN/Shipment.

• Batch ETL lag: use Event Streams for ASNs/GPS; mix streaming + scheduled loads.

• Dashboard sprawl: one semantic model; certified datasets; RLS by region/customer.

• Alert fatigue: start with 3–5 high-impact rules; add severity & cool-downs.

• Automation without guardrails: require approvals for high-cost actions; log everything.

---

Security & governance notes

• RLS by site/brand/customer; DLP on external connectors; sensitivity labels on gold.

• Data contracts per source (SLAs, freshness, schema change policy).

• Ops runbook: who’s paged on which alert, time windows, and escalation paths.

• Audit trail: persist every recommendation + action + override for compliance.

---

FAQs

Q: We already have D365 SCM dashboards—why Fabric?
A: Fabric unifies all events (SCM + TMS/WMS + carrier + IoT) into OneLake with real-time processing, predictive alerts, and closed-loop automation—beyond transactional reporting.

Q: Do we need IoT to start?
A: No. Begin with D365 SCM + carrier feeds; add sensors later for higher fidelity.

Q: How “real-time” is this?
A: Mix streaming for events (ASNs/GPS) with 5–15-minute mini-batches for operational tables.

Q: What’s the first alert to turn on?
A: Stockout risk on A/AX SKUs and ETA variance on revenue-critical orders.