SCM Master: Executive Cockpit

Procurement & supply-chain analytics · live API
💡 Click any category, supplier or product to filter the whole board.
drag to orbit · scroll to zoom · right-drag to pan
Period: scopes spend, supplier share & the demand chart · the capacity tile is live "now"
Forecast Accuracy
–
backtested · all years
Total Spend
–
across – suppliers
Top Supplier Share
–
concentration risk
Stockout Risk
–
SKUs below cover

Predicted vs Actual demand — walk-forward backtest

Supplier concentration (click to filter)

Spend by category (click to filter)

📋 Procurement insights deterministic · auditable

    Steering to the fleet plan: every KPI, its goal, who owns it
    Export CSV
    The five warehouse KPIs · availability, capital tied up, cover, aging, write-down risk · the role's standard measures, as the backend groups them
    Ordering: delivery reliability and forecast quality · the flow around the five, from the same response

    How the goals are set

    Cockpit scorecard · computed in this cockpit from the inventory, spend and backtest reads, not the backend's KPI list · the steering numbers are above
    📦 Inventory & warehouse cycle · live snapshot (now)
    🎯 Demand planning & forecast · full backtest (all years)
    🏭 Sourcing & supplier risk · all-time spend (year view on the Spend tab)

    How to read this scorecard

    🟢 on target · 🟡 watch · 🔴 act. Metrics are computed live from the SCM Master API (inventory, spend, and 12-month forecast backtest). Metrics that require PO / receipt / GL data not exposed by the API (Perfect Order %, OTIF, Cash-to-Cash, PPV) are intentionally omitted rather than fabricated.
    Period:
    Total Spend
    –
    all-time · selected scope
    Units
    –
    items procured
    Avg Spend / Unit
    –
    blended
    Categories
    –
    in scope

    Spend by category (click to filter)

    Supplier split (click to filter)

    Product detail (click a row to filter)

    ProductCategoryUnitsSpend% of total
    SKUs Tracked
    –
    active products
    Below Safety Stock
    –
    replenish now
    Stockout Risk
    –
    cover < lead time
    Inventory Value
    –
    on-hand × price

    On-hand vs safety stock

    Days of cover (red = below lead time)

    Replenishment pipeline · Reorder Point = burn × lead + safety

    ProductOn handSafetyLead (d)Reorder ptDays to reorderActionNext ETA
    Forecast accuracy ⓘ
    –
    1 − WMAPE · pooled · higher = better
    Forecast bias
    –
    over / under · drives safety stock
    Forecasts
    –
    evaluated
    Worst Category
    –
    improvement target
    MAPE (simple mean) sits in the detail below — it is inflated by intermittent SKUs with near-zero actuals (dividing by a tiny number), so WMAPE + bias are the reliable headline for this demand profile.

    Predicted vs Actual demand

    Accuracy by category · 1 − WMAPE (higher = better)

    Error trend over forecast months

    🧮 Clean-sheet should-cost — the margin lever
    Addressable Savings
    –
    vs fair target price
    Avg Cost Gap
    –
    quote vs target
    Products Modelled
    –
    with a BOM
    Total Margin Stacked
    –
    quote vs cost floor

    Quote vs Target vs Floor (per product)

    What this is

    A clean-sheet teardown rebuilds each box from components — memory & flash indexed to commodity markets (DRAM/NAND), metal & PCB to their indices, and CPU/GPU as a list-price benchmark band. That yields a defensible cost floor.

    Headline gap = quote vs target (floor + a fair margin the supplier would accept) — no one sells at cost, so this is the credible ask. The floor is the backstop: total margin stacked in the quote, a ranking signal for where to push. The floor moves with the memory market, so the number stays current as DRAM swings.

    Gap by supplier (who's furthest above floor)

    ProductFloorTargetQuoteGap→targetGap %
    🧾 Total cost of ownership, per device

    TCO per device
    –
    –
    Cost per month in service
    –
    –
    Rent per month
    –
    –
    Margin per month
    –
    –

    Where the money goes, per device

    Layerper deviceper monthshare

    Where to get better

    Every model, full TCO click a column to sort, a row to open the model and its serials · the darker a layer cell, the more euros in it

    The rates behind the numbers quantities are counted in the data; a rate is a public price with its source, or a placeholder with the role that sets it

    RateSmartphoneTabletLaptopUnitOwnerBasis
    📦 Orders — what's arriving, and what it covers · current year
    Open orders
    –
    in flight
    Incoming units
    –
    not yet delivered
    Delivered (YTD)
    –
    orders
    Inbound value
    –
    open POs

    Order pipeline click a row to see its contents · soonest arrival first

    POSupplierStatusUnitsValueCovers gapPromisedActual / nowSlip
    AUTONOMOUS PROCUREMENT

    Decision loop & audit trail

    The agent advises; deterministic code decides. Run the gate (dry-run by default), watch the three checks resolve to a tier, and read the permanent audit log below — click any row to trace a decision to its inputs and, when placed, its purchase order.
    DRY-RUN

    Intake — run the weekly gate

    Runs the gate over the whole catalogue as a dry run; nothing is placed. Not instant: the model is asked once per proposed line, so a run takes a while. The button counts the seconds while it runs.
    Adjust the look-back (default 7 days, the weekly cycle)
    How far back the run looks for what changed: devices decommissioned in that window become replacement demand, and the same number of days is the demand period of the stock position the run nets against. It is not a forecast horizon; the reorder floor and the forecast shortfall do not depend on it.
    For one model, manufacturer or class, how much is needed and why: the console's Purchasing › Ordering mask. This gate is a cycle over everything, not a per-product question.

    Decision audit trail (append-only · newest first)

    WhenProductQtyTotalTierConfPO