Skip to main content
Inventory · DIO · Fill Rate · MAPE

The number on the system
is not the number on the floor.

Inventory analytics for manufacturers, FMCG and distributors. DIO, fill rate, slow-moving stock, demand variability and safety-stock recalculation — on Microsoft Fabric. Surfaces the ERP-to-warehouse gap before the next stockout writes the cheque.

The Problem

Patterns we see in every engagement

ERP perpetual inventory is 60–80% accurate at most mid-market industrials. The demand planner does not trust it. They live on Excel. The Excel is also wrong, just differently. These are the four numbers that close the gap.

01

Safety stock formula running on lead times from 2019.

Your formula assumes 14-day lead time. The last six lead times averaged 23 days. Nobody updated the formula. The reorder dashboard is still running the old number. You are either stocked out three times a quarter or carrying 38% more working capital than you need to.

02

Perpetual inventory vs cycle-count drift.

The system says 480 units. The warehouse cycle count says 421. Variance gets explained as 'shrinkage' until it is 12% of total stock value. The dashboard surfaces drift per location per SKU class — drift becomes the metric, not the surprise.

03

Slow-moving stock that grew into an obsolescence problem.

SKUs with zero movement in 180 days. SKUs with batch-expiry approaching. SKUs that were promoted four quarters ago. The CFO finds out at year-end when the write-off lands. The dashboard names the exposure 90 days in advance.

04

Fill rate measured at customer level, not order-line level.

Your customer scorecard measures order-line fill. Your internal dashboard measures order-header fill. The gap between the two numbers is 4–7 percentage points and it shows up as customer chargebacks the procurement team cannot reconcile.

What we build

What we build

Eight dashboards. Each one answers a question an Operations Director or Demand Planner asks weekly.

01

DIO and Days of Inventory per SKU class

Replaces

The single 'total inventory days' number that hides A-class stockouts and C-class overstock under one average.

  • DIO calculated per SKU, rolled to category, location and SKU-class (A/B/C/D)
  • 30 / 90 / 365 day trend so seasonal pattern is visible separately from underlying drift
  • Target DIO per class — A-class lean, C-class loose, D-class flagged for delist review
  • Drill from aggregate to the specific SKU and the specific lot driving the gap

Aggregate DIO stops hiding the truth. Working capital conversations get specific about which SKUs to act on this week.

02

Fill rate — order-line, case, value

Replaces

The internal '95% fill rate' that does not reconcile to the customer scorecard saying 89%.

  • Order-line fill, case fill, value fill — all three computed and reconciled
  • Per customer, per SKU, per lane — drill to which combination is driving the customer chargebacks
  • Threshold alert when any top-20 customer drops below the agreed SLA
  • Reconciliation to customer-side scorecard (Lulu, Carrefour, Coles, Woolworths) — gap explained, not denied

Internal number reconciles to the customer scorecard. Chargeback disputes have evidence, not opinion.

03

Safety stock recalculation per SKU

Replaces

The static safety-stock figure set in 2019 that has not been adjusted for the lead-time variability since.

  • Per-SKU recalc using current lead-time variability and current demand variability
  • Service-level target per SKU class — 99% on A, 95% on B, 90% on C
  • Refreshed weekly — recalculated values flow back to the ERP if your team chooses to act on them
  • Show the as-was vs the recalculated number side by side so demand planner sees the gap

Demand planner stops working from a 2019 formula. Safety stock reflects 2026 reality.

04

Slow-moving and obsolete stock alerts

Replaces

The year-end obsolescence write-off that surprises finance every January.

  • SKUs with zero movement in 90 / 180 / 365 days flagged with replacement-cost exposure
  • Batch-expiry within 30, 60, 90 days — value at risk per SKU
  • First-in-first-out adherence — when batches are issued out of FIFO order, it surfaces
  • Action queue feeds back to the planning team — markdown, return-to-vendor, scrap, hold

Obsolescence becomes a 90-day-out conversation, not a year-end surprise.

05

Demand forecast accuracy (MAPE)

Replaces

The annual forecast review that says 'we are mostly accurate' without naming where the bias is.

  • MAPE per SKU per region — where are we systematically over-forecasting, where under
  • Bias indicator separate from accuracy — chronic over-forecasting bleeds working capital
  • Hierarchy of MAPE — total → region → category → SKU so the conversation can drill to root cause
  • Forecast-versus-actual with rolling 13-week trend — pattern shifts surface before the next plan cycle

Forecast quality stops being asserted and starts being measured per SKU. Planning conversations have a metric.

06

On-hand vs perpetual gap with cycle-count age

Replaces

The 'shrinkage line' that finance carries on the P&L because nobody knows the root cause.

  • Per-location, per-SKU variance between ERP perpetual and last cycle count
  • Cycle-count age — when was each SKU last verified
  • Drift trend so chronic offenders surface (people, process or system)
  • Daily exception report to warehouse manager and inventory controller

Cycle-count discipline becomes measurable. Shrinkage becomes investigable, not accepted.

How we work

From data audit to live dashboard in 6 weeks

We start with the data quality conversation. Inventory analytics with 60% accurate source data lies confidently. We surface the gap first.

01

Discover — score the ERP inventory module

Two weeks. We pull 6 months of inventory transactions, cycle-count history, fill-rate logs. Score data quality per SKU class. Define the 4 numbers with operations. Identify the warehouses where signal is strong enough to act and the ones where data discipline needs to come first.

02

Prototype — one warehouse, top-100 SKUs

Two weeks. Build the 4-number dashboard for one warehouse and the top 100 SKUs by value. Parallel-run against the existing inventory report for 2 weeks. Demand planner sees both. Trust gets built.

03

Deploy — all warehouses, automated alerts

Three to five weeks. Roll out to all warehouses and SKUs. Wire alerts. Train the demand planner and warehouse manager on the new workflow. Add automated safety-stock recalc and predictive replenishment in the Expand phase.

Technology stack

Lakehouse

Microsoft FabricOneLakeDelta LakeDirect Lake

Pipelines

Fabric Data PipelinesAzure Data FactoryCDCFabric Mirroring

Visualisation

Power BI Direct LakeSemantic ModelsPower BI MobilePaginated Reports

Source ERP

SAP S/4HANASAP ByDesignMicrosoft Dynamics 365NetSuiteEpicorSageAcumatica

Source Ops

WMS (Manhattan, Blue Yonder, Korber)TMSCycle-count systemBarcode scan data

Forecasting

Azure MLPREDICT() in FabricFabric NotebooksDemand sensing via POS-out

Common questions

What buyers ask us

Our ERP is SAP S/4HANA Embedded Analytics. Why not just use that?

Embedded Analytics is a good SAP-internal tool. It struggles when you need to combine inventory with sell-out (FMCG), with planned production (Manufacturing), with customer-level fill rate. The Fabric Lakehouse holds all three. The Power BI dashboard sits over the whole picture. SAP-internal reports stay useful for statutory and SAP-specific drill — they do not need to also do cross-system.

How long until the safety-stock recalc actually changes anything?

The dashboard surfaces the gap from day one. Acting on it is a workflow change — the demand planner has to start using the recalculated number instead of their gut. That is typically a 6–12 week change-management curve. We support that but it is your team's lift.

What about Snowflake or Databricks for the data layer?

We have shipped on both. For most mid-market industrial businesses already on Microsoft 365 and Azure, Fabric is the right answer on cost and time-to-value. For some businesses with heavy data-science workload or multi-cloud strategy, Databricks is the right answer. We assess — see our stack-choice blog series.

How much does it cost?

Discover USD 8,000–12,000. Prototype USD 18,000–32,000 depending on ERP count. Deploy USD 40,000–110,000 depending on warehouse count, SKU count, ERP count. Quoted precisely after Discover.

Can you integrate Project44 / FourKites for in-transit?

Yes. We integrate the major in-transit visibility providers as supply-side feeds. The dashboard combines on-hand inventory with in-transit so the demand planner has the full picture, not just what is in the warehouse today.

Ready to move

Book a 30-minute Inventory diagnostic

30 minutes with Amit. No slides. No pitch deck. No obligation to proceed. We walk through your current ERP inventory module, your safety-stock formula and where the dashboard would land first.