Blog · Procurement and supply chain · Supply chain
Supply chain analytics uses operational records to explain service, inventory and cost performance. Learn which exports to bring, how to define comparable measures and how to reconcile a worked supplier comparison before acting.
Supply chain analytics uses operational records to explain delivery, inventory, forecast and cost performance. It turns exports into comparisons a procurement, planning or operations team can act on, starting with a business question and checking each result against its underlying rows.
A supplier can deliver most units while missing several promised dates. A site can hold more stock while providing worse service. One measure can conceal the trade-off.
Choose questions before asking for files:
| Business question | Measure | Records needed |
|---|---|---|
| Did suppliers meet their commitments? | On-time order-line rate | Orders, agreed dates, receipts |
| Was customer demand fulfilled? | Unit fill rate | Demand lines and matched shipments |
| Did inventory build or fall? | Stock movement and closing balance | Opening balances, receipts, issues |
| How close was the plan? | Absolute forecast error relative to actual demand | Saved forecast and actual demand |
| What activity absorbed spending? | Cost by activity and driver | Ledger, freight and handling records |
Inventory is a balance at a date; demand and shipments accumulate over a period. That distinction also appears in the Census Bureau's manufacturing and trade survey descriptions, which distinguish sales from end-of-month inventories. Match a monthly flow to the right month and a closing balance to the right cutoff.
Use one row per order line for delivery measures, with supplier, site, item, quantity due, original agreed due date and actual receipt date. Combine partial receipts before deciding whether a line was complete on time. Use a separate item-site-period table for demand and forecast comparisons.
On-time order-line rate = Lines received in full by their agreed due date / Lines due
Unit fill rate = Units fulfilled against the included demand / Units demanded
Weighted absolute percentage error = Sum of absolute forecast errors / Sum of actual demand
Agree on these definitions and the treatment of cancellations, early receipts, backorders and returns. Save the forecast available when a decision was made; a revised version created afterward is different evidence.
For technique selection, see analytical methods for supply chain management.
An illustrative US distributor reviews one site for September 2026. The suppliers serve different items; all quantities below have been converted to individual units on a common counting basis. The delivery table summarizes the order-line export.
| Supplier | Lines due | Lines complete on time | Units demanded | Units fulfilled |
|---|---|---|---|---|
| A | 10 | 9 | 1,000 | 980 |
| B | 8 | 6 | 800 | 700 |
| C | 12 | 12 | 1,200 | 1,200 |
| Total | 30 | 27 | 3,000 | 2,880 |
The company on-time rate is 27 / 30 = 90.0%. Its unit fill rate is 2,880 / 3,000 = 96.0%. They differ because one counts order lines and the other counts quantities; neither percentage replaces the other.
The forecast comparison uses the same demand period, with forecasts frozen before September began:
| Supplier's item group | Actual demand | Saved forecast | Forecast minus actual | Absolute error |
|---|---|---|---|---|
| A | 1,000 | 1,000 | 0 | 0 |
| B | 800 | 900 | 100 | 100 |
| C | 1,200 | 1,100 | -100 | 100 |
| Total | 3,000 | 3,000 | 0 | 200 |
Total forecast equals total demand, yet absolute error is 200 units. Weighted absolute percentage error is 200 / 3,000 = 6.7%, rounded to one decimal place. The offsetting errors prove why a matching grand total does not establish an accurate item-level plan.
First, sum the source rows independently: 10 + 8 + 12 = 30 lines due, and 9 + 6 + 12 = 27 complete on time. Reconcile the 3,000 units of demand to the demand export and the 2,880 fulfilled units to the matched shipment rows. Keep unmatched lines in an exception list rather than silently discarding them.
Then reconcile inventory on its own movement basis:
| Inventory movement | Units |
|---|---|
| Opening inventory | 1,200 |
| Receipts | 3,100 |
| Issues | -3,000 |
| Closing inventory | 1,300 |
The identity is 1,200 + 3,100 - 3,000 = 1,300. Inventory issues can include transfers, internal use or other movements beyond customer fulfillment. Explain those categories before expecting this issue total to equal the shipment total above.
Check the spread as well as the aggregate. NIST's guide to measures of scale explains that variance uses squared deviations from the mean and can be affected strongly by extreme observations. A comparison of average lead times should therefore retain unusual late receipts for investigation, with any exclusions stated.
Procurement should investigate supplier B's missed lines and operations the unfulfilled demand. Planning should examine the opposite forecast errors. Finance should approve cost estimates attached to proposed actions.
Check order changes, receiving delays and substitutions before assigning a cause. Use predictive supply chain analytics for future outcomes and validation, and supply-chain risk analytics for continuity exposures.
Bring a sample of your own order, receipt, inventory and forecast exports; Covirage maps the columns, and its tools compute and check comparisons from those files. The external AI model explains the results with every figure traced to its rows. Create a free Covirage account, then set up your business and bring your data.
See AI analytics for supply chain for the product workflow and Covirage for supply-chain teams for the business context.
Start with the question. Delivery analysis needs order lines, promised dates and receipts. Inventory analysis needs dated balances and movements. Forecast analysis needs the forecast saved before demand occurred and the matching actuals. Consistent supplier, item, site and period identifiers make the comparisons usable.
Descriptive analysis measures what happened. Diagnostic analysis compares possible reasons. Predictive analysis estimates future outcomes and requires validation against later actuals. Scenario analysis compares decisions under stated assumptions. Each serves a different question, and a descriptive pattern alone does not establish cause.
Ownership is shared but should be explicit. Procurement owns supplier definitions and follow-up; operations owns receipts, movements and service records; planning owns forecast versions; finance approves cost boundaries and reconciliations. One named report owner should maintain definitions and resolve exceptions before the results are distributed.