A supply chain dashboard is not a wall of numbers; it is a small set of metrics, defined precisely enough to be argued about, reviewed on a fixed cadence. This article defines the core KPI set for an ecommerce supply chain across cost, service and inventory, with formulas, sensible target ranges and the review rhythm that turns readings into decisions. It is written for operators building a first real dashboard, or repairing one nobody opens.
What a dashboard is actually for
A KPI exists to trigger a decision, and any metric that never changes a decision is decoration. That test explains most failed dashboards: they measure what is easy to extract rather than what is expensive to ignore, and they present readings without thresholds, so every number looks approximately fine. A working dashboard has three properties. Every metric has a written definition — formula, data source, owner — so a changed reading can be investigated rather than debated. Every metric has a threshold or band, so "red" means something agreed in advance rather than something argued after the fact. And every metric lands on a named review — weekly or monthly, with people who can act. The metric definitions below follow the same discipline used in SLA design, covered in our guide to designing supply chain SLAs; the difference is audience. An SLA holds a partner to standards; a dashboard tells you whether your whole operation is healthy.
Cost metrics
Cost metrics answer whether the supply chain is becoming more or less expensive per unit of revenue, and they age faster than most operators expect because freight, tariffs and fulfillment fees all move independently. Landed cost ratio — total landed cost of goods sold divided by revenue — is the anchor: it absorbs product cost, freight, duties and inbound handling in one number, and its trend line is the earliest honest signal of margin pressure. Freight share of landed cost isolates the most volatile component; when it rises for two consecutive quarters, the mode mix or the lane mix deserves a review, using the trade-offs in our guide to air versus sea freight. Cost to serve per order — pick, pack, packaging, outbound shipping, payment fees and a per-order share of support — catches fulfillment drift that per-shipment rate cards hide, especially as packaging materials and carrier surcharges change. And returns cost as a share of revenue captures the reverse flow, which most dashboards omit entirely even though it compounds with the defect rate.
Service metrics
Service metrics answer whether the operation keeps the promises the storefront makes. Dispatch on-time rate measures orders leaving the warehouse within the promised handling window after cut-off, and it is the single most customer-visible number on the board. Transit in-range rate measures delivered parcels landing inside the published range per lane — a carrier-owned number, but one you own the consequences of, since published ranges are your promises. Customer-reported defect rate, coded from support tickets and verified against inspection records, connects the customer experience back to production batches. Exception rate per hundred orders counts the orders that needed human intervention — address fixes, reships, carrier claims, stalled parcels — and it is the best early-warning number on the board, because exceptions rise before refunds do. WISMO share, the fraction of support contacts that are "where is my order", measures the tracking loop rather than the customers asking.
Inventory metrics
Inventory metrics answer whether the cash sitting in stock is sized to the demand it serves. Weeks of cover per SKU — units on hand divided by average weekly demand — is the working metric for replenishment, and its distribution matters more than its average: depth on the hero SKUs, thinness on the tail. Stockout days for load-bearing SKUs count the revenue that never happened, and they belong on the dashboard precisely because nobody complains about them internally. Inventory turns — cost of goods sold divided by average inventory value — is the aggregate efficiency read, useful for direction rather than for daily decisions. Aged stock share — the fraction of inventory value older than a defined age — is the markdown preview, and it rises quietly for quarters before it becomes a clearance event.
The KPI reference table
The core set on one page. Target ranges are common starting points drawn from industry practice — they vary significantly by category, lane and program, and each program should set its bands from its own trailing data rather than from any published table:
| Metric | Definition | Common starting range | Cadence |
|---|---|---|---|
| Landed cost ratio | Landed COGS divided by revenue | Watch the trend, not an absolute (varies) | Monthly |
| Freight share of landed cost | Freight and duties divided by landed COGS | Stable within a few points quarter to quarter (varies) | Monthly |
| Cost to serve per order | Fulfillment, packaging, shipping, fees, support per shipped order | Stable with volume growth (varies) | Monthly |
| Dispatch on-time rate | Orders dispatched within promised window, per WMS scan | 97–99% (varies) | Weekly |
| Transit in-range rate | Deliveries inside published range, per lane, per carrier scans | 90–95% (varies by lane) | Weekly |
| Customer defect rate | Verified defect complaints per 100 orders | 0.5–2% (varies by category) | Weekly |
| Exception rate | Orders requiring manual intervention per 100 orders | Low single digits, trending down (varies) | Weekly |
| WISMO share | Order-status contacts divided by total support tickets | Under 30% (varies) | Weekly |
| Weeks of cover (hero SKUs) | On-hand units divided by average weekly demand | Reorder point plus buffer per SKU (varies) | Weekly |
| Stockout days | Days a load-bearing SKU was unsellable | Near zero on hero SKUs | Weekly |
| Inventory turns | COGS divided by average inventory value | Higher is leaner; compare to own history (varies) | Monthly |
| Aged stock share | Inventory value older than the defined age threshold | Small and stable (varies) | Monthly |
Twelve lines is deliberately close to the working maximum. Programs add metrics easily and subtract them never; the discipline is keeping the board small enough that a weekly review actually reads all of it.
Designing the review cadence
The cadence turns readings into decisions, and it works when each meeting has a fixed menu. Weekly, thirty minutes on the service block: dispatch, transit, defects, exceptions, WISMO, stockouts — operational owners present, each red reading leaving with an owner and a date. Monthly, an hour on the cost and inventory blocks: landed cost ratio, freight share, cost to serve, turns, aged stock — and a look at whether any weekly metric's trend needs a threshold recalibrated. Quarterly, a structural pass: are the metrics still the right ones, are the bands still honest given current volumes and lanes, and which metric changed a decision this quarter — because a metric that influenced nothing is a candidate for removal. Trends beat readings on every cadence: a defect rate at 1.2% means little alone, but 1.2% after three quarters of 0.6% is a conversation. Annotate the timeline with events — campaign launches, carrier changes, policy shifts like the US de minimis suspension documented in our guide to the end of US de minimis — because unexplained cliffs get rationalized and annotated ones get investigated.
A build checklist
- Every metric has a written definition: formula, data source, owner.
- Thresholds and bands set from trailing data, not aspiration, and agreed with whoever answers for the number.
- One system of record per metric — no dueling spreadsheets.
- Weekly service review scheduled with a fixed menu and named owners.
- Monthly cost and inventory review scheduled, trends on top of readings.
- Quarterly recalibration on the calendar: bands, metrics, removals.
- Events annotated on the timeline the day they happen.
- Partner-visible metrics shared from the same scoreboard both parties read.
If your fulfillment or sourcing partner reports their own numbers, insist on shared definitions rather than parallel reports — the evaluation questions that keep a partner honest on measurement overlap with the ones in our guide to choosing a 3PL. Two truth systems for the same shipments is not redundancy; it is the start of every measurement dispute.
Frequently asked questions
How many KPIs should a supply chain dashboard have?+
Eight to twelve for most growing programs: three or four cost, four or five service, four inventory, with overlap between blocks deliberate. Beyond that, reviews stop reading the whole board and start skimming it. Add a metric only when one has been removed, or when a new failure mode has proven expensive enough to deserve permanent visibility.
Which metrics should be reviewed weekly versus monthly?+
Weekly: anything customer-visible and fast-moving — dispatch, transit, defects, exceptions, stockouts. Monthly: cost aggregates and inventory structure — landed cost ratio, cost to serve, turns, aged stock — which move slowly and reward trend reading. Weekly numbers reviewed monthly lose their actionability; monthly numbers reviewed weekly just generate noise.
What defect rate should we target?+
There is no honest universal number — categories differ enormously in what counts as a defect and how customers report one. Set the target from your own verified trailing rate, tighten it as inspection and supplier quality improve, and treat sudden movements in either direction as investigation triggers: a falling defect rate can mean better production or worse reporting discipline.
Which metric should we fix first when several are red?+
The one that compounds: exceptions. A rising exception rate feeds refunds, support load, WISMO share and reviews, so stabilizing it usually drags the others toward their bands. Stockouts on hero SKUs are the close second, since that revenue is unrecoverable rather than delayed.
