RSKRisk, cost & sustainabilityStop 3 of 9
Service level policy
Setting service targets by segment, based on what a stockout really costs, instead of one blanket number for everything.
Read atChange any time
“We target 98% service on everything” sounds ambitious and simple. In practice it overstocks items where a stockout barely matters and may still under-protect items where a stockout is a disaster. A service level policy sets targets deliberately, by segment.
The economics behind it: each extra point of service costs more inventory than the one before. Going from 95% to 99% cycle service raises safety stock by about 41%, and pushing towards 99.9% nearly doubles it. So the right target depends on the cost of the alternative:
- High targets where stockouts are expensive: items that stop a customer’s production line, key accounts with penalties, high-margin hero products, critical medicines.
- Moderate targets for regular items with reasonable substitutes.
- Lower targets for cheap, slow or easily substituted items, or ones customers will wait for.
A practical policy combines product segments (ABC-XYZ) with customer tiers, for example 98% for A items to key accounts, 95% for B items, 90% for slow C items. Plotting total inventory against overall service for different policies gives an efficient frontier: the best service achievable for each level of inventory investment. Leadership then picks a point on the curve knowingly, instead of guessing.
The policy should be reviewed as costs, margins and customer needs change.
Words to use
- Differentiated service
- Different service targets for different products or customer groups.
- Stockout cost
- What a shortage really costs: lost margin, penalties, a stopped line or lost customers.
- Efficient frontier
- The best service achievable for each level of inventory investment.
Same average service, less inventory
Two items with identical demand and identical uncertainty over the lead time ( = 100 units). Item A costs ₹1,000; item C costs ₹50. Safety stock = .
| Policy | A target | C target | Safety stock value |
|---|---|---|---|
| Uniform | 97% | 97% | ₹1,88,079 + ₹9,404 = ₹1,97,483 |
| Differentiated | 95% | 99% | ₹1,64,485 + ₹11,632 = ₹1,76,117 |
Both policies average 97% cycle service across the two items, but the differentiated one costs ₹21,366 (11%) less, because extra service is cheap on the cheap item and expensive on the expensive one.
Across thousands of items, this logic is applied with an optimization: maximize the overall service measure for a given inventory budget, or minimize inventory for a target, by adding safety stock wherever it buys the most service per rupee.
The efficient frontier
Solving that problem for different budgets traces the efficient frontier: the best service achievable for each level of inventory. Today’s position is usually below the frontier, so the first step is often to reach the frontier (same service, less stock, or more service, same stock) before deciding where on it to sit.
Writing the policy
A practical service policy states:
- the measure (cycle service, fill rate or OTIF) for each segment,
- targets by product class and customer tier, for example 98% for A items to key accounts, 95% for regular items, 90% for slow C items unless they are critical,
- exceptions: critical items, contractual commitments, launches,
- who owns the targets and how often they are reviewed.
What does a stockout cost?
Cost-based targets need an estimate of the cost of being short, which differs by situation:
- Customer waits (backorder): expediting, a discount or penalty, some goodwill.
- Customer buys a substitute from you: little lost margin, possibly a gain or loss in mix.
- Customer buys elsewhere this time: the lost margin on that sale.
- Customer leaves for good: the lifetime value of the relationship.
- Contract penalties: OTIF fines or service credits.
Data on substitution and repeat purchase after stockouts, plus contract terms, turns these into numbers. Even rough estimates show that some items deserve much higher targets than others. With a stockout cost per unit and a holding cost, the newsvendor logic gives a target directly: the critical ratio.
Choose the measure for each segment
- Cycle service level for items replenished in regular cycles where any stockout matters.
- Fill rate for items where the share of demand served matters more than the number of stockout events.
- OTIF for customers who measure complete, on-time orders.
- Availability (machines running) for service parts.
Targets that move
Targets can change with the season (higher before the festive peak for key items), with the lifecycle (higher at launch to build the brand, lower during phase-out) and with supply conditions (temporarily lower when supply is constrained, with allocation rules).
Governance
Service targets are a joint decision: sales knows customer needs, finance knows the cost of inventory, supply chain knows what each target requires. A policy signed off by all three, reviewed periodically and tracked by segment, prevents both “99% on everything” and quiet cuts to service that sales discovers later.
Interview kit
Practise it out loud
Answer each question yourself before opening the model answer.
I’d set service targets by segment rather than one blanket number, because safety stock rises steeply near 100% and stockout costs differ widely. High targets go where stockouts are expensive, such as critical items, key accounts and high margins; lower ones where substitutes exist or customers will wait. Comparing policies on an inventory-versus-service curve lets leadership choose the trade-off explicitly.
Leadership wants 99% service on every SKU. How do you respond?
Show the cost: moving from 95% to 99% raises safety stock by about 41% per item, and much of it would sit on cheap or substitutable items where stockouts barely matter. Propose differentiated targets by product and customer segment, with an inventory-versus-service curve showing the overall service and investment of each option.
What information do you need to set a service target for an item?
Its margin and the cost of a stockout (lost sale, penalty, a stopped customer line, damage to the relationship), the holding cost of the extra stock, demand variability and lead time, and whether customers accept substitutes or backorders.
Connected stops
Where this idea feeds in, or gets its inputs from.
Sources and further reading
- Silver, E. A., Pyke, D. F. & Thomas, D. J. (2017). Inventory and Production Management in Supply Chains (4th ed.). CRC Press.
- Chopra, S. & Meindl, P. Supply Chain Management: Strategy, Planning, and Operation. Pearson. Chapter on determining the optimal level of product availability.
- Teunter, R. H., Babai, M. Z., & Syntetos, A. A. (2010). ABC classification: Service levels and inventory costs. Production and Operations Management, 19(3), 343–352.