There is a formula for safety stock. You may have learned it at university, or found it in a textbook, or inherited it in a spreadsheet from someone who left the business in 2019. It looks authoritative. It has a square root in it. It feels like maths, and maths feels like rigour.

The formula is not wrong, exactly. It is just quietly assuming things about your business that have not been true for some time, if they ever were.

What the textbook assumes

The standard safety stock calculation makes three assumptions.

First, that demand follows a normal distribution. In practice, most demand patterns are lumpy. A handful of large orders from key accounts create spikes that a normal distribution handles poorly. If your top three customers represent 40% of revenue, your demand is not normally distributed. It is event-driven, and your safety stock calculation should reflect that. (And if you are running a forecast, the real question is not how variable demand is but how wrong the forecast is, which is a different calculation entirely.)

Second, that lead times are stable and predictable. Anyone who has sourced from East Asia, Southern Europe, or anywhere with a port involved will find this assumption generous. Lead time variability is frequently the dominant factor in safety stock calculations, yet it is the one most often entered as a single average number rather than a range. An average lead time of 42 days is meaningless if the actual range is 28 to 65.

Third, that every SKU deserves the same service level target. The formula asks you to pick a service level (95%, 97.5%, 99%) and applies it uniformly. But a 99% service level on a slow-moving C-class item with a six-month shelf life is a very different commercial decision from 99% on your top-selling A-class product. Applying the same target everywhere is not rigorous. It is expensive.

Where the real damage happens

The formula itself is secondary to the conversation around it, or more precisely, the absence of one.

In most businesses, safety stock is set once and then left alone. Someone ran the numbers when the ERP was implemented, or when a new product launched, and the parameters have been sitting in the system ever since. Demand has shifted. Suppliers have changed. Lead times have drifted. But the safety stock sits there like a geological feature, quietly costing money.

The second common failure is treating safety stock as a lever rather than a system output. Safety stock is not an independent variable. It is a consequence of demand variability (or forecast accuracy), supply variability, and service level ambition. In order to reduce safety stock, the answer is not to override the formula. It is to reduce the variability that feeds it. Shorter lead times, more reliable suppliers, better demand visibility: these are the actual levers. The safety stock number is the scoreboard, not the game.

The third, and perhaps most corrosive, failure is political. Safety stock becomes a negotiation between sales (who need everything available immediately) and finance (who need inventory as low as possible). Without a shared framework, this negotiation produces arbitrary compromises that satisfy nobody and optimise nothing.

And then there is the cardinal sin, which is so common it qualifies more as a rite of passage.

A safety stock setting from three years ago, still sitting in the system on a product that nobody is really selling anymore. Multiply that across a tail of dying SKUs and you have serious working capital tied up in inventory that nobody asked for and nobody needs.

If you have never found one of these lurking in your system, it is only a matter of time.

A more sustainable approach

None of this requires sophisticated software or a dedicated inventory planner. What it does require is a willingness to interrogate the inputs rather than trust the output.

Start with an ABC segmentation that reflects actual commercial reality, not just volume. The A-class items are not necessarily your highest sellers. They are the ones where a stockout has the most painful consequences, whether that is a lost key account order, a production line stoppage, or a contractual penalty. Segment by consequence, not just by revenue.

Then look at the demand and lead time data honestly. Not the averages. The ranges. If you have 24 months of purchase order history, you have enough data to calculate a meaningful standard deviation for lead time per supplier. If the spread is wide, that is not a data problem. That is a supplier performance problem, and it should be managed as one.

Set differentiated service levels by segment. Your A-class items might warrant 98%. Your C-class items might be perfectly well served at 90%, or even managed to order rather than stocked at all. The point is to make the trade-off consciously rather than by default.

Finally, review quarterly. Not annually, not never. Quarterly. Safety stock parameters should be living numbers that reflect current conditions. A fifteen-minute review of your top 20 SKUs each quarter will do more for your inventory health than any annual deep-dive. It is also the most reliable way to catch those forgotten settings on dying products before they quietly eat your warehouse.

The honest truth about the formula

The safety stock formula is just a tool. It gives you exactly the answer you deserve based on the quality of what you feed it. If you feed it averages, assumptions, and parameters from three years ago, it will return a number that feels precise and is almost certainly wrong.

The businesses that manage inventory well are not the ones with the best formula. They are the ones that treat safety stock as a living decision, revisited regularly, segmented thoughtfully, and grounded in data they have actually looked at recently.

The square root is the easy part.