In short: A uniform percentage cut applied across a network misallocates by construction, because the cost of removing a week of cover differs by item, location and lead time, so the same percentage buys very different amounts of risk in different places. The instrument that replaces a single target is a service and inventory frontier showing what each level of stock buys in availability, which turns a board number into a choice between points on a curve. Presenting inventory as a portfolio, with working, safety, cycle and dead stock separated, makes clear that only one of those can be removed without giving up service. Dead stock is the part of any target deliverable without spending service, and it usually needs a write-off allowance rather than a planning change.
The slide said inventory would come down fifteen percent by year end. It came out of a working capital review, it was reasonable at the level it was written, and it went into the board pack as a single line with a single number.
Ten weeks later the number is moving. Two categories in one region are short, the expedite freight bill has doubled, and the eleven million of slow moving stock that everyone agrees is dead is sitting in exactly the racks it occupied when the slide was written. Nothing went wrong procedurally. A percentage was set, it was cascaded, and each unit delivered its share. The allocation of that fifteen percent across the network is where the damage happened, and the allocation was never decided by anybody.
Why a uniform percentage misallocates by construction
A percentage target on a balance sheet number carries no information about service. It instructs every unit to reduce by the same proportion regardless of what that stock is doing, so the only remaining thing to optimise is how hard the reduction is to achieve. The cut lands where it is cheapest to execute, and cheapest to execute has close to nothing in common with cheapest in service terms.
Four patterns show up nearly every time.
The reduction gets taken from fast movers, because that is where the currency is. A high turning category holds a lot of value and looks like the obvious place to find fifteen percent, and it is also the category where every unit of cover is protecting the most revenue. Slow moving stock holds less value per line and takes more effort to remove, so it survives.
Dead stock survives for a specific accounting reason. Removing it means recognising a write-off, which hits the P&L this period while improving the balance sheet. A unit manager asked to reduce inventory with no corresponding P&L allowance will avoid the one category where reduction is free in service terms, because it is the only one that costs them their own number.
The reduction gets delivered by deferral rather than by structural change. Inbound orders due in the closing weeks get pushed out, the closing number lands, and the stock arrives three weeks later alongside the shortage it caused.
Political cover determines the rest of it. Units with the most credible service argument, or the most senior leader, negotiate their way to a smaller share, and the residual settles on whoever pushes back least. Each of those responses is rational for the person making it, and together they produce the allocation you would expect from asking for a proportion when nobody has priced the alternatives.
The frontier as the instrument that replaces the target
The replacement for a single number is a curve. For a given network, demand model and set of lead times, you can compute the maximum service achievable at each level of inventory investment. That relationship is concave, so the first tranche of investment buys a lot of service and later tranches buy progressively less. The calculation belongs to the buffer sizing and placement work and is covered elsewhere. What matters here is what having the curve does to the conversation.
It converts the board's question from "how much should inventory come down" into "which point on this curve do we want", and the second question has content the first does not. Every point carries a cash number and a service number, and the segment between two candidate points has a price attached: this much cash released, this much fill rate given up, at these customers.
The first finding is usually the useful one. Most networks sit inside the curve, holding more inventory than their current service level requires because the allocation across items and locations is inherited rather than chosen. Where that is true, the cut and a service improvement are available at the same time and the fifteen percent can be delivered without the argument. Where a network is already on the curve, the cut has a price and the price is quotable, which is a far better position to be in when the board asks.
The other thing a frontier gives finance is a marginal number they already know how to use. The return on the last million of inventory can be set against the return on any other million the business could deploy. If the marginal million of stock returns six percent against a hurdle rate of twelve, the reduction is correct and the board should push harder. If it returns forty percent, the target is destroying value and somebody in the room should be arguing for more inventory.
Presenting inventory as a portfolio rather than a number
The framing that works with a finance audience is one they already use elsewhere. Every unit of stock is a position taken against a forecast. It has a payoff if the demand arrives, which is the margin captured plus whatever the customer relationship is worth, and a loss if it does not, which is the carrying cost plus the probability of markdown or write-off.
Grouping the balance by segment gives each group risk and return characteristics that a single company-wide number hides. High margin fast movers are positions with a good expected return and low variance. The long tail is a set of positions with poor expected returns and high variance, which is where a portfolio manager would cut first. Pre-build against a known capacity constraint is closer to an arbitrage with a fairly certain payoff. A promotional buy is a directional bet with a date on it. Those four things deserve different treatment and a percentage target gives them the same one.
The distributional point is worth making explicitly, because it is the one that changes minds in the room. A company holding seventy days of cover on average holds twelve days on some lines and four hundred on others, with both the money and the risk concentrated in the tail. A target applied to the average is applied to a number that describes nothing in the building.
Chen, Frank and Wu (2005), in Management Science, examined what happened to the inventories of American public companies between 1981 and 2000 and looked at the relationship with long run stock returns. Firms carrying abnormally high inventory had poor long run returns, which is the result everyone expects. The part worth quoting to a board is the other end of the range: firms holding the very lowest inventories did not outperform, and the strongest long run returns sat with firms holding somewhat below average inventory. Lower is directionally right up to a point, and stops being right after it.
Four kinds of stock, and only one of them is free
Splitting the balance into its components puts the whole argument on one slide, because the category a blanket target should be attacking is rarely the category it reaches.
Cycle stock exists because you order in batches. It is a function of order frequency, minimum order quantities and container economics. It reduces by ordering more often, which trades inventory against freight cost, handling cost and whatever price break gets lost. The trade is real and quotable, so a cycle stock reduction can be presented with its cost already attached.
Safety stock exists because demand and supply vary. It reduces either by accepting a lower service level or by reducing the underlying variability, and there is no third route. The sizing method sits elsewhere and is genuinely a different piece of work. What finance needs from it is the price: this much buffer removed, this much fill rate given up, on this named list of customers.
Pre-build and strategic stock exists because somebody made a decision. A shutdown, a seasonal peak the plant cannot meet in real time, an announced price increase, a supplier transition. Reducing it means reversing the decision that created it, and that decision had an owner and a reason. It belongs in the board pack as a named item with its rationale, so that removing it is a conscious reversal rather than a cascade casualty.
Dead and excess stock exists because something went wrong: an over-order, a discontinued line, a launch that did not land, stock sitting in the wrong location for the demand that remains. This is the only category free to remove in service terms, and it is the one a percentage target reliably misses because of the write-off. Securing an explicit P&L allowance for the clearance is the highest value thing a finance partner can do for a working capital programme.
In-transit stock is a fifth line that belongs on the same slide even though planning can barely move it. It responds to sourcing origin and transport mode rather than to coverage policy, and leaving it inside the total lets a switch from air to ocean read as a planning failure.
What a defensible inventory plan looks like
A pack that survives contact with a CFO tends to have a consistent shape.
The current position decomposed into those categories, in currency and in days, with the dead stock figure stated separately and plainly rather than buried in a total. Reporting days as well as value matters more than it sounds, because a value-only number moves with cost inflation, mix and exchange rates, and a business can report a reduction in value while carrying more weeks of cover than it did last year.
The frontier with the current position marked on it, and two or three candidate points, each showing the cash, the service level and the customers affected. Where the network is inside the curve, say so, and separate the free improvement from the priced one so that nobody spends the free part twice.
The assumptions the plan depends on, listed with the direction each one moves the answer. Demand growth, supplier lead times and their spread, payment and delivery terms, and any pre-build being maintained. A plan whose assumptions were never written down cannot be reviewed when it misses, and it will be reviewed anyway, on memory.
And the downside stated at the right size. Hendricks and Singhal (2005), in Production and Operations Management, studied firms that announced supply chain disruptions and found substantial, persistent underperformance in stock returns over the following period along with an increase in equity risk. Lost margin on an unserved order is the floor on what a service failure costs. A board deciding how much buffer to remove should see the cost framed that way rather than as a line item in a fill rate report.
The limit
The frontier gives you the efficient options. Choosing among them is a different act, and the model cannot perform it.
The choice depends on things the model does not contain. How much covenant headroom exists. Whether the growth plan needs the cash this year or next. Which customer contracts are up for renewal, and what a stockout would cost inside a negotiation nobody has told planning about. How much service variability the business will tolerate as against how much it expects on average. Those are risk appetite questions and they belong to the board, which is the body that owns the covenant, the growth plan and the customer commitments. What planning owes them is honest pricing of every option on the curve.
The frontier also inherits the quality of the demand model behind it. A curve computed from a forecast that systematically understates variability will show a cheaper trade-off than the one that exists, and the error runs in the direction that makes cutting look attractive. Recomputing after each planning cycle and checking whether realised service matched the point you chose is the only honest test of it.
A curve is also a snapshot. Lead times move, suppliers change, the range changes, and a frontier six months old describes a network that has stopped existing. Treat it as a recurring calculation rather than as a study that gets commissioned.
Decompose your current inventory into those four categories this week and take the dead stock figure to your finance partner with a request for a write-off allowance, because that is the part of any target you can deliver without spending service to do it.