The newsvendor model is a fundamental framework in operations management that helps you decide how much inventory to order when you only get one chance to order and you don’t know what demand will be. It gets its name from a newspaper vendor who must decide how many papers to buy each morning before knowing how many customers will show up. Order too many and you’re stuck with unsold papers. Order too few and you miss out on sales. The model gives you a precise way to find the order quantity that minimizes total expected cost.
The Core Trade-Off
Every newsvendor decision comes down to two competing risks: ordering too much and ordering too little. The model assigns a specific cost to each.
The overage cost is what you lose on every extra unit you ordered but couldn’t sell. It equals the cost you paid minus whatever you can recover by discounting, returning, or salvaging the item. If you buy holiday ornaments at $55 each and can sell the leftovers at half price ($40) in January, your overage cost is $15 per unsold ornament.
The underage cost is the profit you forfeited on every unit a customer wanted but you didn’t have in stock. It equals the selling price minus your cost. For those same ornaments retailing at $80, the underage cost is $25 per ornament you could have sold but didn’t order.
These two costs are almost never equal, and that asymmetry is the entire point. When the cost of ordering too few is much higher than the cost of ordering too many, you should lean toward ordering more, and vice versa. The model tells you exactly how far to lean.
How the Critical Fractile Works
The newsvendor model finds the optimal order quantity using a ratio called the critical fractile (sometimes called the critical ratio). It looks like this:
Critical fractile = underage cost / (underage cost + overage cost)
This ratio always falls between 0 and 1, and it represents the probability of demand being at or below your optimal order quantity. In other words, it tells you how much of the demand distribution you should aim to cover.
Take those holiday ornaments: the underage cost is $25 and the overage cost is $15. The critical fractile is 25 / (25 + 15) = 0.625. You should order enough ornaments to satisfy demand about 62.5% of the time. That means you’re accepting a roughly 37.5% chance of having leftovers, because the penalty for under-ordering outweighs the penalty for over-ordering.
A more dramatic example: a newsstand buys academic journals for $1.00, sells them for $4.00, and can return unsold copies to the publisher for $0.50. The underage cost is $3.00 and the overage cost is $0.50. The critical fractile is 3.00 / 3.50 = 0.857. The newsstand should stock enough to cover demand roughly 86% of the time, because the lost profit on a missed sale dwarfs the cost of a returned copy.
Turning the Ratio Into an Order Quantity
Once you have the critical fractile, you need a demand forecast to convert it into an actual number of units. The model treats demand as a random variable with a known (or estimated) probability distribution. If you assume demand follows a normal distribution, you need two inputs: the average expected demand and the standard deviation (a measure of how much demand fluctuates).
You then find the point on the demand distribution where the cumulative probability equals your critical fractile. With normal demand, this means looking up or calculating the corresponding z-score and plugging it into: optimal order quantity = mean demand + (z-score × standard deviation). A critical fractile above 0.5 pushes the order quantity above the mean, building in a buffer. A critical fractile below 0.5 pulls it below the mean, deliberately planning to run short sometimes.
The normal distribution assumption is common and works well when demand comes from many independent customers, since aggregating lots of small, independent purchases tends to produce a bell-shaped pattern. But the model works with any demand distribution as long as you can identify the point where the cumulative probability matches the critical fractile.
What Salvage Value Does to the Math
Salvage value is the amount you can recover per unsold unit, whether by discounting, selling on a secondary market, or returning to the supplier. It directly reduces your overage cost. The higher the salvage value, the lower the penalty for over-ordering, which pushes the critical fractile up and encourages larger orders.
Consider a shoe company that buys pairs at $40, sells them at $60, and can clear leftovers at $30. The overage cost is $40 minus $30 = $10 per pair, and the underage cost is $60 minus $40 = $20 per pair. The critical fractile is 20 / 30 = 0.667. But if the clearance price dropped to $10, the overage cost would jump to $30, the critical fractile would fall to 20 / 50 = 0.40, and the company would order fewer pairs because the sting of excess inventory got much worse.
Key Assumptions and Limitations
The newsvendor model rests on a few simplifying assumptions that are important to understand:
- Single period. You place one order for one selling season. There’s no chance to reorder midway through. This fits products with long lead times or short selling windows: holiday merchandise, fashion collections, concert T-shirts, perishable food.
- One order before demand is known. You commit to a quantity before you observe any real customer behavior. The entire problem exists because of this timing mismatch.
- Linear costs. Each extra unit of overstock costs the same amount, and each extra unit of unmet demand costs the same amount. There are no volume discounts or escalating penalties built into the basic model.
- Known demand distribution. You don’t know exact demand, but you have a reasonable estimate of its average and variability. The quality of your answer depends heavily on the quality of this estimate.
In practice, these assumptions rarely hold perfectly. Demand distributions are estimated from limited data and can be wrong. Costs may not be perfectly linear. Customers who can’t find what they want may come back later or may never return, and the true cost of lost goodwill is hard to pin down. Still, the model provides a disciplined starting point that consistently outperforms gut-feel ordering.
Where the Model Gets Used
The classic example is a newspaper vendor, but the model applies far beyond newsstands. Any business placing a one-time order before uncertain demand faces a newsvendor problem. Fashion retailers ordering seasonal collections months before they hit stores. Bakeries deciding how many loaves to bake before the morning rush. Event organizers printing a fixed number of programs. Vaccine manufacturers committing to production quantities before flu season.
The logic also extends beyond physical inventory. At its core, the model captures any situation where you must make a firm commitment before a random outcome is revealed, and the penalties for overshooting and undershooting differ. Booking hotel room blocks, reserving server capacity, and setting staffing levels for a shift all share this structure. The math is identical: estimate the cost of having too much, estimate the cost of having too little, compute the critical fractile, and use your best demand forecast to pick a quantity.