Every buyer has a method for this. Most methods are reasonable and most are missing the same thing.
What the standard method looks like
Ask ten buyers at mid-size distributors how they size a first order for a new item and you will hear roughly one answer. Pull up two or three comparable products from the same category or the same supplier. Look at what they sold in their first months. Ask the sales reps who cover the accounts most likely to buy it. Add ten or fifteen percent because launches get a push. Round up to the carton or pallet size, because the supplier will round it up anyway.
This is not a bad method. Comparable products genuinely are the best evidence available for an item that has never been sold, and rounding to a valid order size is not a compromise — it is the actual constraint. Surveys of planning practice keep finding the same thing: most of this work happens in spreadsheets, and the dedicated inventory tools in the market are built for replenishing existing items, not for the one decision where there is no history at all.
What it leaves out
The gap is not accuracy. It is that the method produces a single number, and a single number carries no information about how wrong it might be. "We expect to sell about 600 in the first quarter" describes a world with one outcome. The real world contains a version where a large contractor standardises on the item and you sell 1,400, and a version where the specification slips a season and you sell 180. Those two versions have wildly different consequences for the same order quantity, and nothing in the standard method distinguishes between them.
The second omission follows from the first: without a spread you cannot weigh the two ways of being wrong against each other. Ordering too few and ordering too many are not symmetric mistakes, and in distribution they are not even close. Under-ordering costs the gross margin you did not earn, and it costs it at exactly the moment when interest in the product is highest. Over-ordering costs carrying charges — storage, handling, shrinkage, damage, and the capital tied up — which run to roughly a quarter of the inventory's value per year for a typical distributor, and it ends in a liquidation that recovers somewhere between ten and forty percent of cost.
The three questions that replace the buffer
First: what period does this order actually have to cover? Not "the launch" and not a round six months. The first buy has to last until a replenishment order can physically arrive on your shelf — the supplier's lead time plus however long it takes you to notice and place the order. For an item with a ten-week lead time reviewed monthly, that is fourteen weeks, and demand beyond week fourteen is somebody else's problem, because by then you can reorder. Sizing the first buy against a six-month expectation when you can replenish in fourteen weeks is one of the most common and most expensive mistakes in new-item buying.
Second: what does the range of plausible demand look like over that period? Take the comparable products you already pulled up, and instead of averaging them, look at the spread between them. If three similar items sold 320, 540 and 890 units in their first fourteen weeks, your honest input is that range, not its average of 583. The width of that spread is information, and it is the input that determines how much cover you should buy.
Third: what does each kind of mistake cost you? Write down the gross margin per unit, and write down what a unit that never sells actually costs after you have carried it for a year and liquidated it. The ratio between those two numbers tells you how much of the demand range to cover. High margin and recoverable stock means you should cover the optimistic end and expect leftovers. Thin margin and stock that gets written off means covering the middle of the range and accepting that you will sometimes run out.
Why the answer is a quantile, not an average
Once you have a range and two costs, the arithmetic has a known answer, and it is over a century old: order enough to cover demand up to a specific point in its distribution, determined by the ratio of the two costs. Cover too little and the expected margin you forgo exceeds the carrying cost you avoid; cover too much and the reverse is true. The crossing point is the optimum. It is almost never the average demand, and the direction it moves away from the average tells you something useful about your own product. We wrote this up in plain language in the newsvendor model, explained for buyers.
One property of that curve deserves emphasis, because it is genuinely reassuring: it is flat near the top. Being a hundred units off the optimum costs very little. Being at half or double the optimum costs a great deal. The purpose of doing this properly is not to hit an exact number — it is to avoid being in the wrong neighbourhood, which the intuitive buffer method does regularly and silently.
Two things that quietly break the method
Your comparable products are censored. If one of the items you are comparing against went out of stock in its second month, its sales curve shows what you shipped, not what customers wanted. It understates true demand, and it usually does so precisely because its own first buy was too small. Averaging that curve into your new estimate repeats the mistake. This is the single most under-appreciated issue in new-item planning and we gave it its own guide.
The MOQ may make the whole question moot. If your analysis says 400 and the supplier's minimum is 1,000, the exercise is no longer about sizing — it is about whether to launch at all, and what to say in the negotiation. The useful output there is the sell-through you would need to break even and how likely that is; the MOQ Risk Calculator gives both.
Doing it in five minutes
None of this requires new software or a data science team. It requires keeping the spread instead of collapsing it, being honest about the two costs, and measuring against the replenishment horizon rather than a round number. The Initial Order Quantity Calculator does exactly that from a handful of inputs, and FirstBuy AI will do it from your own launch history if you have a few years of it in a spreadsheet — which most buyers do, in the export they already run for the annual line review.