How do you forecast demand for a new product with no sales history?

A new product with no sales history is forecast from borrowed evidence instead of its own. The three inputs that hold up best are an analogue SKU you already sell, the retailer's own commitment in doors and facings, and the first real weeks of sell-through once the product ships. Each one is trustworthy for a limited stretch of the launch, so the job is to know which input you are relying on this week and to hand over to real sales as soon as the sell-through shows a repeatable pattern.

What is an analogue SKU in new product forecasting?

An analogue SKU is an existing product whose launch looked like the one you are planning: same category, similar price, same channels, same customer, and similar launch support. Its first weeks of sales become the starting shape for the new product's forecast, adjusted for the differences you can name. A good analogue is one you launched yourself, recently, into the same retailers, because a competitor's product or a launch from years ago carries conditions you cannot see.

Is a retailer's opening order a demand forecast?

A retailer's opening order is a pipeline fill, not a forecast of consumer demand. The opening order stocks shelves and distribution centres across every door at once, so it arrives as one large order that will not repeat at that size. The number worth planning against is the reorder rate after the shelves are full, which comes from doors, facings and units sold per door per week. Treating the opening order as a run rate is how brands overbuy the second production run.

When should you stop using the analogue and trust real sales data?

Real sales data should replace the analogue once the new product has sold long enough to show a pattern that survives the launch spike, which usually means several weeks of steady sales after existing customers and early promotion have washed through. Before that point, early sales mostly measure your launch marketing. The practical test is whether the most recent weeks agree with each other; when they do, and they disagree with the analogue, the real weeks win.

How big should the first order for a new product be?

The first order for a new product should cover the low end of your forecast through the date a second order could arrive, not the whole launch season. Supplier lead time sets that date, and minimum order quantity sets the smallest order you can place. When the MOQ forces a larger first order than the low case needs, the extra units are a decision about carrying cost and shelf life to make on purpose, not a forecast to defend.

How to Forecast Demand for New Products

To forecast demand for a new product, borrow evidence until you have your own. Start from an analogue SKU you already sell, scale it with the retailer's commitment in doors and facings, and hand over to real sell-through once the launch spike has passed. Know which input you are relying on each week, and when it stops being trustworthy.

Why can't you forecast a new product the normal way?

A statistical forecast fits a model to a product's own sales history. A new SKU has none, so there is nothing to fit. Every method for new product forecasting is a way of borrowing evidence from somewhere else and then replacing it as quickly as the real sales allow. How the normal version works is covered in what demand forecasting is.

That makes the question less "which method is most accurate" and more "which input is least wrong this week". The answer changes as the launch goes on:

Input	What it tells you	Trust it for	Stops being trustworthy when
Analogue SKU	The shape of the launch curve	The first order and the first weeks	Real sales settle and disagree with it
Retailer commitment	The scale: doors × facings × rate of sale	Sizing the retail pipeline fill	The opening order ships and reorders start
Pre-launch signals	Whether interest is above or below the analogue	Nudging the first order up or down	The product ships and real orders exist
Top-down market sizing	Whether the plan is plausible at all	A sanity check on the other three	Always. It is never an order quantity
Early sell-through	What customers actually do	Everything after the launch spike	Never, once it has settled

The rest of this page takes the three that carry weight in order: the analogue, the retailer, and the first real weeks.

How do you forecast from an analogue SKU?

An analogue is an existing product whose launch looked like this one. Its early sales become the starting shape of the new forecast. It is the most useful single input for a new SKU in a line you already sell, and the most commonly misused.

Pick the analogue on four conditions

Same category and format. A new flavour of a bar you already sell is a strong analogue. A new format, such as a powder next to a line of bars, is a weak one.
Similar price. Price changes how fast a product turns, even inside one category.
Same channels and retailers. A DTC-only launch and a launch into a national grocery chain have different shapes. Use an analogue that launched where this one will.
Similar launch support. A launch with a paid campaign and a launch that went out in the newsletter are different launches, whatever the product.

Prefer one of your own launches from the last couple of years to a competitor's product. You know what happened around your own launch. You do not know what happened around theirs.

Adjust for the differences you can name, and only those

For example, suppose your last two flavour launches sold 400 and 520 units a week across DTC and Amazon once they settled, and the new flavour launches into the same channels at the same price with similar support. A starting forecast near 450 units a week is defensible. If the new flavour also goes into a retailer the analogues never had, forecast that retailer separately from its own commitment rather than stretching the analogue to cover it.

Resist adjustments you cannot defend. "The category has grown" and "this flavour tested better" are both real, and both are the kind of adjustment that turns an analogue into the number you wanted.

Where the analogue breaks

The analogue stops being trustworthy the moment the new product's settled sales disagree with it. It also breaks early if the new SKU takes sales from the products you based it on. That split between new demand and moved demand is its own problem, covered in demand cannibalization at a CPG launch.

How do you use the retailer's commitment?

For a retail launch, the retailer gives you the most concrete scale you will get before the product ships: how many doors, how many facings, and the rate of sale the buyer expects. Multiply them and you have a unit rate you can plan against.

Weekly retail units ≈ doors × facings × units sold per facing per week

For example, a launch into 300 doors with two facings each, at one unit per facing per week, is 600 units a week of consumer demand. That number is what the retailer will need to replenish once the shelves are full.

The opening order is not the run rate

The first order a retailer places fills every shelf and every distribution centre at once. It is a pipeline fill. It arrives as one large order, it will not repeat at that size, and it tells you about the retailer's stocking policy rather than about shoppers. The number that matters is the reorder that follows it.

This is the most expensive single mistake in launch planning. A brand that reads the opening order as demand buys the second production run at the same size, and the second run is the one that turns into dead stock.

When the retailer's number stops being trustworthy

The commitment is trustworthy for sizing the pipeline fill and not for much longer. Once reorders start, the retailer's expected rate of sale is replaced by the actual one. If the actual rate is well below the buyer's expectation, the risk is no longer your forecast. It is losing the facings.

When do the first real weeks take over?

Real sell-through replaces every borrowed input eventually. The skill is not handing over too early.

The first weeks after launch are distorted in predictable ways:

Existing customers buy first. On DTC, your email list and repeat buyers try the new product in the first days. That demand is real, but it does not repeat at the same rate.
Launch support ends. A paid campaign, a feature on the homepage, an end-cap at the retailer. When it stops, so does part of the rate.
Stockouts hide demand. If the product sells out in week two, weeks three and four look like low demand when they were no stock. Exclude those weeks rather than learning from them.
Retail sell-through lags. Point-of-sale data arrives later than your own shipments, so the retail picture is always behind the DTC one.

The working rule: trust the real weeks when the most recent weeks agree with each other. When they agree with each other and disagree with the analogue, the real weeks win, and the forecast should move to them in full rather than being averaged back toward the assumption.

How big should the first order be?

Forecasting gives you a range. The first order has to be one number, and the forecast is not what sets it.

Cover the low case through the date a second order could land. Supplier lead time sets that date. The first order is a bridge to the second, not a bet on the whole season.
Check the MOQ before the forecast. If the supplier's minimum is larger than the low case, the extra units are a carrying-cost decision to make on purpose. What MOQ means and how to decide on one covers that trade.
Hold safety stock against the uncertainty you actually have. A new SKU's uncertainty is higher than a mature one's, so the buffer should be sized to that, and reduced as real sales arrive. Safety stock covers the sizing.
Know your shelf life. On food, beverage and supplements, the upside case can expire on the shelf. The upper limit on the first order is often a date, not a quantity.

Then decide the second order before the first one ships: what sell-through would make you order more, what would make you wait, and the latest date you can place it.

Where Planster fits, and where it doesn't

Planster pulls sales and inventory from 150+ systems, fits a statistical model to each SKU's own history, and builds one demand forecast across DTC, Amazon, retail and wholesale. It nets that forecast against what is on hand, on order and already committed, and applies lead times and MOQs to work out what to order and when. That is the engine.

For a launch, the engine has nothing to fit, and it says so. So you set a manual plan for the new SKU from your analogue and keep it next to the statistical forecast rather than on top of it. Once real sales arrive and the forecast starts disagreeing with your assumption, you can see the gap. For retail launches, Pipeline puts the deal in the plan at full volume and shows the date you have to order by to hit the launch. Master Plan shows what a bigger first order commits before you place it. Each morning the agent brings you the short list, with the purchase orders already drafted. Every order still needs you to approve it. Plan, the full app, is $999/month, for consumable CPG brands, manufacturers and co-manufacturers between $50M and $500M.

Where it does not help: Planster does not pick your analogue or tell you what a new product will sell. No system can do that from sales history the product does not have. That first number is still your judgement. What the plan does is hold that judgement explicitly, run it through the same ordering maths as every other SKU, and tell you when real sales stop agreeing with it.