Forecasting new products without sales history: analogues, category rates and early reads

Every forecasting method in the textbook starts from history. A new product has none. Yet the first purchase order for a launch is often the biggest bet you'll make on that product, placed with the least information.

You can't make the uncertainty go away, but you can make a disciplined first estimate, update it quickly once sales start, and size the first order so that being wrong in either direction is survivable. This guide covers all three.

Key takeaways

• Start from analogues: similar past launches, scaled for the differences you know about.
• Cross-check with a top-down category estimate, and remember that some of the new product's sales will come from your existing range.
• Read early sales against how analogues sold in the same launch weeks, not against their steady-state rate.
• Size the first order to last until you can act on a reliable read, plus the reorder lead time.
• Buy above or below the central estimate depending on the cost of running short versus the cost of leftovers.

Method 1: analogues

An analogue is a past product you expect to behave like the new one. For a new candle scent, the best analogues are your previous scent launches. For a new colourway, previous colourway launches in the same style.

Suppose you've launched three scents before. Their first eight weeks:

The starting forecast for the new scent is 40 units per week over its first eight weeks, with a plausible range of roughly 30 to 50 based on what the analogues actually did.

Then adjust for known differences — but only the ones you can justify:

• Marketing support. If the new launch gets an email to the full list and the others didn't, it will likely start faster.
• Price. A higher price point usually means fewer units.
• Distribution. Launching in two stores rather than one, or on a marketplace as well.

Write each adjustment down with its reason. When you review the launch later, you'll want to know which assumptions were right.

Pick analogues for similarity of customer and channel, not similarity of product. A new scent sold to your existing candle buyers through your own store is a better analogue for another scent than a competitor's best-seller is.

Method 2: category rates

A top-down cross-check uses the category instead of individual products. If your candle category sells 600 units a week and you expect the new scent to take 8% of the category's sales once established:

new scent = 600 × 8% = 48 units per week

That's in the same range as the analogue estimate of 40, which is reassuring. If the two methods disagree badly, find out why before ordering.

The category view also forces you to think about cannibalisation. If half the new scent's sales come from customers who would otherwise have bought another scent, the category only grows by:

48 × 50% = 24 units per week

and your existing scents lose 24 units per week between them. If you don't lower their forecasts, you'll over-buy the existing range at the same time as you're buying the new one. It's an easy mistake to miss, because each forecast looks reasonable on its own.

Category rates are also how you handle seasonality for a product with no history: a new product launched in September inherits its category's seasonal pattern, so its forecast rises into Q4 along with the category. Stokrio's demand forecasting does this automatically — new products borrow their category's seasonal pattern and rate until they have enough history of their own.

Method 3: the early read

Once the product goes live, history accumulates fast, and the question becomes: what does the first fortnight tell us about the next six weeks?

The common mistake is to compare early sales with the analogues' average weekly rate. Launches don't sell evenly. There's usually an initial burst from your existing customers, email and social, followed by a lower steady rate. Compare like with like: the new product's first two weeks against the analogues' first two weeks.

From the table, the analogues sold 90 units in weeks 1–2 out of 320 in weeks 1–8:

share of 8-week sales in weeks 1–2 = 90 ÷ 320 = 28.1%

The new scent sells 110 units in its first two weeks. Projecting that forward:

projected 8-week sales = 110 ÷ 0.281 = 391 units
remaining 6 weeks      = 391 − 110 = 281 units, or about 47 per week

So the early read lifts the forecast from 40 to about 47 per week for the rest of the period. Compare that with the naive approach of dividing 110 by 2 and getting 55 per week — which would over-forecast the remaining weeks by about 8 units a week, because it mistakes the launch burst for the steady rate.

Treat the early read with care:

• Two weeks is a small sample. Blend it with the analogue estimate rather than replacing it. Weighting the early read more each week as data builds up is a sensible approach.
• Check for stockouts. If a size or variant sold out in week one, its early sales understate demand.
• Check the marketing calendar. If this launch had a bigger email push than the analogues, its early share will be higher, and the projection will overstate the steady rate.

Sizing the first order

The forecast gives you a central estimate and a range. The first order has to decide how much of that range to fund. Start with timing:

first order must last = weeks until a reliable read + reorder lead time

If you'll have a reliable read by the end of week 3 and the supplier lead time is 6 weeks, the first order must last 9 weeks. At the analogue range of 30–50 per week, with a likely case of 40:

Which number should you order? It depends on what happens when you're wrong.

• Under-ordering costs you the margin on sales you miss.
• Over-ordering costs you whatever it takes to deal with the leftovers.

One standard way to weigh these is the critical ratio:

critical ratio = cost of being short ÷ (cost of being short + cost of being over)

The ratio tells you roughly how far up the range to buy. Two illustrative cases — substitute your own figures:

A core-range scent sells for $40, costs $12 and will keep selling indefinitely. Missing a sale costs $28 of margin. A leftover unit just sells a few months later, costing perhaps $3 in holding and capital. The ratio is 28 ÷ (28 + 3) = 0.90 — buy towards the high scenario, around 450.

A limited-edition holiday scent sells for $30 and costs $15, and leftovers in January clear at $5. Missing a sale costs $15; a leftover costs $15 − $5 = $10. The ratio is 15 ÷ (15 + 10) = 0.60 — buy a little above the likely case, around 360–380.

Same forecast, different orders, because the consequences of error differ.

When the supplier's minimum is the real constraint

Often the supplier's minimum order quantity is bigger than any of your scenarios. If the MOQ is 1,000 units, the forecast question changes from "how many should we buy?" to "how long will 1,000 units last, and can we afford that?" At the likely rate of 40 per week, 1,000 units is 25 weeks of stock. That may be fine for a core product. For a limited edition it's a reason to negotiate the minimum, split the run with another variant, or reconsider the launch.

After the launch

Once a product has eight to twelve weeks of history, its own data starts to outweigh the analogues. From there, it should graduate into your normal forecasting process. Two habits make the next launch easier:

1. Record each launch's first eight weeks, alongside the marketing support it had. That's your analogue library.
2. Compare the launch forecast with what happened. If launches consistently beat or miss forecast, you have a bias to correct — see our guide to forecast accuracy and bias.

If you plan launches in Stokrio, you can add the launch to the forecast as a fixed quantity or monthly target using events and promotions, and it flows straight into replenishment quantities.