Forecasting for a new product is one of the hardest parts of demand planning, because the usual statistical models that learn from history have no history to learn from. Yet these decisions are typically expensive: the initial order quantity, production capacity and inventory capital are all set before the first real sale is seen.
The first approach is analogy to similar products. Select one or more products in your portfolio that are similar in price, target market and sales channel, and use their first-months sales curve as a pattern. What matters is writing down the similarity criteria in advance, not choosing, after seeing the result, the analog that gives the desired output.
The second approach is a top-down or bottom-up estimate. Top-down estimates market size, target share and penetration rate. Bottom-up counts the number of points of sale, the expected rate of sale per point and the coverage plan. A large gap between the two is useful information in itself: it shows which assumptions are fragile.
The third approach is pre-launch evidence: pre-registered orders, pilot results at a few customers or cities, and feedback from sales negotiations. These signals tend to be optimistic because customers do not equate interest with purchase, so they must be adjusted by a historical conversion factor derived from your own earlier launches.
The fourth and most important approach is designing scenarios rather than a single number. Instead of one forecast, define low, base and high scenarios with approximate probabilities and build decisions around them: how much to commit now, and how much to keep with a later decision option (for example a short-lead-time second order). The aim is that if the low scenario happens the loss is limited, and if the high scenario happens you can respond.
A common mistake is staying loyal to the initial number. A new-product forecast should be updated quickly with real data in the first weeks. Three or four weeks of sales data, even incomplete, often contain more information than months of market research. Set a review checkpoint in advance and name the decision owner who can change course.
Another mistake is over-protecting with large stock. A new product has higher obsolescence risk, and if sales stay low heavy excess inventory remains. If you can reorder quickly, take a smaller initial volume and be ready to top up; if lead times are long, use flexible contracts (a minimum volume plus an option to increase).
A practical starting point: for the next launch, do three things — a scenario table (low/base/high) with explicit assumptions, one analog product with written similarity criteria, and a review meeting in week four. After launch, compare forecast to actual so you have a better adjustment factor for later launches.