How do you forecast demand for a new product?

Neon Blue builds simulations of human behavior. You forecast demand for a new product by building it up from parts you can estimate: how many people in the target market will become aware of it, how many of those will try it, how many will buy again, and how often. With no sales history, each part comes from research, analog products or both. A diffusion curve then spreads first purchases over time. The forecast is only as good as its weakest input, so the work is in pinning down trial and repeat for the right buyers before launch.

What goes into a new product demand forecast?

Get the target buyer count first. The target audience guide covers sizing the segments the forecast is built on.

How do diffusion curves help forecast new product demand?

The Bass diffusion model, from 1969, splits adopters into two groups: people who adopt on their own (driven by the coefficient of innovation, p) and people who adopt because others have (the coefficient of imitation, q). In 1990, Fareena Sultan, John Farley and Donald Lehmann published a meta-analysis of 213 applications of diffusion models in the Journal of Marketing Research. They found average values of about 0.03 for p and 0.38 for q, with wide variation by product and country.

Those averages are useful as a starting point when you have nothing else: a product with typical word of mouth reaches its peak sales years, not months, after launch. They are a poor substitute for your own inputs. Research on the target audience tells you whether your product looks more like a fast-spreading or a slow-spreading one. The way the launch decision uses this curve is covered in the go or no-go guide.

What questions feed a demand forecast?

1. How likely would you be to buy this product within the next four weeks if it were available where you shop, at [price]?
2. If you tried it and it worked as described, how often would you buy it?
3. What would you buy less of if you bought this?
4. How did you first hear about the last new product you tried in this category? [Friend, ad, store shelf, online review, other]
5. Would you recommend it to someone you know? Who?

Questions 1 to 3 feed trial, frequency and source of volume. Questions 4 and 5 hint at how much spread comes from word of mouth, which helps choose between fast and slow diffusion assumptions. The wording of question 1 and how to read it are set out in measuring purchase intent.

What does a forecast input study look like on a simulated audience?

Hypothetical: a home goods company is planning a smart water filter pitcher that reorders cartridges by subscription.

• Audience: US households that use a water filter or buy bottled water weekly, split by filter owners versus bottled-water buyers, renters versus owners, and households with kids.
• Design: concept and price, the five questions above, and a cartridge subscription question, fielded on a simulated audience, with two price cells drawn from a Van Westendorp range.
• Readout: adjusted trial and repeat inputs by segment, source of volume, and word-of-mouth signals.

Example output layout (illustrative, not a real result)

The finance team takes the adjusted inputs into the volume build. If bottled-water buyers show high trial but low subscription uptake, the forecast splits pitcher sales from cartridge revenue rather than assuming both.

How do you make a new product forecast more reliable?

• Adjust stated intent. Apply a discount to top-box intent, based on your own past launches if you have them.
• Use analogs. Find two or three past launches in a similar category and check your curve against theirs.
• Run scenarios. Low, base and high forecasts, each with stated assumptions. Leadership can then see which input moves the answer most.
• Close the loop with a market test. A regional or simulated test market narrows the range before a national commitment, and early sales let you refit the curve.

Teams who need the full research sequence before a forecast can start with testing a product idea. Pricing inputs come from pricing research.

Questions about forecasting new product demand

How do you forecast sales with no historical data?

Build volume from target buyers, awareness, trial, repeat and frequency, using research and analog products for each input, then check the result against similar past launches.

What is the best method for new product demand forecasting?

For packaged and consumer goods, trial and repeat build-ups calibrated with research. For durables and technology, diffusion models with research-based or analog coefficients. Many teams use both.

How far off can a new product forecast be?

They carry wide ranges. Present low, base and high cases and update as soon as early sales arrive.

How does Neon Blue help with demand forecasting?

Neon Blue supplies forecast inputs, trial, repeat, frequency and source of volume by segment, from a simulated audience of your target buyers, all marked as simulated, for your team's volume model. Start a free first study.

More on launch planning

• How do you test a product idea before you launch it?
• What questions should you ask in a concept test?
• Synthetic panels vs traditional panels: which should you use?
• Marketing and product teams: testing ideas before launch