Home / Insights

Strategy frameworks

How to forecast a market that doesn't exist yet

A century of technology adoption curves gives owners a practical way to forecast a new market, and to keep the forecast honest as it unfolds.

By Dan Lee · November 2026

A new market has no sales history to measure. It can still be forecast. The forecast has to be built from the parts that can be known today, use history for the parts that can't, and be updated as the market reveals itself.

The warning sign comes from Clayton Christensen. In The Innovator's Dilemma, he found that disk-drive analysts forecast existing markets within single digits but missed new ones by several hundred percent, using the same methods. The lesson isn't that new markets can't be forecast. It's that you can't forecast them by asking experts for one number and extending a line.

In brief

• Every adoption forecast answers three questions: the ceiling, the takeoff and the speed.
• History gives usable ranges. The median technology took 15 years after launch to win its early adopters, then about 9 more to win the early majority.
• Lower cost and a better experience raise the ceiling, as Ford's Model T and Uber showed. New technologies take off fastest when they ride on an installed base.
• Build the forecast in scenarios, name the first customers in each adopter segment, and let leading indicators decide the next investment.

Every adoption forecast is three questions

Any forecast for a new product or technology comes down to three questions:

1. The ceiling. How many customers could ever buy it, at what price, and how much would each one spend?
2. The takeoff. How long until it wins the first customers willing to take a chance on it?
3. The speed. Once the mainstream starts buying, how many years until it reaches half of the ceiling, and then most of it?

The ceiling can usually be built from data that exists today. Takeoff and speed are where forecasts go wrong, and that is where history helps most.

Who adopts, in what order

In Diffusion of Innovations (1962), Everett Rogers sorted adopters into five groups, each with its own reason to buy:

The segments matter for a forecast because each group needs different proof. Geoffrey Moore's Crossing the Chasm (1991) argued that the hardest step is the one between early adopters and the early majority. The early majority does not buy on vision. It buys on references from people like them.

What a century of adoption curves shows

Here are thirteen technologies that changed how Americans live, plotted against Rogers' segments:

Measured segment by segment, the history is consistent:

Six lessons for anyone building a forecast:

• Winning the early adopters is the slowest phase. The median technology took 15 years from commercial launch to reach 16% of households, the end of the early adopters. For technologies launched after 1950 it was 13. The range runs from 6 years (radio) to 35 (air conditioning). That long early stretch is Moore's chasm showing up in the data. A plan that needs mainstream volume in year two is betting against a century of history.
• The early majority comes fastest. Once a technology passed 16%, the median took only 9 more years to reach 50%, and those that passed 16% after 1980 took a median of 6. Social media did it in 3. When references exist, the middle of the market moves quickly, so capacity has to be ready before the evidence is overwhelming.
• The late majority is slower, and the ceiling may be below 100%. Getting from 50% to 84% took a median of about 12 years. The automobile took 50, with the Depression and a world war in between. Washing machines, home computers and social media never reached 84% in the data. Model a slow finish, and test whether your ceiling really is "everyone."
• Price and experience move the ceiling. A ceiling is not fixed. It depends on what the product costs and how easy it is to use, and both can change.
  • Cost down, demand up. Henry Ford set a price the next group of buyers could afford, then drove cost down to meet it. The Model T touring car sold for $850 in 1908. By the mid-1920s a Model T cost under $300, and Ford built about 2 million cars in 1923. Ford's share of U.S. car production went from about 15% in 1908 to 57% in 1923. In the chart above, the automobile went from 16% of households to 50% in nine years, the Model T years. Ford's own view was that a market that looks saturated is often just priced above what the next buyer can pay.
  • A better experience creates demand that didn't exist. In 2014, valuation professor Aswath Damodaran sized Uber against a global taxi market of about $100 billion and valued the company at about $6 billion. Bill Gurley, an Uber investor and board member, argued in "How to Miss by a Mile" that the relevant market was not taxi rides but how people get around, including car ownership. Gurley turned out to be closer. Uber went public in 2019 at a valuation more than ten times Damodaran's figure. In New York, app-based ride services alone averaged about 701,000 trips a day in 2019, well above the roughly 464,000 daily yellow-cab trips of 2010. A ride you could summon from your phone, at a known price, created trips that would never have been taxi trips.
• New markets come out of old ones. Landline telephones peaked at 95% of households in 2002 and had fallen to 42% by 2019, while cell phones reached 96%. The ceiling for the new technology was largely the market for the old one.
• New technologies ride on the ones already in place. The fastest takeoffs came when the hardware and habits were already installed. The internet passed 16% of households in 1996, 7 years after commercial access began, when a third of households already owned a computer. Social media passed 16% of adults in 2008, when most households were already online and had a cell phone, then reached 50% three years later. Neither had to wait for its own infrastructure. Each was adopted on top of a base someone else had already paid for.

These are consumer curves. Industrial and B2B adoption usually moves more slowly, because qualification cycles, capital budgets and program awards set the pace. Adjust the analogs for that.

How to build the forecast

1. Build the ceiling from the market it replaces or serves. Most "new" markets have an existing base underneath: the vehicles, machines, plants or processes the new product would go into. Count that base and estimate content or spend per unit. This is a bottom-up sizing exercise, done in exactly the same way as for an existing niche market. Then make the ceiling depend on price and experience. Estimate how many more buyers come in at each lower price point. Model your cost falling as volume grows. Ask what the product makes possible that the old way did not. A forecast sized only as a share of the market it replaces will miss what Ford and Uber added. If the product depends on an enabling technology (connected equipment for a software service, installed chargers for an EV component), the enabler's installed base is the practical ceiling. Forecast the enabler first.

2. Name the takeoff trigger. Adoption rarely takes off on enthusiasm. It needs a trigger: cost reaching parity with the incumbent, the necessary infrastructure being in place, a qualification passed, or a regulation taking effect. Put a date range on when each trigger is likely to be met. That range is when you cross from early adopters to the early majority. Check it against the analogs: if your plan gets past the first 16% of its ceiling much faster than the 13 to 15 years history suggests, the trigger has to explain why. The most common good explanation is an installed base the product can ride on, the way the internet rode on home computers.

3. Name the customers in each segment. In an industrial market the segments are not anonymous. The innovators are the one or two customers who will co-develop and run a pilot. The early adopters are the companies that win something specific by moving first. The early majority will buy once those first customers can show results. Write down who they are and what proof each group needs. That list is your sales plan and the most direct test of your takeoff assumption.

4. Choose analogs for the speed. Pick past technologies that share the features that matter: who the buyer is, how much capital adoption requires, whether it depends on other infrastructure, and whether one customer adopting makes it more useful to the next. The analogs give you a slow, base and fast speed. Marketing researchers formalized this as the Bass diffusion model. A simple S-curve calibrated to analogs gets most of the value.

5. Forecast in scenarios, not a single line. Combine the ceiling, the takeoff range and the speed range into three S-curves. Show the board all three. Most of the decision value is in the gap between the slow and the fast case, because that gap determines how much capacity to commit and when.

6. Pick the leading indicators and update. For each scenario, write down what you would expect to see in the next one to two years: which named early adopters have bought, pilot orders, price points, competitor announcements, qualification results. Check them quarterly and move the forecast toward the scenario the evidence supports. This is where Christensen's advice fits: treat the first forecast as a hypothesis, and make the indicators decide the next investment, not the original model.

A caution on incentive-driven demand

Tax credits and grants create real orders. They can also make a market look further along its curve than it is. Christensen warned that demand created by mandate can distort the search for a market. Track how much forecast demand depends on the incentive, and build a scenario where the incentive changes.

What it looks like in practice

Take a composite case: a family-owned supplier is asked to support a component for a new energy application, and the customer's outlook calls for a fast ramp.

• The ceiling comes from the installed base the component would go into, times content per unit.
• The takeoff trigger is cost parity with the incumbent part, which the supplier dates within a two-to-four-year range.
• The speed comes from analog technologies sold into capital equipment, slower than the consumer curves above.

The result is three curves, not one. The supplier commits a pilot line sized for the slow case. It agrees with its bank that a production order and two quarters of cost data will release funding for the second line. That is the onion theory of risk applied to a forecast: each indicator peels away one layer.

What to do on Monday

Take your biggest growth bet and write down its ceiling, its takeoff trigger, the speed your plan assumes, and the names of the first five customers you expect to buy. Find two or three past technologies that took off in a similar way and check your speed against theirs. If your plan is faster than every analog, you need a reason. If you have one, write it down and name the indicator that will show whether it is holding.