Forecasting Early Revenue for New Offers
Here is the trap with a new offer: you have to forecast revenue for something that has never sold. There is no history to lean on, so the forecast quietly becomes a hope dressed up as a number. Then the launch misses, and the miss looks like a demand problem when it was really a forecasting problem. Simon-Kucher’s global research found that 72 percent of new products fail to hit their revenue and profit targets, and an optimistic, unexamined forecast is one of the quietest reasons why.
You cannot remove the uncertainty, but you can replace guesswork with discipline. This piece shows how to build an early revenue forecast you can actually defend. If you want the short weekly version of ideas like this, subscribe to the AI Profit Pulse newsletter for one evidence-based read each week.
Why are new-offer forecasts so often wrong?
Two forces work against you, and both are predictable.
The first is the blank slate. Traditional forecasting leans on history, and a new offer has none, so teams fall back on a top-down target (“we need $2M in year one”) and reverse-engineer a demand story to match. That is not a forecast, it is a wish with a spreadsheet.
The second is optimism bias. The people closest to a launch believe in it, which is a strength in the build and a liability in the forecast. Enthusiasm inflates conversion assumptions, compresses the sales cycle, and ignores the substitutes buyers will actually weigh. The result is a number that only works if everything goes right.
The cost is not abstract. Over-forecast and you overbuild, overhire, and overspend against revenue that never arrives. Under-forecast and you stock out, cap your own launch, and hand momentum to a competitor. Better forecasting is not a finance nicety. It is how you protect cash and capacity in the riskiest quarter a product has.
What does AI-enabled forecasting actually add?
AI does not conjure certainty out of a blank slate. What it does well is find structure in signals you already have but do not use in one place.
The measured gains are real. McKinsey’s operations research, still the most cited benchmark in the field, puts AI-driven forecasting at a 20 to 50 percent reduction in forecast error versus traditional methods, with product unavailability falling by as much as 65 percent. The knock-on effect on the top line is measurable too: a McKinsey Global Institute analysis estimated that a 10 to 20 percent improvement in forecast accuracy yields a 2 to 3 percent revenue increase for consumer-goods companies, from fewer stockouts and more responsive supply.
One honest caveat, because this brand does not do hype. Those numbers come mostly from forecasting existing products with sales history. A brand-new offer is harder, and no model will hand you a precise year-one figure with confidence. The value of AI here is not false precision. It is speed and structure: turning analogs, external signals, and early demand into a defensible range, and updating it fast as real data lands. If you want the deeper mechanics of how the models work, our guide to how predictive analytics actually forecast prices and demand goes under the hood.
How do you forecast an offer with no history?
No history is not the same as no signal. Build the number from the ground up in four moves.
Anchor on analogs. Your closest comparable products and past launches are your base rate. How did similar offers ramp in their first 90 days? What conversion did comparable price points earn? Start from evidence, not from a target.
Read pre-launch demand signals. A new offer generates data before it earns a dollar. Waitlist sign-ups, pre-orders and deposits, search interest, landing-page conversion, and pricing-page clicks are all real willingness-to-buy signals. Weight them honestly and they sharpen the forecast considerably. Pricing sits underneath all of this, so it is worth grounding your number in the hidden math of willingness-to-pay analysis before you model revenue.
Forecast a range, not a point. A single number hides the risk and invites false confidence. Model a best, base, and worst case, ideally by segment, and plan capacity against the base while staying ready for the best. A range is not indecision. It is an accurate picture of a genuinely uncertain launch.
Re-forecast fast. The first weeks of real orders tell you more than any pre-launch model. Update weekly, not quarterly. Early accuracy, corrected quickly, beats a precise-looking forecast you refuse to touch. For the launch price itself, pair this with our guide to setting the right pricing strategy for a new product launch.
What good looks like
A strong early forecast is legible and honest. Anyone on the team can see which analogs it rests on, which signals moved it, and how wide the range is. It states its assumptions out loud, so when reality diverges you know exactly which assumption broke. And it is built to be revised, not defended, because the goal is an accurate picture, not a number you have to protect in the next review.
Traditional methods leave median forecast accuracy somewhere around 70 to 79 percent; disciplined, signal-driven forecasting pushes that meaningfully higher. Gartner expects 70 percent of large organizations to adopt AI-based demand forecasting by 2030, so the discipline is becoming standard. Adopting it now, on your next launch, is still an edge.
The takeaway
You cannot forecast a new offer from history you do not have. You can forecast it from analogs you do have, demand signals it generates before launch, an honest range instead of a point, and a fast cadence of updates. That is how you avoid being one of the 72 percent that miss, without pretending to a precision no one has.
Want one evidence-based pricing and revenue idea in your inbox each week? Subscribe to AI Profit Pulse. And if you want a fast read on where your current pricing is leaking margin before you forecast the next launch, the Pricing Pulse Audit is a short, evidence-based place to start.