How to Forecast Demand for a New Product Launch

The standard advice for demand forecasting assumes you have data. Use a weighted moving average. Apply a seasonality index. Run a regression. All of that requires history — months or years of actual sales to extrapolate from.

New product launches don't have that. You're forecasting something that has never sold, at a velocity that's never been observed, in a channel that may be new to your brand entirely. The statistical toolkit doesn't apply, at least not directly.

Most content on this topic either pretends the data problem doesn't exist or gives you a generic framework that doesn't account for how decisions actually get made in a small brand. So here's what practitioners who have launched products at Unilever and at founder-led brands actually do — and why one planner's decision to add a 33% buffer to a forecast saved a mass retail launch from running out of stock in week two.

Why New Product Forecasting Is Different

When you're forecasting an existing product, errors tend to average out over time. You over-call one month, under-call the next, and your inventory position self-corrects through the reorder cycle.

New products don't give you that cushion. The cost of getting it wrong in one direction is severe:

Over-forecast: You produce too much. Capital is tied up in inventory that may not move. If the product has a shelf life — and most CPG products do — the clock starts immediately. In a regulated category, you may not be able to discount or donate the excess.

Under-forecast: You run out of stock during the launch window. For DTC, that means lost sales and a gap in your customer acquisition funnel. For retail, it's worse: empty shelves in week three of a new product launch signal to the buyer that you can't execute. Some buyers won't reset your product until the next planogram cycle, which could be six months away.

The consequence of under-forecasting at retail is not just lost sales today — it's lost distribution for the next half year.

This asymmetry matters. It means the right approach to new product forecasting isn't neutral. It's deliberately biased toward having enough.

Step 1: Start With an Analog

The first move when you have no history on a product is to find a product that does have history and is close enough to use as a baseline.

This is called analog forecasting, and it's the most reliable starting point for new product launches when you have any comparable data at all.

The analog doesn't have to be identical. It has to be close enough on the dimensions that drive velocity:

• Category and consumer need — A new moisturizer can benchmark against your existing moisturizer line, not against your cleanser
• Price point — A $28 product will sell differently than a $14 product, even in the same category. Adjust.
• Channel — DTC velocity and retail velocity are different. Use the right analog for the channel you're launching into.
• Format — A new scent or size variant of an existing product has much stronger analog data than a completely new product form

Once you've identified the closest analog, apply your adjustments. If the analog sells 800 units/month DTC and your new product is in the same category at a 15% higher price point with a slightly narrower use case, you might apply a 0.75 factor and start at 600 units/month as your base assumption.

This is a judgment call. That's fine. The goal at this stage is a range, not a precise number.

Step 2: Run a Delphi Process

When analog data is weak or unavailable — which is common for genuinely new product categories, new formats, or new channels — the Delphi method is the most reliable way to synthesize what you actually know.

The Delphi process gathers independent input from multiple people who have relevant knowledge, aggregates it, and iterates until the estimates converge. In a large CPG company, this might be a formal multi-round survey. In a founder-led brand, it's a structured conversation between the people who know the most.

Who to include:

• The demand or inventory planner (responsible for the final number)
• Sales lead or key account manager (knows the channel, knows the buyer's expectations)
• Marketing (owns the promotional calendar and launch investment)
• Someone with direct consumer insight — customer service, a category buyer, retail field reps if you have them

How to run it:

Ask each person to provide their independent estimate of launch velocity before the group discussion. This is critical. If the demand planner shares their number first, everyone anchors to it. You want uncontaminated input.

Then aggregate the estimates. If the range is tight (within 20–30%), you have reasonable consensus. If the range is wide — marketing is calling 2,000 units/month and sales is calling 800 — the gap tells you something important about what's unknown. Don't average your way through it. Ask what assumptions are driving the divergence and resolve those first.

The output of a Delphi process isn't a single number. It's a range: a conservative case, a base case, and an upside case. All three should go into your initial production planning.

Step 3: Build Your FLM Meeting

For larger launches — a mass retail debut, a new product line, or a channel expansion — the Delphi input feeds into what enterprise planners call an FLM: a Flawless Launch Management meeting.

The FLM is a cross-functional review designed specifically for new product launches. It's different from a regular S&OP meeting in a few ways:

It's launch-specific. You're not reviewing the whole portfolio — you're pressure-testing the assumptions behind one product. Who's the target consumer, and how realistic is the velocity assumption based on category benchmarks? What's the promotional support behind it, and what lift does that actually buy you? What are the lead time constraints, and when do you need to place the initial production order to have inventory available on launch day?

It's assumption-explicit. The purpose is to surface and challenge every assumption that went into the forecast. Not to defend the number — to stress-test it. Someone in the room should be asking: what would have to be true for this forecast to be right?

It's decision-forcing. The FLM ends with a production decision, not a follow-up meeting. You've reviewed the range, you've stress-tested the assumptions, and you're committing to a number. That commitment includes the buffer.

At a large CPG company, the FLM may involve category managers, retail buyers, and sales VPs. At a $10M brand, it might be four people on a video call. The format doesn't matter. The discipline does.

Step 4: Add the Buffer — And Understand What It's For

Here's the part that separates experienced demand planners from founders who are forecasting for the first time.

The buffer is not padding. It's not pessimism. It's a deliberate adjustment that reflects the asymmetry of launch risk.

A demand planner working on a mass retail launch for a cleansing oil — entering Target and Walmart at an accessible price point — looked at marketing's forecast and thought it was too low. The product had the characteristics of an outperformer: a disruptive price point, a format that hadn't been widely distributed at mass, and strong early signals from limited DTC data.

Rather than override marketing's number in a meeting, she applied a 33% buffer above their forecast and built the production plan to that number. No drama. Just a judgment call based on category knowledge and launch history.

The product came in 40–50% above forecast. Without the buffer, they would have run out of stock at Target, Walmart, and drug within the first two weeks of the launch — during the exact window when velocity is being established, buyers are watching, and retail momentum is either built or lost.

With the buffer, there was enough time to scramble: pulling forward production, allocating across retailers, managing through the shortfall without losing distribution. The launch survived.

When to add a buffer, and how much:

• New product, established category, existing retail distribution: 15–20% above base forecast
• New product, new category for your brand: 20–30% above base forecast
• New product launching into a new retail account: 25–35% above base forecast
• New product at mass retail (Target, Walmart, drug): 30–40% above base forecast
• Relaunch or reformulation with existing consumer base: 10–15% above base forecast

These aren't rules. They're starting points. Adjust based on your lead time, your production flexibility, and your downside exposure. If you can reorder in 4 weeks, you need less buffer than if your co-man is overseas and lead time is 14 weeks.

The key question is: if demand comes in at the high end of my range and I'm wrong in the optimistic direction, can I recover? If the answer is yes, the buffer can be modest. If the answer is no — if running out of stock in retail means losing shelf placement — size the buffer to make sure that doesn't happen.

Step 5: Set a Review Trigger, Not Just a Review Calendar

Most brands review their forecasts monthly. For a new product launch, monthly isn't enough.

Set a trigger-based review point in advance: "If we've sold through 40% of initial inventory in the first 30 days, we reconvene and pull forward the next production run." Define the signal before the launch, not after you see the velocity.

The same logic applies on the downside: "If sell-through in week 4 is below X, we pause the production queue and reassess." Define that number too.

Launch forecasting is inherently uncertain. The goal isn't to get the forecast right before the launch — it's to be watching the right signals early enough to adjust before the decision window closes.

A Note on New Product Forecasting With Very Limited Data

If you're a small brand launching your first product, or launching into a category where you have no comparable history at all, the honest answer is: your initial forecast will be wrong. The question is how wrong, and in which direction.

In that situation, the framework still applies, but the weight shifts:

• Lean harder on Delphi input from people with category experience — buyers, brokers, retail reps who know the velocity of comparable products
• Use conservative case assumptions for production, and build a fast-reorder plan if demand exceeds expectations
• Price the downside explicitly: if this sells at 50% of plan, what does my inventory position look like in 12 months, and can I absorb it?

The worst outcome isn't a forecast error. It's a forecast error you didn't plan for.

About to launch a new product and not sure your forecast assumptions are solid?

Izba's free 20-minute demand planning audit is a fast pressure-test on your launch assumptions — we'll look at your forecast, your buffer, and your lead time logic and tell you where you're exposed.

Book your audit at izba.co