How to do demand planning with limited historical data

Most demand planning advice starts with the same assumption: you have at least a year of sales history to work with.

Run a moving average. Apply a seasonality index. Calculate your MAPE. Pick the technique with the lowest error rate.

All of that is useful — once you have the data to run it on. But if your brand launched six months ago, or you're about to launch a new product line with no comparable history, the standard approach doesn't give you much to work with.

This is one of the most common situations founder-led brands face, and one of the least well-served by the planning literature. The answer isn't to wait until you have enough data. It's to use a different set of methods that are built for uncertainty rather than pattern recognition.

Here's how to build a demand plan when history isn't an option.

First: understand what stage you're actually in

Limited data means different things depending on where you are.

Zero to three months of sales. You're in pure launch territory. There is no meaningful trend to extrapolate from. Statistical forecasting models will produce numbers that look precise and mean very little. At this stage, your plan needs to be built on external evidence and expert judgment rather than your own history.

Three to six months of sales. You have early signals but no full seasonal cycle. You can see whether demand is growing, flat, or declining, but you can't yet distinguish a genuine trend from the noise of a new product finding its footing. Use this data as a directional input, not a statistical foundation.

Six to twelve months of sales. You're starting to have something to work with. A weighted moving average can begin to capture recent momentum. You still won't have a full seasonality picture, but you can build a more grounded baseline than in the earlier stages.

Each stage needs a different approach. The mistake is applying a statistical model to data that isn't dense enough to support it — which produces false confidence in numbers that aren't actually reliable.

Method 1: Analogous product benchmarking

The most useful question you can ask when you have no sales history is: what does a product like this usually do?

Every category has patterns. A new personal care product launching at mass market at a mid-range price point has a reasonably knowable range of velocity at a store like Target or Walmart. A supplement launching DTC has comparable products whose trajectories are visible — how fast they ramped, where they plateaued, what their repeat purchase rate looked like.

The goal is to find two or three products that are genuinely comparable — similar category, similar price point, similar channel, similar distribution footprint — and use their performance as a benchmark for your own range.

Where to find this data:

Syndicated data providers like Circana and Nielsen publish category velocity benchmarks. If you're entering retail, your retail buyer may also share category data — it's worth asking directly.

Your own portfolio. If you've launched products before, how did they ramp? A new SKU in an existing line often follows a similar trajectory to the products that came before it — adjusted for any meaningful differences in price, format, or distribution.

Industry contacts and category experts. People who have launched in your category before have a working mental model of what typical performance looks like. This is informal but often surprisingly accurate. A founder who has launched three personal care products at Target will have a strong intuition for what week-one velocity looks like.

The output of this exercise isn't one number. It's a range — a lower bound based on conservative analogues and an upper bound based on optimistic ones. That range becomes the foundation for your scenario planning.

Method 2: The Delphi approach

When numerical benchmarks aren't available or aren't reliable enough, structured expert judgment is the next best tool.

The Delphi method — originally developed for long-range forecasting in complex systems — works by gathering independent estimates from multiple people with relevant knowledge, sharing those estimates with the group, and iterating toward a consensus range.

In practice, for a founder-led brand, this doesn't require a formal process. It means deliberately gathering input from the people most likely to have a useful view on your product's demand:

Your sales team or retail broker. They've seen how comparable products perform at the accounts you're entering. They have a view on whether the category is growing or contracting, whether the timing is favorable, and what a realistic initial order cycle looks like.

Your retail buyer. If you have one, they know what their customers are buying. They won't give you a number, but they'll tell you whether they're seeing strong growth in your category, what their modular space looks like, and whether they're planning any promotional support.

Category veterans. Advisors, investors, or consultants who have worked in your category have seen multiple launches. Their intuition about what "good," "average," and "disappointing" look like is a form of data.

Your own marketing team. Pre-launch signals — website traffic, email list growth, waitlist sign-ups, social engagement — can be converted into rough demand proxies if you have comparable launch data to calibrate against.

Gather these inputs independently first, then bring them together. Where they converge is your base case. Where they diverge is the source of your range.

The value of this approach is that it forces you to be explicit about your assumptions. When every estimate is written down and compared, the optimistic ones and the conservative ones become visible — which makes it harder to let the most enthusiastic voice in the room set the plan.

Method 3: Conservative range forecasting

When data is limited, the single biggest mistake is planning to one number.

A single-point forecast on a new product implies a level of confidence you don't have. It also sets up a binary outcome — you were right or you were wrong — rather than a framework for making decisions under uncertainty.

Range forecasting gives you three scenarios to plan against:

Base case: Your best estimate given all available information. What you genuinely expect to happen if nothing unusual occurs in either direction.

Conservative case: What happens if the product takes longer to ramp than expected, distribution is slower to build, or category conditions are less favorable. This is typically 60 to 70 percent of your base case.

Optimistic case: What happens if the product outperforms, distribution builds faster, or you get unexpected media attention or a viral moment. This is typically 130 to 150 percent of your base case.

The three scenarios aren't just intellectual exercises. They produce three different inventory requirements — and the gap between them is where you make your actual planning decisions.

**Produce to the base case. Hold buffers for the optimistic case. **The buffer doesn't need to be full finished goods. Depending on your supply chain, it might be raw materials held at your co-manufacturer, a confirmed but uncommitted production slot, or an air freight option identified in advance. The goal is to be able to respond quickly if demand comes in above base — without having committed the cash in advance for inventory you might not need.

Identify your floor. The conservative case tells you the minimum you need to be able to absorb without a financial crisis. If your conservative case results in inventory you genuinely can't sell through, your production commitment is too high. Adjust the base case down until the conservative case is manageable.

What to do as data accumulates

The methods above are not permanent replacements for statistical forecasting. They're the right tools for a specific stage.

As your sales history builds, start transitioning:

At six months, introduce a weighted moving average alongside your judgment-based forecast. Compare the two. See whether the model is tracking with reality or diverging from it.

At twelve months, you have enough data to calculate seasonality — whether certain months are reliably higher or lower than average. Add that layer to your model.

At eighteen to twenty-four months, you have a full picture. Statistical techniques become your primary tool and expert judgment becomes a check on the model rather than the foundation of the plan.

The transition is gradual rather than a hard switch. The goal throughout is to reduce the proportion of the forecast that relies on assumption and increase the proportion that relies on data — without ever pretending you have more confidence than the evidence supports.

The honest summary

Demand planning with limited data is harder than demand planning with a full history. Anyone who tells you otherwise is selling something.

What you can do is be structured about your uncertainty rather than pretending it doesn't exist. Use analogues to anchor your range. Use expert judgment to pressure-test your assumptions. Build scenarios rather than a single number. And hold buffers that give you room to respond to outcomes you didn't predict.

That won't make the forecast right. It will make being wrong less expensive.

If you're navigating a launch or a new product line and want a second set of eyes on your demand assumptions, we're happy to take a look.