The distinction between a demand forecast and a sales forecast may sound simple, but in practice it often causes confusion.
Understanding the difference is important, because each type of forecast serves a different purpose in inventory planning.
Demand Forecast vs Sales Forecast
A demand forecast estimates the quantity of a product that customers are willing to buy at a given moment in time.
It reflects customer intent, assuming the product is available and visible.
A sales forecast, by contrast, represents the quantity you are realistically able to sell.
It depends not only on demand, but also on:
- • how much inventory you have,
- • when replenishment arrives,
- • operational constraints such as Buy Box availability, listing visibility, and other conversion factors.
In other words:
- • Demand forecast: what customers want.
- • Sales forecast: what you can actually sell given your stock and supply plan.
When tools or dashboards present a "forecast" without clarification, it is not always obvious which of the two they represent.
Why Forecasting Demand Is Both a Science and an Art
Although the idea of forecasting demand seems straightforward, the practice is far from trivial. Several challenges arise even before discussing specific models or methods.
1. You need far more historical data than you may expect
Even sophisticated models (including those powered by machine learning) require significantly more history than the length of the forecast itself.
Short or fragmented data sets limit the model's ability to detect patterns.
2. You need enough data points to identify seasonality
Seasonality cannot be inferred from a few peaks.
It requires repeated patterns across years, with enough observations to separate:
- • genuine seasonal fluctuations,
- • one-time events,
- • and noise.
3. Past trends do not guarantee future trends
Consumer demand can shift due to competition, price changes, ranking changes, advertising, reviews, or macro factors.
A model trained on past behaviour may not recognize when conditions have fundamentally changed.
4. External factors have significant influence
Weather, holidays, promotions, media coverage, and competitor actions can all distort historical numbers.
Most automated systems cannot fully interpret these influences without explicit context.
5. Seasonality is more complex than it appears
It depends not only on months, but on:
- • the day of the week,
- • the proximity to holidays,
- • the number of consecutive non-working days,
- • pre-holiday vs. post-holiday behaviour,
- • school vacations,
- • regional patterns.
Such fine-grained patterns require a significant amount of detailed data.
6. The choice of time granularity matters
Whether you forecast in days, weeks, or months affects the result.
- • A more granular forecast captures nuances (e.g., demand spikes right before holidays).
- • But higher granularity usually decreases statistical accuracy because each interval contains fewer observations.
Choosing the wrong forecast granularity can lead to misleading results.
Why no algorithm can produce a perfect forecast
Even with strong data and advanced models, demand forecasting remains inherently uncertain.
There will always be unexpected events, trends that change, or factors that the model is unable to observe.
Forecasting, in practice, cannot be reduced to a fully self-sufficient automated process, because it depends on assumptions, data quality, and factors outside the model's visibility.
Algorithms can recognize patterns, but only a person can interpret nuances, context, and real-world events.
A purely automated forecast will inevitably miss certain signals.
Human oversight is essential
Because no model can incorporate every nuance, there must always be a way to adjust the machine-generated forecast manually.
Human input allows you to account for:
- • upcoming promotions,
- • known external events,
- • planned advertising changes,
- • supply constraints,
- • new competitors,
- • changes in marketplace rules,
- • insights that models cannot infer from data alone.
Forecasting works best when automation provides a structured baseline, and a person refines it using knowledge that is not captured in historical numbers.
The bottom line
A demand forecast and a sales forecast serve different purposes, and automated systems cannot perfectly predict either one.
Demand forecasting requires sufficient data, careful interpretation, and human judgment.
Recognizing its limitations, and the difference between demand and sales, leads to more informed decisions and more realistic expectations from forecasting tools.
In the previous article, we brought time back into inventory decisions using the cash conversion cycle.