Every purchase order is a prediction. A production run booked in March assumes a certain number of customers will want that product in June. Order too little and it sells out early. Order too much and the extra units sitting on shelves, tying up cash. Demand forecasting is how that number gets set with evidence instead of instinct, and some of the best evidence sits in the warehouse.
TL;DR
- Demand forecasting estimates how many units customers will buy, when, and where. The longer your lead time, the further ahead you have to predict.
- The main types come in pairs, and the right method depends on history: judgment for new products, statistical models or AI for established ones.
- Past sales understate demand whenever stock ran out, so those weeks need adjusting.
- Forecasts that double as sales goals run high, leaving extra stock in storage.
- Live warehouse data shows when a forecast is off while there is still time to react.
What Is Demand Forecasting?
A simple demand forecasting definition is the following: estimating future customer demand for a product, usually by SKU, channel, and week or month. The output is a number, such as 4,000 units of a shampoo on Shopify in November.
In demand planning and forecasting, the forecast is the estimate, and the plan is the response to it: purchase orders, production runs, and the labor to ship it all.
The Need For Demand Forecasting Comes Down To Lead Time
The need for demand forecasting exists because inventory takes longer to arrive than customers are willing to wait. That gap is lead time, covering production, freight, and the days it takes to shelve incoming stock.
Lead time sets how far ahead you forecast. With a 12-week lead time, you are forecasting 12 weeks out, and week 12 is always harder to call than week 2. Demand forecasting in the supply chain is really about what will sell by the time a new order lands.
So a shorter lead time improves the forecast, because there is less to guess. Merchdrop’s Chatsworth warehouse sits among roughly 20 cosmetic manufacturers within one to three miles, so a beauty brand producing locally can bring finished goods to the dock in a short trip. Add 3PL fulfillment services that report units in transit and send reorder alerts, and lead time becomes a figure you track rather than assume.
Types Of Demand Forecasting
Most forecast types come in pairs. Each one of them settles a different question on the warehouse floor.
Short-Term vs Long-Term
- Short-term forecasts cover the next few weeks to a year, drawing data from recent sales, promotions, and seasons. They drive reorders and staffing.
- Long-term forecasts look a year or more ahead, at growth plans and new channels. They tell you how much space you will need.
Macro vs Micro
- Macro (external) forecasts follow the economy and category trends, which shapes how cautious your next big buy should be.
- Micro (internal) forecasts zoom in on one SKU, channel, or region, and set how many units of each SKU to hold.
Passive vs Active
- Passive forecasts carry past results forward, which works well for replenishing steady sellers.
- Active forecasts build in planned launches and campaigns, so stock is in place before the push begins.
Demand Forecasting Methods And Techniques
Qualitative Methods For Products With No History
With no sales data, judgment fills the gap:
- Expert panels (the Delphi method): Several people forecast separately, compare, and revise until their numbers settle.
- Sales team input: What your team hears from retail buyers and distributors.
- Market signals: Pre-orders, waitlists, and sales of a similar product you already carry.
These methods usually decide the size of the first production run. Merchdrop’s brand and product development team works with production partners across the US, which helps when you are weighing one large first run against a smaller one with a fast reorder behind it.
Quantitative Demand Forecasting Models
Once a product has a sales record, the math takes over:
- Moving averages: Recent periods averaged and projected forward.
- Exponential smoothing: Recent weeks count more, so the forecast reacts faster.
- Seasonal time series: Separates the trend from repeating peaks such as the holidays.
- Regression models: Link demand to drivers like price or promotions.
All of them work best on steady volume. A SKU moving 2,000 units a week shows a pattern, while one moving three a week mostly shows noise. So a focused, high-volume catalog is easier to forecast.
AI Demand Forecasting
AI demand forecasting uses machine learning to read many inputs at once, from channel sales and pricing to weather and social media. Its limit is the same as any model’s: it learns from the history you give it, and that history has a gap most brands never check.
Your Sales History Is Not Your Demand History
Sales records show what you sold, not what customers wanted. When a SKU runs out, sales drop to zero, but shoppers keep arriving and leave empty-handed. A model reading those weeks concludes the product stopped selling, so the next order is smaller and the product runs out again.
The gap can be large. In a benchmark dataset of 50,000 store and product combinations from fresh food retail, approximately 44% of daily observations were affected by stockouts, according to a 2026 study from France’s FEMTO-ST Institute. Fresh food sells out faster than most categories, but the lesson holds for ecommerce demand forecasting too: adjust the weeks when a product was out of stock before using them in a forecast.
The warehouse holds the records that make that possible, such as the day stock hit zero and the day replenishment landed. The storefront holds the rest, in back-in-stock requests and visits to sold-out pages. When one team runs your online store and the shelves behind it, demand analysis and forecasting can draw on both.
A Forecast Is Not A Sales Target
Forecasts also get mixed up with goals. When the number that drives purchasing is also the number a sales team is judged on, it creeps upward. The University of Tennessee’s Global Supply Chain Institute notes that forecasts used internally as business commitments have historically run high, exceeding actual demand more than 90% of the time.
In a warehouse, that optimism shows up as pallets paying storage fees every month. For beauty and wellness brands, there is also a deadline, since expired products become write-offs. Merchdrop tracks every unit by lot and expiry date so older batches move out first, but the cleaner fix is to keep one number for what you expect to sell and another for what you hope to sell, and buy against the first.
Every Forecast Will Miss, So Plan For It
No method gets it exactly right. What decides a quarter is how quickly you spot the miss. When stock levels, sell-through, and low-stock alerts come in live from the team that owns and runs the warehouse, a forecast running hot or cold shows up within days, while there is still time to reorder or slow down. If you want your next forecast built on what actually happened on the shelf, contact Merchdrop.
Ready To Find Out If Merchdrop Is A Good Fit?
The fastest way to find out whether we are a good match for your needs and growth plans is a short conversation, not a long form. Send us your numbers and we will come back with a quote within 24 hours.
FAQ
What is demand forecasting?
Estimating how much of each product customers will buy over a coming period, so the right amount is in stock before they buy it.
How is demand forecasting different from demand planning?
Forecasting estimates what customers will buy. Planning decides what to do about it: what to order, when to produce it, where to store it, and how many people to schedule to ship it.
What are the main demand forecasting methods?
Qualitative methods, such as expert panels and market research, suit new products. Quantitative methods, such as moving averages and exponential smoothing, suit products with a sales record. AI models combine many inputs at once.
What is an example of demand forecasting?
A skincare brand sold 3,000 units of a serum last November. This year it adjusts that number for growth and a bigger promotion, corrects for the four days the serum was out of stock, and orders early enough to have stock shelved before the campaign.
How can a 3PL help with forecasting demand?
A 3PL does not set your forecast, but it holds much of the data behind it, including stock on hand, units in transit, and stockout dates. Live dashboards and reorder alerts show early when a forecast is drifting.
