Demand Planning for Ecommerce Brands
Demand planning answers one question before any money is spent: how many units of each product will customers actually buy, and when. This page covers how a demand forecast is built, the forecasting methods compared, and the traps that make sales history a poor guide to demand. RitePrep runs it as a planning service, separate from the warehouse work.
What Is Demand Planning?
Demand planning is the process of forecasting how many units of each product customers will buy, by period and by sales channel, then adjusting that forecast for known events such as promotions, launches and seasonal peaks. It produces one agreed demand number that purchasing and inventory decisions are built on, and it is revised every planning cycle.
It belongs to a different family of work from the warehouse. Receiving, storage, picking and shipping all handle stock that already exists. Planning decides what stock should exist in the first place. Demand planning is the first step of that: it says what will sell, before anyone decides what to buy.
Done well, it is less about a clever formula and more about honest inputs. The forecast is only as good as the history it learns from, and ecommerce history is full of things that look like demand but are not: stockout weeks that read as zero, promotion spikes, a single wholesale order, or a launch that will never repeat.
Demand Planning vs Demand Forecasting vs S&OP
Demand forecasting is the statistical step: take past demand, find the level, trend and seasonal pattern, and project it forward. Demand planning wraps that step in judgment. It cleans the history first, layers known future events on top of the statistical number, gets the people who own sales and marketing to agree it, and measures how far off it was.
Sales and operations planning (S&OP) sits one level above both. It is a cross-functional process, usually run monthly, that sets the agreed demand plan against supply, capacity and cash and decides what the business will actually do. Demand planning feeds the demand review inside S&OP. A small brand rarely needs the full ceremony, but it still needs the same discipline: one demand number, agreed on a fixed cadence.
Where Demand Planning Stops and Inventory Planning Starts
Demand planning ends at a forecast: expected units per SKU, per channel, per week or month. It does not say how much to order. Turning that forecast into reorder points, safety stock and buy quantities is inventory planning, which also has to account for supplier lead times, minimum order quantities and the cash you are willing to tie up.
Keeping the two jobs separate is what keeps each one honest. If the forecast is quietly padded to avoid stockouts, the buffer gets counted twice, once in the forecast and again in safety stock. The forecast should be the best estimate of demand, and protection against being wrong belongs in the inventory plan.
Why Sales History Is Not the Same as Demand
Sales only record demand you were able to serve. In a week a SKU was out of stock, sales show zero or close to it, but customers kept arriving. A forecast trained on that history learns to expect less, which causes the next stockout, which lowers the forecast again. Correcting stockout periods, for example by using the rate of sale on days the item was in stock, breaks that loop.
Shipments are not demand either. Units you send into Amazon fulfillment centers or to a retailer's distribution center are stock moving between locations. The demand is what end customers buy from there. Forecasting the shipments instead of the sell-through is how a brand ends up chasing its own replenishment pattern.
Inventory Forecasting Methods Compared
There is no single best forecasting method. Each SKU's history decides which one is reasonable, and most catalogs need several. These are the common methods, what each suits, and where each one goes wrong.
| Method | How it works | Suits | Watch out for |
|---|---|---|---|
| Naive | Next period's forecast equals the last period's actual demand. | A benchmark that any other method should beat. | Copies every random swing straight into the next order. |
| Moving average | Averages the last few periods, often four to twelve weeks. | Steady items with no trend and no seasonal pattern. | Lags behind growth or decline, and flattens seasonal peaks. |
| Seasonal naive | Uses demand from the same period last season, usually last year. | Strongly seasonal items with a stable yearly shape. | Needs at least a full season of clean history, and ignores growth. |
| Simple exponential smoothing | A weighted average where recent periods count most, set by a factor between 0 and 1. | Items whose level drifts slowly, with no clear trend or season. | Produces a flat forecast, so it misses trend and seasonality. |
| Holt-Winters smoothing | Exponential smoothing with separate level, trend and seasonal parts. | Established items with both growth and a repeating seasonal pattern. | Typically needs two full seasons of history to estimate the pattern. |
| Promotion-adjusted baseline | Forecasts the baseline from normal periods, then adds an estimated lift for each promo. | Brands that run planned sales, discount events or ad pushes. | Lift is often followed by a dip, because some buyers simply bought early. |
| Like-item (analog) | Borrows the sales curve of a similar past product, scaled to the new one. | New SKUs, new colorways and line extensions with no history. | A poor choice of comparison item carries its error straight across. |
| Judgmental overlay | A planner adjusts the statistical number for known future events. | Launch dates, price changes, lost listings, new retail doors. | Untracked overrides add bias. Record each reason and check if it helped. |
Accuracy is measured after the fact, not promised in advance. A method earns its place on a SKU by beating the naive forecast on that SKU's own history.
Why Ecommerce Demand Is Hard to Forecast
Most forecasting advice assumes years of steady history on a stable catalog. Ecommerce brands rarely have that.
Thin history: new SKUs, variants and bundles arrive with weeks of data, or none.
Seasonality: gifting peaks, weather and back-to-school shift demand by month, and one year is a small sample.
Promotions and launches: a sale or a launch spike inflates the baseline unless it is separated out.
Stockouts: weeks spent out of stock show up as low sales and pull the next forecast down.
Channels differ: a Shopify store, Amazon and wholesale accounts each have their own rhythm and order size.
Set-once forecasts: a number built in January is stale by March if nobody compares it with what happened.
None of this is solved by a better formula alone. It is solved by cleaning the inputs and reviewing the output on a fixed cycle.
What Our Demand Planning Covers
The output is a forecast you can buy against, plus the record of how it was built so the next cycle starts from evidence.
Cleaned Demand History
Order history by SKU and channel, with stockout periods corrected and promotion spikes, one-off bulk orders and returns flagged so they do not distort the baseline.
Baseline Forecast by SKU and Channel
A forecasting method chosen per SKU from its own pattern, tested against a naive benchmark, with family-level totals that roll up cleanly.
Events Planned on Top
Promotions, launches, price changes and marketing dates added as separate, documented adjustments, so the baseline stays clean.
New Product Forecasts
Low, expected and high scenarios for SKUs with no history, built from comparable items and your launch plan, then replaced by real data as it arrives.
Accuracy and Bias Review
Every cycle, last period's forecast is checked against real orders. Error and persistent over or under forecasting are reported by SKU, and the biggest misses get a reason.
Hand-off to the Buy Plan
The agreed forecast is passed on in a form inventory planning can use directly to set reorder points, safety stock and order quantities.
Demand planning is one part of our supply chain services. The forecast feeds inventory planning, which turns it into what to buy and when, and purchase order management then places and chases those orders with your suppliers. Deciding which SKUs to carry at all is assortment planning, and planning stock for Amazon FBA is covered under Amazon inventory management. Terms like safety stock and reorder point are in our glossary, the warehouse side of keeping counts accurate is ecommerce inventory management, and how unserved orders are recorded is explained in what is a backorder.
How a Demand Planning Cycle Runs
Gather the Inputs
Order history by SKU, channel and date, stock levels or stockout dates, your promotion and launch calendar, and any wholesale commitments.
Clean the History
Stockout periods are corrected, promo spikes and one-off orders are tagged, and discontinued or renamed SKUs are linked so their history is not lost.
Build the Baseline
Each SKU gets a method that fits its pattern, checked against how well it would have forecast the recent past.
Add Known Events
Planned promotions, launches and channel changes are layered on as separate adjustments, each with a written reason.
Agree the Number
The forecast is reviewed with whoever owns sales and marketing, so the plan the buy is based on is one the business stands behind.
Measure and Repeat
At the next cycle, last period's numbers are set beside real orders, the largest gaps are explained, and the forecast rolls forward.
Who Demand Planning Is Built For
It earns its keep when a wrong guess about demand is expensive, either in stockouts or in cash sitting in slow stock.
How We Approach Demand Planning
The principles behind every forecast we build, whatever the size of the catalog.
Demand, Not Shipments
We forecast what end customers buy, not what you send into Amazon or a retailer's warehouse.
Stockouts Corrected
Out-of-stock periods are treated as missing demand, not as low demand.
Every Override Has a Reason
Manual adjustments are written down and checked later to see whether they improved the forecast.
Ranges for New Products
A launch gets a low, expected and high case rather than one confident number.
Forecast and Buffer Kept Apart
The forecast is a best estimate. Protection against error is set in the inventory plan, so it is not counted twice.
Counts From a Real Warehouse
If your stock is in our Austin, Texas warehouse, on-hand figures and stockout dates come from the same scans that ship your orders.
Questions, answered
Demand planning is the process of forecasting how many units of each product customers will buy, by period and channel, then adjusting that forecast for known events such as promotions, launches and seasonal peaks. The result is one agreed demand number that purchasing and inventory decisions are built on, revised every planning cycle.
Start With a Forecast You Can Buy Against
Send us a sample of your sales history and tell us how you buy today. We will show you where the history misleads, and what a cleaned, SKU-level forecast for your catalog would look like.