Multi-Channel Demand Planning: What a Demand Planning System Has to Do

What is a demand planning system?

Demand planning is the work of deciding how much of each product customers will order, by week or month, far enough ahead to buy materials or finished goods in time. A demand planning system is whatever holds that work: the data, the forecast, the overrides and the record of who agreed to what.

For a consumable CPG brand or manufacturer, the system has five jobs. The table sets out each one, and why selling through several channels makes it harder.

Demand planning ends where supply planning begins: demand planning says what will sell, and supply planning says what to buy or make to meet it. The forecasting step itself is covered in what is demand forecasting.

Why does each channel need its own forecast?

A single SKU sold through four channels is really four demand patterns that happen to draw on the same inventory. Each has its own driver and its own best input.

If you forecast the total of all four, the combined line looks smoother than any channel really is. A strong DTC month can hide a retailer that has stopped reordering until that retailer's next order fails to arrive. A distributor's large quarterly order can look like a trend when it is a single event.

Retail needs its own inputs

Retail shipments are the retailer's orders to you, not consumer sales. They arrive in batches, with an initial fill for a new listing followed by replenishment that depends on how the product sells in store. The earliest signal is the sell-through, so a retail forecast is best built bottom-up from door counts and units per store per week, then converted into the orders the retailer's distribution centres will place. Planning around door count covers how that changes when a retailer adds stores.

Retail and DTC behave differently

DTC demand moves with decisions you control. Retail demand moves with decisions the retailer controls. Retail vs DTC inventory works through the differences in more detail.

When should channel forecasts be combined?

Combine them at the point you decide what to buy or make. Until then, keep each channel separate so that every override and every promotion stays attached to the channel it belongs to.

At the buying decision, the combined forecast is netted against what you already have:

1. Add the channel forecasts for each SKU and week.
2. Subtract stock on hand, stock on order and stock already committed to open customer orders.
3. Add the safety stock you hold against forecast error.
4. Time the result back by the supplier's lead time and round it to the MOQ.

Where inventory sits matters as well. Stock at an Amazon fulfilment centre cannot fill a retailer's order, and stock at a 3PL in one region may be the wrong stock for a customer in another. If channels are fulfilled from different locations, plan demand per location as well as per channel.

What happens when supply is short?

When there is not enough stock for every channel, the system has to support an allocation rule that someone has agreed in advance:

• Priority. One channel is filled first, usually the one with the largest penalty for a short shipment, such as a retailer that charges for late or incomplete orders.
• Proportional. Each channel gets the same share of its forecast.
• Reserved. A fixed quantity is held for a channel, and the rest is filled in order.

The rule is a commercial decision, not a planning one. The system's job is to show the shortfall early enough for the people who own each channel to make it.

What should a demand planning system do that a spreadsheet cannot?

A spreadsheet is a capable demand planning system for a small catalogue. It becomes the weak point when the plan cannot be rebuilt as often as the business changes. A dedicated system should:

• Fit a model per SKU and channel, and change the model when the pattern changes, rather than applying one method to the whole catalogue.
• Refresh every night from your sales channels, warehouses and ERP, so the plan is current without a manual export.
• Record every override with who made it, when and why, so the S&OP meeting can review changes rather than argue about whose version is right.
• Model scenarios such as a promotion, a new retailer or a lost listing, and show the purchase orders and cash each one commits before anyone approves it.
• Flag exceptions, such as SKUs selling well above or below forecast, so planners spend their time on the items that changed.

When is a spreadsheet the better fit?

A spreadsheet is the better fit when you sell a few dozen SKUs through one or two channels, demand is stable, and one person owns the plan and rebuilds it weekly. In that situation a planning tool adds setup work without changing many decisions. Spreadsheets vs planning software sets out the signs that the balance has shifted.

Can the ERP do it instead?

ERPs include demand planning features, and they are worth checking before buying anything else. NetSuite Demand Planning, for instance, is sold as an add-on module with four projection methods: linear regression, moving average, seasonal average and a sales forecast you maintain. Companies with many SKUs across several channels often keep the ERP as the system of record for orders, receipts and inventory, and plan demand in a separate layer that reads from it.

Where does Planster fit?

Planster is demand and supply planning software for consumable CPG brands and manufacturers between $50M and $500M in revenue: food, beverage, supplements, beauty and household goods. On the demand side it works as described above.

• Each SKU gets the forecasting model that fits its history best: Prophet, ETS, SARIMA or linear regression. The model is re-selected when the pattern changes.
• Channels are planned separately and then combined: retail, wholesale, DTC and Amazon. See multi-channel planning.
• Retail is planned bottom-up per retailer, from door counts and velocity, on the retail planning view.
• Retail deals go into the plan at full volume with the date you have to order by to make the launch, through retail deal planning (Pipeline).
• Sales and inventory sync nightly from 150+ warehouse, 3PL and sales-channel connections, and from NetSuite through Trackstar or Sage 100 Premium through a native connector.
• Safety stock is set in days of cover for each SKU and warehouse, and Planster converts it to units from the forecast for the weeks ahead.

Planster needs about a year of sales history to forecast a SKU from its own data. It drafts purchase orders and flags what needs a decision; a person approves each one.

What to do first

1. List every channel each SKU sells through, and where each channel's sales data lives today.
2. Check whether your current forecast is per channel or a single total.
3. Agree the allocation rule for a shortfall before you need it.
4. Decide who owns each channel's part of the forecast, and review overrides in the S&OP meeting.