Demand planning and demand forecasting are not the same thing

Two phrases get used as if they were interchangeable, and the confusion costs real money. Demand forecasting and demand planning are not the same activity. One predicts what customers will want. The other decides what you will do about it. When a brand runs them as a single blurred step, the plan inherits the forecast's fragility, and it drifts the instant the forecast is wrong.

Here is the clean version. A forecast is a statement about the future: this SKU, in this size, in this channel, will sell roughly this many units over the next twelve weeks, with this much uncertainty. A plan is a decision constrained by reality: given that forecast, and given my working-capital ceiling, my lead times, my minimum order quantities and my markdown headroom, this is what I will actually buy, allocate and reorder.

The forecast has no constraints. The plan is nothing but constraints. That is the whole distinction, and getting it wrong is why so many plans look precise in January and describe nobody's business by March.

The vocabulary itself invites the confusion. Both phrases share the word demand, both live in the same team and often the same meeting, and both produce numbers that go in the same spreadsheet. So people treat them as two names for one activity. They are not. They are two activities that happen to be adjacent, and running them as one is a bit like calling the weather forecast and your decision to bring an umbrella the same thing. One is a prediction you do not control. The other is a choice you fully control, made in light of that prediction and everything else you know.

Watch what happens when the two steps are welded together. A brand builds a forecast, treats it as the plan, translates it straight into a buy, and locks it. There is no separate decision layer holding the constraints, no record of the trade-offs made, no way to re-decide without re-forecasting from scratch. The forecast and the buy are the same object.

Then real sell-through arrives and disagrees with the forecast, which it always does, because a forecast is a distribution and reality is one draw from it. Now the buy is wrong, but there is no plan to adjust, because the plan was never a separate thing. The only move is to reopen the whole exercise, which takes a week nobody has, so the buyers stop touching it and manage by feel. The plan drifts because it was never built to move. It was a forecast wearing a plan's clothes.

A brand that keeps the two separate has somewhere to stand when the forecast moves. The forecast updates. The plan re-decides against the new forecast, inside the same constraints, and only the parts that changed get changed. The working-capital ceiling still holds. The lead-time floor still holds. The plan absorbs the new information instead of shattering on it.

That capital is not freed by planning less. It is freed by keeping the decision layer alive: money moves out of the styles the forecast has cooled on and into the ones it has warmed to, while there is still time to place the order and receive it in season.

You can see the conflation in the language people use. When a plan misses, the post-mortem says "the forecast was wrong," as if that explains the miss. But a forecast being wrong is not a failure; it is the expected behavior of a forecast, because it is a probability, not a promise. The real failure is that the plan had no mechanism to respond when the forecast turned out to be one of the wrong ones, which it was always statistically going to be some of the time. Blaming the forecast for a plan that could not adapt is like blaming the weather report for the fact that you left the house without a coat.

A common instinct is to assume that a good enough forecast makes planning unnecessary. If you knew exactly what would sell, surely you would just buy that. But you cannot buy the forecast. The forecast says sell 1,200 units across the season. The plan has to reckon with a supplier minimum of 2,000, a lead time that eats the first four selling weeks, and a working-capital budget that is already committed elsewhere. The forecast is an input to that decision. It is not the decision.

This is why the two need different tools and different rhythms. The forecast should re-run constantly against live data, because prediction gets better the more recent the signal. The plan should re-decide on that fresh forecast but always through the guardrails: the ceilings, floors and headroom that keep the buy inside the business. Tightly's platform runs them as two connected layers, so the forecast can move hourly while the plan stays governed, and the agent (Tia) stages the buy and reorder moves the new forecast implies for the buyer to approve.

The rhythm mismatch is the part most brands get wrong even when they nominally separate the two. They forecast once a week and plan once a week, in the same meeting, and call that a live process. It is not. The forecast wants to move as often as the signal moves, which for a DTC brand is continuously. The plan wants to re-decide as often as it is useful to place or adjust an order, which is frequent but bounded by lead times and supplier windows. Force both onto the same weekly clock and you throttle the forecast down to the plan's speed, which throws away exactly the recency that made the forecast worth having.

Fourteen points of full-price sell-through is the difference between a healthy season and a promotional one. Much of it comes from this one structural choice: whether the plan can re-decide when the forecast moves, or whether it can only be defended until it fails.

The practical shape is two connected layers, not one blurred step. The forecasting layer predicts demand at SKU by size by channel, re-runs against live sell-through, returns and channel mix, and carries a confidence range on every number. The planning layer takes that forecast and re-decides the buy, the allocation and the reorders inside the working-capital ceiling, the lead-time floor, the MOQs and the markdown headroom, staging the moves for a buyer to approve in one pass.

The connection between them is where the confidence range earns its keep. A plan does not just need a forecast; it needs to know how sure the forecast is, because certainty and uncertainty call for different decisions. A high-confidence prediction can be bought close to the number. A low-confidence one gets a wider berth, or a smaller opening order with room to chase. When the plan can read the range, not just the point, it stops treating every forecast as equally trustworthy and starts sizing the buy to the actual risk. That is only possible when forecasting and planning are separate layers that talk to each other, rather than a single number written straight into a purchase order.

A forecast you cannot act on is trivia. A plan without a forecast underneath it is a guess. You need both, and you need them kept apart.

Draw the line clearly and the drift problem largely solves itself. The forecast is allowed to be wrong, because it is a prediction and predictions carry uncertainty. The plan is allowed to change, because it was built to re-decide. What is no longer allowed is the thing that quietly wrecks most seasons: a frozen buy that everyone knows is stale by week three and nobody has the week to reopen.