What a planning week actually looks like.

STEP ONE · SHAPE THE PLAN

Time, place, and what you are planning.

Three things define a plan. A planning calendar cuts time into buckets — the weeks or months your business talks in. Demand locations name the places demand actually lands: a customer’s distribution centre, a region, a channel. And the plan itself binds a calendar to a horizon and a status.

The bucket is the unit everything else hangs off. A forecast quantity belongs to a bucket. An open order falls into a bucket. Accuracy is measured on buckets that have fully elapsed. Get the bucket right and the rest of the system follows.

Demand locations are what make customer-level planning possible. Business Central knows a sales order has a ship-to address; it does not know that the address is Costco’s Bay Area DC and that you plan it separately from their San Diego DC. Mapping ship-to codes to demand locations is what closes that gap.

STEP THREE · WORK THE WEEK

Real orders arrive and meet the plan.

This is where most of a planner’s week goes. Sales orders come in, get resolved to the right customer and demand location through their ship-to mapping, and land in the bucket their shipment date falls in. The plan and reality sit on the same screen.

Variance is the difference between them, coloured against thresholds you set. Green is fine. Amber says look. Red says booked orders have already passed what you planned — and somebody downstream is about to be surprised.

Where the gap is material, a recommendation appears with its evidence attached: this is the forecast, these are the orders, this is the gap. Accept it, change the number, or reject it. Rejected suggestions stay rejected.

Demand that arrives with no forecast at all shows up too, so a new item or a new ship-to cannot quietly go unplanned.

STEP FOUR · EXPLAIN THE CHANGE

Next year, somebody will ask why.

A number in a spreadsheet carries no memory. Six months later nobody remembers whether a spike was a promotion, a distribution gain, or a typo — and the person who knew has often moved on.

Reason codes attach a why to a change. You decide how hard the system pushes: optional, required only when a change is large enough to matter, or required every time. Demand events go further, describing something happening in the world — a promotion window, an assortment switch, a phase-out — with dates, a reason, and a note.

Both land in forecast history: one line per quantity change with the old value, the new value, who made it, when, where it came from, and the reason they gave. That is the change tracker most planners are already keeping by hand, kept by the system instead.

Approval gates and freeze windows sit alongside. A freeze window locks buckets close enough to now that changing them would disrupt people already committed to them.

STEP SIX · LOOK BACK

Was the forecast any good?

Accuracy is only meaningful against what you believed at the time. So each cycle the plan is archived as an as-of snapshot, and performance compares those archives against what actually happened.

Results are grouped by lag — how far ahead the forecast was made. A number set one week out should be better than one set six months out, and grouping by lag shows whether that is true for you. Only fully elapsed buckets are measured, so a half-finished week never drags the score.

Accuracy, bias, and mean absolute error are reported per band. Bias is often the more useful of the three: it tells you whether you are consistently over or under, which is a habit you can correct.