A common planning failure is applying one approach across every time horizon. Detailed scheduling stretched twelve months out produces work that gets discarded. Aggregate thinking applied to next week produces plans the floor can’t execute. The horizons need different methods, different data, and different levels of detail.

Long range: twelve to thirty-six months

The questions here are structural. Do you need more capacity, another site, different equipment, a new supply base?

Forecasts at this distance are unreliable and pretending otherwise wastes effort. Work in aggregate volume by product family, and use scenarios rather than a single number — what happens under high growth, flat, and decline. The purpose isn’t prediction; it’s ensuring the decisions you make now don’t foreclose options you’ll want later.

Detail at this horizon is actively counterproductive. Nobody needs a SKU-level forecast for month twenty-eight, and producing one creates false confidence.

Medium range: one to six months

This is where most planning value sits and where most attention should go.

The decisions are shift patterns, labour, subcontracting, long-lead material commitments, and inventory build for seasonal peaks. Product mix matters here in a way it doesn’t in aggregate long-range planning, because mix determines which constraint binds.

The right level is product family or major SKU, in weekly or monthly buckets. Forecasts are meaningful at this range and worth investing in. This is the horizon your S&OP cycle should cover.

Short range: one to six weeks

Execution. Sequencing, allocation, expediting, daily decisions.

Here you work mostly from firm orders rather than forecasts, at SKU and operation level, in daily buckets. What matters is current data and fast response rather than analytical sophistication — a good plan built on yesterday’s stock position is worse than an adequate plan built on today’s.

Where the horizons connect

The transitions are where planning usually breaks, and both directions matter.

Downward, each horizon should constrain the next. The medium-range plan sets the envelope within which short-range sequencing happens. If short-range decisions routinely violate the medium-range plan, either the plan was infeasible or the process isn’t being followed — and it’s worth knowing which.

Upward, execution reality should inform planning assumptions. If demonstrated capacity keeps coming in below the planning figure, the planning figure needs to change. This feedback loop is missing in most manufacturers, which is why medium-range plans stay optimistic indefinitely.

Freeze windows

Define how far ahead the plan is fixed and enforce it. Without a freeze, short-range chaos propagates upward and no plan means anything.

Set the freeze period by your longest binding constraint — usually material lead time or changeover economics. Anything inside it should require an explicit trade-off decision rather than a quiet resequence.

A practical test of whether your horizons are working: ask what decision each planning output actually drives. If a report exists that nobody acts on, it’s being produced at the wrong horizon or the wrong detail level, and dropping it frees time for the horizon that needs the attention.

ticktick.ai plans at appropriate granularity per horizon, with freeze enforcement and feedback from demonstrated capacity into medium-range assumptions.