Its operations team is preparing for next Tuesday and needs to anticipate demand before committing stock, packing capacity, and delivery arrangements.
Start with the business question: what sales revenue is plausible on that day, given what we know today?
A useful forecast explains the central estimate, a range of possible outcomes, and the assumptions—not just a precise-looking dollar figure.
📦 Northstar’s planning decision
Imagine Northstar's operations team preparing for next Tuesday.
It needs a realistic view of demand before committing money and capacity—not a promise that one precise sales number will happen.
Northstar forecasting use case: the decision, forecast, assumptions, and action
🏪The business
Northstar is a fictional online retailer.
Daily sales vary with the customer base, weekday shopping habits, prices, promotions, holidays, and unexpected events.
🗓️The question
Given the information available today, what daily sales revenue is plausible next Tuesday?
State the date, revenue units, and assumptions; a forecast for next month answers a different question.
🎯One-number answer
‘We expect $32,000 in sales’ is a point forecast: one central estimate, not an exact outcome or a guaranteed minimum.
This is an illustrative teaching number, not a fitted Northstar business result.
↔️A more useful answer
Add a prediction interval to show plausible future daily sales under the model and assumptions.
A calibrated 80% procedure should cover about 80% of outcomes over repeated forecasts; individual misses are expected.
🚧What could change it?
An unplanned discount, an outage, or different customer demand could change sales.
Say which promotions and calendar events were known when the forecast was made, and use scenarios for uncertain inputs.
📦How the team uses it
Use the central estimate as a planning anchor and inspect low/high scenarios before committing capacity.
Staffing and stock decisions also need order counts, units, product mix, and costs—not revenue alone.
A forecast is a range: concepts and Northstar examples. Illustrative Northstar examples, not measured business results.
Concept
What it means
Northstar example
↔️Uncertainty belongs in the answer.
An 80% prediction interval is designed to cover 80% of future observations over repeated forecasts, if the model and its assumptions are appropriate.
It isn’t a guarantee.
🛒 Northstar compares a quiet Tuesday with a busy one rather than staffing as though exactly $32,000 must occur.
An interval is about possible sales, not a guaranteed minimum or maximum.
🗓️Further away usually means less certain.
There is more time for surprises to accumulate.
The fan below widens with the horizon; its center is not a promise.
🛒 Tomorrow's order pipeline is partly visible; a forecast six weeks out also depends on future promotions and demand changes.
The playground lets Northstar inspect how illustrative uncertainty grows with the horizon.
🧭Forecastability comes first.
An established store has repeat customers and years of history.
A brand-new product has much less evidence.
Neither deserves false precision.
🛒 Established categories have repeated weekly patterns and sales history.
A newly launched product deserves wider uncertainty and explicit assumptions, not the same confidence as a mature category.
FIG 01.1
The further ahead, the wider the range
WHAT THIS SHOWS
Northstar's central forecast is surrounded by illustrative ranges. Move the horizon or hide the bands; hiding uncertainty does not remove it.
The question it answers: Why is a single point estimate not enough for a planning decision?
Sales historyPoint forecastPrediction intervals
Point forecast$33.1K
Illustrative 80% interval$29K – $37.1K
Interval width$8.1K
Drag from 1 to 60 days, then hide the range. The line looks just as confident—but the uncertainty hasn’t disappeared. Bands here use a simple normal-innovation random-walk illustration (50%, 80%, 95%).
Fine print & assumptions
Constructed teaching data and illustrative calculations, not a production forecasting service. Read the diagram's method notes and distinguish supplied simulator components from fitted estimates.
CHECK YOURSELF · CHOOSE AN ANSWER
An 80% interval misses one day’s sales. Does that prove the model is wrong?