You can estimate before you know.
The aim is not to guess a famous number. It is to build a short chain of decisions that another person can inspect.
Start with a quantity, not an operation
Before reaching for a calculator, describe what your answer will count. “How much water?” could mean litres falling in a storm, water stored in a tank or daily drinking demand. All three concern water, but they need different models. Write the object, place, time period and unit in one sentence.
A useful scope statement for a photo archive is: “Pixel data for 5,000 uncompressed RGB images, in decimal gigabytes.” It excludes several plausible interpretations before any arithmetic. If a detail would change the result by a large factor, make it explicit.
Choose a few anchors you can defend
An anchor is a starting quantity, such as a room’s approximate floor area or the distance covered by one wheel revolution. It can be measured, documented or assumed. Label which it is. A remembered fact is not automatically more reliable than a carefully stated assumption.
Use two or three decisive anchors first. Adding ten poorly known factors can make the answer look sophisticated while increasing uncertainty. If you split a population into groups, their definitions must not overlap; if you multiply by a participation fraction, check that the original total does not already count participants.
Write the structure before choosing numbers
- Stock: how much exists at one moment? A room’s air mass is volume × density.
- Flow: how much happens over a period? A device’s energy use is average power × operating time.
- Balance: what remains after arrivals and departures? Pending orders = initial orders + arrivals − deliveries.
- Cycles: how many repeated movements make the total? Wheel revolutions = route length ÷ rolling circumference.
- Transformation: how much input produces the output? Flour demand and grain demand are connected by extraction yield, not by assuming they are identical.
The shape of the model matters more than the final decimal. Choosing a stock model for a throughput question can produce a neatly calculated answer to the wrong question.
Let units catch your mistakes
Write units beside intermediate numbers. Converting millimetres to metres before calculating a volume avoids a thousandfold error. Kilowatts describe power; kilowatt-hours describe accumulated energy. A percentage needs a denominator, and bits are not bytes. These distinctions often matter more than refining an anchor by ten per cent.
Try a second plausible scenario
Suppose one uncertain anchor could reasonably be half or twice your first choice. Recalculate only that anchor and see what changes. In a simple product, doubling one factor doubles the answer. Changing all factors at once hides which uncertainty matters. Keep quantities fixed when the question explicitly fixes them.
An optimistic and a pessimistic scenario are not automatically a confidence interval. They are choices that show sensitivity. Communicate the base case and the reason for your alternatives rather than presenting many significant figures as certainty.
Check the scale and explain your result
Compare your result with something familiar. Is the classroom’s air measured in grams or hundreds of kilograms? Does the photo library resemble a few messages or a substantial backup? A comparison is a check, not proof: if it seems surprising, inspect the model instead of forcing the answer to feel familiar.
Finish with one sentence: “Under these assumptions, about X units; Y is the most uncertain anchor.” Then explore the worked examples or use the estimation lab to compare guesses and change an image-storage scenario.