Simple Random Sampling: How to Pick a Fair Sample

You have a list of 800 customers, 650 invoices or 1,000 survey responses, and there is no time to look at all of them. Checking a sample is the obvious answer, but which ones? Picking the first fifty, the most recent or the ones that “look interesting” quietly skews the results. Simple random sampling solves this by giving every item the same chance of selection. Here is how to do it properly in a few minutes.

What simple random sampling means

In a simple random sample, every member of the population has an equal chance of being chosen, and every possible group of the chosen size is equally likely. It is the baseline method taught in statistics courses because it is easy to explain and easy to defend.

The key ingredients are a complete list (the sampling frame), a unique number for each item and a random source that nobody can influence. Draw without replacement so the same item is not picked twice.

Step-by-step method

1. Build the list. Export every item — customers, invoices, products — into a spreadsheet. Missing items cannot be sampled, so check the list is complete.
2. Number the rows. Give each item a number from 1 to N. With up to 1,000 items, a 1–1000 range fits perfectly.
3. Choose the sample size. Decide how many items you can realistically review.
4. Draw unique random numbers. Generate that many numbers without duplicates.
5. Match and record. Pull the rows with those numbers and save the list of drawn numbers with the date, so anyone can audit the process later.

Drawing the numbers with RandNow

The generator embedded above is preset to the range 1 to 1000:

1. If your list is shorter, change the range to match, for example 1 to 650. Numbers outside your list would waste draws.
2. Set How many? to your sample size — you can draw up to 1,000 numbers at once.
3. Keep No duplicates switched on (the default) so every row can be picked only once.
4. Choose Sort ascending to make the numbers easy to find in your spreadsheet.
5. Export the result as CSV, JSON or TXT and keep it with your notes. A share link records your settings.

Every number has exactly the same chance, and the randomness comes from the Web Crypto API in your browser, so nothing is stored and nobody can predict the picks.

How big should the sample be?

There is no single magic number. The right size depends on how precise you need to be, how varied the items are and how much time you have. As rough guidance:

• Spot checks such as a quick quality review often use a few dozen items.
• Surveys that report percentages need larger samples, and precision improves more slowly than you might hope — doubling the sample does not halve the error.
• Small populations sometimes justify checking everything instead.

If the results will support an important decision, ask someone with statistics training to set the size before you draw.

Biases that sneak in anyway

• Incomplete lists. If inactive customers were never exported, they cannot appear in the sample.
• Redrawing. Replacing an “awkward” pick with a new number breaks the randomness. Decide in advance how to handle unusable items.
• Non-response. In surveys, the people who answer may differ from those who ignore you.
• Sorting by importance first. Random numbers are fair only if the row order does not affect which items are eligible.

When the population falls into clear groups, such as regions or departments, consider sampling each group separately — a method called stratified sampling.

Other uses for unique random numbers

The same approach works for raffles with numbered tickets, choosing which pages of a long report to proofread or picking test cases from a numbered list. If your items are names rather than numbers, shuffling the list directly may be simpler:

Pro tips

• Freeze and save the numbered list before drawing so row numbers cannot shift.
• Write down your sample size and redraw rules before you press Generate.
• Sort the drawn numbers ascending to speed up finding rows.
• Keep the exported numbers as evidence that the sample was random.

Frequently asked questions