What it means
A representative sample is a subset of a larger group, called the population, that reflects the population's mix of important features. If 30% of a company's invoices come from overseas suppliers, then roughly 30% of the invoices in the sample should too.
Finance teams rely on samples because checking everything is often too slow or too costly. Auditors test a slice of transactions, marketers survey a slice of customers, and credit teams review a slice of loans, then draw conclusions about the lot.
The usual way to get a representative sample is random selection, where every item has an equal chance of being picked. If the population has distinct groups, stratified sampling (splitting the population into groups first and then sampling each group in proportion) stops a small but important group from being missed.
In auditing, monetary unit sampling gives larger balances a higher chance of selection, so the biggest items are rarely skipped. Sample size matters as much as the method.
A larger sample narrows the margin of error (the range within which the true figure probably sits), but the gain shrinks as the sample grows, so quadrupling the size only halves the margin of error. The most common failure is bias, where the way items are chosen systematically favours some over others.
Surveying only the customers who reply to an email, or testing only the invoices a manager hands over, produces a sample that looks convincing but is not representative. A sample can also be the right size and still mislead if the population has changed since it was drawn.
In practice
Real-world examples.
Example
An external auditor of a regional retail chain draws 400 sales receipts at random from 120,000 receipts across 60 stores. Each store appears roughly in proportion to its sales, so no single branch dominates the test. The error rate found in the sample is then used to judge whether revenue is fairly stated.
Example
A subscription software company wants to know why customers cancel. It surveys 500 customers chosen at random from each pricing tier in proportion to tier size, rather than the first 500 who answered an email. The result mirrors the real customer base, so the product team can trust the churn reasons.
Example
A bank reviewing its $2,000,000,000 small business loan book for credit quality pulls 250 loans, making sure each industry and region is included in proportion. The credit team uses the arrears found in the sample to set the loan loss provision for the whole book.
Formula
Calculation
Required sample size n = (z x z x p x (1 - p)) / (e x e), where z is the confidence factor (1.96 for 95% confidence), p is the expected proportion of errors (use 0.5 if unknown) and e is the margin of error you will accept.
Suppose a finance team wants to test 20,000 supplier invoices worth $10,000,000 in total, at 95% confidence with a margin of error of 5%. Then n = (1.96 x 1.96 x 0.5 x 0.5) / (0.05 x 0.05) = (3.8416 x 0.25) / 0.0025 = 0.9604 / 0.0025 = 384.16, which rounds up to 385 and which the team rounds again to 400 for a tidy test. If 8 of the 400 invoices contain an error, the error rate is 8 / 400 = 2%. Projecting that to the population gives an estimated misstatement of 2% x $10,000,000 = $200,000, assuming the errors are spread evenly by value.Case study
Seen in the real world.
Harbourview Foods is an illustrative, fictional packaged food distributor that checked inventory by counting only the shelves nearest the warehouse door. Counts looked tidy for years, and management assumed the stock records were accurate.
When a new finance controller redesigned the count, she numbered every storage location and used a random number list to pick 300 of them, with extra picks for high-value pallets. The new sample found that slow-moving items stored at the back of the warehouse were overstated by about 6%, which the old convenient sample had never touched.
Extending the 6% finding across the $8,000,000 stock balance suggested an overstatement of roughly $480,000. The illustrative lesson is that a sample is only as good as the way it was drawn, and convenience is the enemy of representativeness.
Watch out
Common mistakes.
- Picking the items that are easiest to reach, such as the first 50 invoices in a folder, and treating them as if they speak for the whole population.
- Believing that a bigger sample always fixes a badly chosen one, when a large biased sample is still biased and only gives false precision.
- Forgetting that important sub-groups, such as large customers or high-risk regions, can be missed entirely unless the sample is deliberately structured to include them.
Questions
People also ask.
How large does a representative sample need to be?
It depends on the confidence level, the margin of error you will accept and how variable the population is, but a few hundred well-chosen items is often enough for large populations.
Is a random sample always representative?
Not always, because chance can produce an unlucky draw, which is why stratifying the population and checking the sample against known totals is good practice.
Can a sample be representative of a population that has changed?
No, a sample mirrors the population at the time it was drawn, so if prices, customers or systems have shifted since, the sample may no longer describe today's reality.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%