What it means
When running a business, checking every single detail is rarely practical. If you want to know if your suppliers are overcharging, reviewing one million invoices takes too much time.
Instead, you pick a smaller, manageable subset to review, and this subset is your sample size. By looking at these specific items, you can draw conclusions about the entire batch.
Getting the number right is vital. If your selection is too small, your results might be skewed by random flukes, and if it is too large, you waste precious time and money.
In finance, sample sizes are used constantly during audits, quality control checks, and market research. For instance, an internal auditor might test fifty expense reports out of five hundred submitted that month.
If they find three errors in that group, they can estimate the error rate for the whole month and decide if further investigation is needed. The key is ensuring your selection truly mirrors the wider population, meaning every item has an equal chance of being chosen.
Non-finance managers often struggle with how many items to review. While there is no magic number, statistical tools can help calculate the ideal count based on how certain you need to be and how much risk you can tolerate.
A higher certainty requirement demands a larger group. Understanding this balance helps you defend your findings to senior leaders or external regulators without spending unnecessary weeks gathering data.
Imagine cooking a large pot of soup. You do not need to eat the entire pot to know if it needs more salt.
You simply stir it well, take one spoonful as your sample, taste it, and judge the whole pot based on that single bite.
In practice
Real-world examples
Example 1
An online startup reviews fifty customer support chat logs out of two thousand monthly interactions to measure overall service satisfaction levels. Because the logs are picked at random, the 2.5% sample gives a fair view of the whole month without reading every conversation.
Example 2
A local bakery tests the weight of twenty loaves of bread daily from a production run of four hundred to ensure consistent quality. If more than one or two loaves are underweight, the baker checks the scales before the next batch is made.
Example 3
A regional transport firm audits twenty-five driver fuel receipts each week out of a total pool of five hundred to spot potential discrepancies. A pattern of missing or altered receipts would trigger a wider review of that depot.
Think of it
Imagine cooking a large pot of soup. You do not need to eat the entire pot to know if it needs more salt. You simply stir it well, take one spoonful as your sample, taste it, and judge the whole pot based on that single bite.
Formula
Calculation
Sample Size = (Z x Z x p x (1 - p)) / (Margin of Error x Margin of Error), where Z is the confidence score and p is the expected proportion of items with the issue (use 0.5 when unsure, as it gives the largest sample).
Example: For 95% confidence (Z = 1.96), p = 0.5 and a 5% (0.05) margin of error, the calculation is (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 items.
If the whole population is small, adjust it: Adjusted Sample = 385 / (1 + (385 - 1) / Population). For a population of 2,000 invoices, that is 385 / (1 + 384 / 2,000) = 385 / 1.192 = about 323 invoices.Case study
Seen in the real world
This is a fictional, illustrative scenario. Oakwood Retail, a mid-sized clothing chain with ten stores, wanted to understand why stock shrinkage, or unaccounted inventory loss, had risen. Examining every single item across all warehouses and shops would have taken months and cost thousands in labour. Instead, the operations manager decided to take a sample. They selected twenty high-value product lines across all locations, representing five hundred individual items.
Upon physical count, they discovered a consistent discrepancy in how returned goods were logged at two specific branches. Armed with this targeted finding, the finance team audited those two stores further and uncovered a training gap in the till-refund process. By using a sensible sample size, Oakwood fixed the leak quickly without halting daily trade. The count took two people three days rather than the several weeks a full stocktake would have needed. Because the sample was drawn from every store, the team could be confident the pattern at the two branches was real and not a coincidence.
Watch out
Common mistakes
- Choosing a group that is too small because it is easier, leading to unreliable results.
- Picking only convenient examples, such as the most recent transactions, which introduces bias.
- Assuming a larger sample size is always better, even when the extra cost outweighs the benefit.
Questions
People also ask
How do I know what sample size to use?
You can use standard statistical calculators that factor in your total population, your desired confidence level, and your acceptable margin of error.
Can my sample size be too large?
Yes. While larger groups reduce error, checking too many items wastes time and money without providing significantly better insights.
What happens if my sample is biased?
A biased group does not truly represent the whole population, which means your financial conclusions or audit results will be incorrect.