We have been providing practical sampling solutions to a variety of questions from the banking sector. Questions such as:
- How many bank accounts should be sampled to obtain a reasonable estimate of statement error? If no errors are found within the sample, what does this mean?
- What is the best estimate of bad debt provision from a large population of accounts with a range of different credit histories?
- How can we optimise a cross-selling opportunity within our existing database by predicting who will be more likely to purchase?
- Where is the source of our most profitable customers and how can I describe them?
In some of the above examples, Stratified Random Sampling (SRS) was used with the primary objective of obtaining the most precise estimate. In SRS, the strata are defined and a random sample is taken from within each stratum. The principle of stratification is to partition the population in such a way that the units within the stratum are as similar as possible to yield the most efficient estimate. The main considerations are to decide how many strata to use, how to describe the strata, and the number of accounts to sample from within each stratum.
In order to optimise the sample allocation, the aim is to allocate a fixed total sample size among the strata in order to minimise the margin of error (i.e. the bands around the bad debt provision estimate). The theory of stratified sampling suggests that more accounts be sampled from strata where there is larger variability and larger population sizes.
The theory for allocating the sample size across the strata is used as a guideline for deciding on the total sample size taking into account other issues such as time, cost and practicality.
If you would like to find out more about how we approach statistics programmes for the banking sector, contact us and let’s begin the discussion.