Sample Selection From Large Data For Audit

In Insurance service various risks are covered underpopulation consists of all the claims; paid or denied or
terms and conditions mentioned in Insurance Policiespending; that became part of the system. Larger the
and in case of claims arising, they are managed orpopulation, larger is the sample size if rest of the
settled accordingly. It is a common practice to auditfactors remain unchanged.
the procedures for underwriting, policy management,2. Variance: This is a measure of variation in the data.
claims management etc. to ensure that healthyIf there is no variation then by checking any one
practices are followed by the organization. Since theitem we can form accurate opinion about the entire
number of policies and claims are usually very high, itlot of item. But, if variations are there, we need to
is undesirable to audit each and every record.check more. Larger the variation, larger is the sample
Sampling techniques are used to reduce the volumesize. Variance and standard deviation both are
of audit work without compromising on quality.measures of variation in the data and represent the
Sampling techniques have been in use in varioussame property of the data.
areas. Pathologists take a sample of blood from a3. Confidence Level required: Suppose we audit 10
human body and test that sample to get an ideaclaims cases and find that in 3 cases a business rule
about all of the blood. Similarly, while boiling rice, wewas not taken care. Further, in 20 audited claims
test a few grains of rice as a sample to concludecases we find this number to be 6. We may conclude
about the entire lot of rice. We know that thisthat in 30% of the cases this rule has been ignored.
technique takes us to the right conclusion. PropertiesBut, how much confident we would be on this
of blood are uniform for pathological tests beingconclusion? If we audit more cases and find similar
conducted. Rice being boiled too is uniformly heatedresults, the confidence level on the conclusion will be
by keeping the size of the container and volume ofhigher. So, If the confidence level required is high, the
water appropriately large. Due to this uniformity, asample size required too is high.
small sample size is sufficient for reaching to the right4. Accuracy: Higher the accuracy required, bigger the
conclusion. If the rice is boiled with lesser water thansample size should be.
what is required for uniform flow of heat, the riceAs it is clear that a sampling technique and the
doesn't get cooked uniformly. In such cases, testingsample size cannot be decided without proper
few grains from top and forming opinion about theunderstanding of the nature of data. An auditor
entire lot may lead to wrong conclusion. Similarly, ifcannot do anything with the population data and
the items being cooked together are not similar thenvariance in that data because these data are
again we need more samples to draw any acceptablegenerated through various operational activities
conclusion.performed in the organization. However, if a very
In Insurance industry, the data related to varioushigh level of accuracy and confidence level is set, the
functions have variations because of its basic naturesample size required becomes very high. It is
of dealing with uncertainties. Keeping the sample sizerecommended to keep these values realistic to keep
large for auditing increases the cost of audit. On thethe cost and time acceptable.
other hand keeping it low may lead to invalid orA good practice is to do a small initial study to
wrong conclusion. Hence, it is important to chooseunderstand the population size, variance and basic
correct sampling technique and sample size.nature of the data. Then use them to compute the
Using sampling techniques effectively is very easy.sample size and conduct the detailed study. While
But it is not that easy that we pre-decide the samplepicking the records for audit, the sample should
size without knowing the nature of data. Sample sizerepresent the population without any bias. For this,
depends on following four factors:generating random numbers and associating them
1. Size of the population: If we want to drawwith actual records by using the statistical distribution
conclusion about claims management system, theof data gives the best result.