Data processing:Data validation methods

Data validation methods

1 Batch total checks. In this, source documents are assembled in batches at the point at which they are col­lected, an average batch size being about 20. So clock cards might be batched, invoices batched, and so on.

Attached to the batch will be a batch control slip. At the point of collection, a clerk will calculate batch totals and write them on the slip. More than one batch total may be calculated. In the case of clock cards, for example, one batch total may be the hours worked as recorded on the cards in the batch, a second may be the sum of the workers' code numbers. In this example, the second total (the sum of the code numbers) has no use or meaning other than its use as a control total, and is called a hash total. The first total (the sum of the hours worked) is meaningful: it might, for instance, show the total hours worked in a certain department.

Each batch is passed to the data entry clerk, who keys the data on each document into the system. The clerk also keys in the batch totals, and the computer checks whether these correspond to the totals it works out from the keyed-in numbers. If they do not, it displays a warning message; either a document has gone missing from the batch or been overlooked, or the clerk has keyed in a number incorrectly.

2 Check digits. This check is often used with numeric codes. It involves appending to the code number a final digit that has a mathematical relationship with the other digits. The check digit is automatically worked out by the computer and appended when the code number is first created, and then checked each time the code is subsequently keyed in. In one method, the check digit is such that the following relationship holds:

1 x (right-most digit, i.e. the check digit) +

2 x (next digit) +

3 x (next digit) +

4 x (next digit) + ...

equals a number which is divisible by 11.

Suppose, for example, a code number is 32305. Then X, the check digit that is to be appended, is calculated as shown in Figure 9.2. The result, shown in this figure, is 50 + X, and this must be divisible by 11. Therefore, X must be 5, and the code number with the check digit appended will be 323055.

You might like to confirm that the final digit in the ISB number on the back of this book is calculated in this way.

3 Range check. This involves a check by the computer that the data lies within certain limits. In the case of clock card data, for example, the computer might check that the weekly hours worked by an employee are less than 60, and display a warning message if the clerk keys in a number greater than this.

4 Parity check. This is different type of check, applied by the computer to check that errors do not arise during processing. It applies this check each time it moves data around within its store. To each byte that makes up the data, it adds an extra parity bit, which will be either a binary 0 or a binary 1 depending on whether the number of 1s in the byte is even or odd. The presence of this bit enables it to check that a bit is not altered when data is moved around. If it is, it repeats the operation. Of course, in the unlikely event of two bits being altered to give a compensating error, this test will fail.

Data verification methods

As we have said, data may be valid and yet wrong. Data verification involves checking that the keyed-in data is actually correct.

1 Informal check. The data entry clerk can informally check the accuracy of his or her work. The computer can assist in this, for in response to keying in the code number of a customer, supplier, or part, for example, it may display the name on the screen.

2 100% check. This is the main type of verification check, and involves keying the data twice, first by one clerk and then by another. The computer checks the second set of entries against the first, and displays a warning message if a discrepancy occurs.

Files

In a computer system, data is stored in files. Chapter 6 explained the main computer filing terms and concepts, including fields and records (see page 130). In data process­ing systems, there are two kinds of file:

• Transaction files, which hold the records of business transactions (such as sales and purchases). The data on these records is called transaction data. Such data is added to a transaction file when (or shortly after) the transaction takes place, and it is deleted from the file after the transaction has been processed.

• Master files, which hold stock records, employee rec­ords, and other non-transaction records. The data on these records, being more permanent than transaction data, is called standing data. Some standing data, such as names and code numbers, hardly ever changes, and this is called reference data. Qther standing data, such as stock balances, may be updated during processing; this is called master data.