Problems with incorrect data
Errors in data can cause all sorts of problems such as:
- Incorrect decisions being made resulting in loss of money
- Goods being sent to the wrong address
- Having to spend time sorting out mistakes
- Loss of goodwill
- Loss of trust
- Being prosecuted under the Data Protection Act of 1988 for not keeping personal data accurate.
How data errors can occur during transcription
Transcription errors are mistakes humans make when either keying in data or filling in forms such as optical mark forms. Often they occur through carelessness and are not picked up by verification methods such as proof reading. Transcription is the process of copying data from one source document such as an order form or an application form. Transcription errors are those errors introduced during the keying process.
Avoiding transcription errors:
- Avoid the use of keyboarding as a data input method
- Use automated methods of data entry (e.g. barcoding, OMR, etc)
- Staff should be trained to be accurate
- Problems must contain appropriate validation checks.
How data errors occur during input
Input data even though it has been verified (i.e. checked against the source documents) and validated, can still be incorrect. Incorrect data can occur in a variety of ways and it is usually any human involvement at either the data collection stage or the input stage that provides the weak link in the system. The only way to avoid errors is to keep human involvement to a minimum and therefore use direct methods of data capture such as MICR, OCR, barcoding, etc. wherever possible.
Keyboarding can produce a large number of errors during input. Even automated methods of data entry such as OMR or MICR can have problems with errors. For example, a form may not be read by the reader, because it may not have been marked correctly. It is important that the system soes not try to guess at what a mark shoudld be, as this would introduce incorrect data. Instead, the system should reject the form and leave it for a human to decide what to do.
Barcode readers in supermarkets make a beep that tells the operator that an item has been scanned correctly. This means that the person can can the products in again if they do not hear the beep, or if that fails, manually enter the number at the bottom of the barocde.
How data errors occur during processing
Mistakes that occur during processing include:
- A programming error that remained undiscovered during testing and only come to light when a series of conditions apply. (i.e. stop running)
- The use of the wrong version of a data file for processing data rather than the latest version. For example, you could use an older version of a spreadsheet by mistake or you could use the wrong set of data to perform a mail merge.
- Incorrect formulas in a spreadsheet that were not deleted and corrected during the testing, means that incorrect processors are being performed, leading to wrong information being output.
- Damage by viruses - viruses are able to delete data or otherwise render it unreadable.
- Equipment malfunction - hard disk drives are like any mechanical device in that they breakdown occasionally. This can cause the loss of data, so it is important to always back up files containing programs and data.
How errors occur during transmission
When data is passed through a communication medium (wireless, metal wire, optical wire), it is important that the data is not corrupted in any way and if it is, it is equally important that this is detected and the data is retransmitted.
The checking of data after it has passed along a communication line is performed using a parity check. Parity checking works in the following way. The computer adds up the number of bits in one byte and if the parity is different to the parity setting, the computer will report and error. It is possible to use either even or odd parity. Taking odd parity, for instance, suppose we are sending the letter C along a communication line. In ASCII code, the series of bits used to represent C is 1000011. Since there are three 1s in this code and odd parity is being used, a 0 is added to the left hand side of the group of bits so that the total for the byte is odd. If even parity were being used, a 1 would need to be added so that the total for the byte would then be an even number.
Communication devices have a chip inside them to deal with parity checks: the sending device adds the parity bits and the receiving units calculates what the parity bit should be. If an error has occurred, transmission parity will no longer be observed, and the corruption is detected. The problem with parity checks is that if more than one error occurs and the errors compensate for each other, parity can still appear to be correct.
Validation is a check performed by a computer program during data entry.
The purpose of validation is the more validation there is in place, the less likely ridiculous data is to be processed.
Validation is the process which ensures that data accepted for processing is sensible and reasonable. Validation is performed by the computer program being used and consists of a series of checks called validation checks.
- Range Check - The data must be a value that falls between two predefined values, e.g. Date of Birth of a student in a school would have to be between 01/09/89 and 31/08/96, any other date should be rejected.
- Presence Check - A presence check requires that a value must be entered, e.g. every patient in a hospital needs an admission number, if the person inputting information about the patient leaves that specific box empty, the computer should tell the user to fill in that field.
- Format Check - The data must conform to a prescribed layout (specific numbers or letters), e.g. Date of Birth being 01/01/1980 means the format is dd/mm/yyyy, any other format should be rejected.
- Data Type - The data must be of a specific type (number, text, boolean, etc.), e.g. data entered for "No. of stock" must be a number. Any other data type should be rejected.
- Fixed Value / Drop down - The data must conform to one of the values in a predefined list, e.g. In a gender field, you can only be Male or Female.
- Check Digit - This is a way of checking for transcription errors in long numbers (ISBN code of a book), the check works by performing a mathematical calculation on the code number and returning an error if expected value isn't the same as actual value.
Verification - checks that the data being typed in matches exactly the data on the document used to supply the information.
The purpose of verification is to ensure the accuracy of information.
- Double Entry - Two different people enter the same data into two different computers. The verification program then compares the two sets of data and identifies the mistakes.
- Sending Back Printouts - This is when a company would send printouts of inputted data to the data source and ask them to confirm the accuracy of the data.
- Proof Reading - Proofreading is reading a proof copy of text for the purpose of detecting errors.