Data Capture and Error Checking

How Data Is Captured

Data capture refers to the methods and tools used to collect data and get it into a computer system. The method chosen has a big impact on accuracy, speed and cost. Manual methods include keyboard entry (typing data into a form or spreadsheet) and handwriting on paper forms that are later typed up. These methods are flexible but slow and prone to human error. Automated methods capture data without (or with minimal) human typing. A barcode scanner reads product codes at a supermarket checkout. A magnetic stripe or chip on a payment card is read by a card reader. Optical Character Recognition (OCR) software converts printed or handwritten text in a scanned image into editable digital text. Optical Mark Recognition (OMR) detects pencil marks in fixed positions on a form — used in multiple-choice exam answer sheets and lottery tickets. Sensors capture environmental data automatically: a temperature sensor on a weather station records readings every minute without any human involvement. Online forms with drop-down menus and checkboxes guide users toward valid entries, reducing errors compared to completely free-text entry.

No data capture method is perfect. Even automated systems can misread a damaged barcode or a smudged form. The data that enters a system must be checked for errors before it is stored or processed, otherwise the resulting information will be unreliable. Two distinct processes are used for this: validation and verification.

Validation

Validation is an automated check carried out by the computer to test whether data is reasonable and in the correct format. It is important to understand what validation cannot do: it cannot confirm that the data is factually true. For example, if a date-of-birth field expects a date in the past and a user types 01/01/2010, validation will accept it as a plausible date — even if the real date of birth is 01/01/2009. Validation simply checks that the data could be correct.

There are several types of validation check:

A range check confirms that a numeric value falls within acceptable limits. For instance, a student's exam mark out of 100 must be between 0 and 100 — the system would reject -5 or 150. A type check (also called a data type check) confirms that the data entered matches the expected data type for that field. A name field expects text; if a user enters numbers, the type check rejects the entry. A presence check ensures that a required field has not been left blank — useful for mandatory fields such as a customer's surname or email address. A format check (also called a picture check) verifies that the data matches a specific pattern. A UK postcode must follow a pattern such as SW1A 1AA; a date must follow DD/MM/YYYY. A length check confirms that the entry is within an acceptable number of characters — a mobile phone number in the UK should be 11 digits, so the system would reject an entry of 8 or 15 digits. Finally, a check digit is an extra digit added to a code (such as an ISBN book number or barcode) that is calculated mathematically from the other digits. When the code is entered, the system recalculates the check digit and compares it to the one entered — a mismatch reveals a keying error.

Verification

Verification checks that data has been copied or transferred correctly from one place to another. Unlike validation, verification does not check whether the data is reasonable — it checks whether it has been faithfully reproduced. There are two main verification methods. Double entry (also called double keying) requires a user to type the data twice — the computer compares the two entries and flags any difference. This is commonly used when setting a new password: you type it once, then again to confirm. If the two entries do not match, you are asked to try again. Proof reading is a human review of data before it is confirmed — a member of staff reads through entered data and compares it against the original source document to spot any errors. This is less automated but can catch subtle mistakes that a computer might miss.

Sources of Error

Errors can enter a system at multiple points. Transcription errors occur when data is copied from one place to another — a typist accidentally hits "8" instead of "9", or two adjacent characters are swapped (a transposition error). Transmission errors happen during the transfer of data over a network — interference or signal degradation can corrupt data in transit. Original source errors occur when the original data is simply wrong from the start — a customer gives an incorrect telephone number, or a sensor is miscalibrated. Data loss can occur during processing if a system crashes before data is saved. Deliberate falsification is where a person intentionally enters false information — for example, giving a false age on a website form. No single checking technique can prevent all of these errors, which is why systems combine multiple approaches.