The importance of data integrity and data governance
Topic 12 is the first topic of Unit 3, an A Level unit examined by WIT13: a two-hour written paper, answered on paper with no computer and no calculator.
Everything from the AS units is assumed, especially the Topic 5 vocabulary of tables, keys and relationships.
Manipulating data has been the heaviest topic on recent WIT13 papers, about a quarter of the marks, and this first section sets up the idea the whole topic serves: data an organisation can trust.
Your syllabus expects you to understand the importance of data integrity and the methods of maintaining it (data governance).
(this applies to every section of this unit)
- The paper: 2 hours, 80 marks, six compulsory questions, no sections, no choice, no calculator, no data files. Each question opens with a one-line context; parts add their own short scenarios. Allow about 1.5 minutes per mark.
- Short items: Give, State and Name are recall worth 1 or 2 marks, one mark per distinct point. A stem that prints one example and asks for "one other" or "two other" gives nothing for repeating it.
- Explain [2] and Describe [2]: always 2 marks: one statement plus one linked expansion. Explain adds a reason ("because", "so that"); Describe develops the point, or gives one step per mark for a process.
- Complete: tables (one mark per correct cell) and part-drawn diagrams (one mark per mark point). The stem says exactly what to add; extra symbols or keys lose marks.
- Create [6] or [12]: third normal form tables or a Gantt chart. Draw [6] or [10]: an interface design or a flowchart.
- Levels items: Discuss [6] and Evaluate [12], judged on accurate knowledge applied to the organisation, developed points on both sides and, for Evaluate, a conclusion that follows.
- Diagrams and models carry about 40% of the paper. Topic 12 supplies the normalisation and logical data model items.
The idea: data you can trust
Lark Hill Veterinary Group runs four surgeries from one shared database.
A receptionist books a dog in for a vaccination; a vet records the dose; the pharmacy system reduces the vaccine stock; the accounts system sends the bill.
If any one of those records is wrong, missing or out of step with the others, the next person acts on bad data: the reminder goes to the wrong owner, the stock count says there is vaccine when the fridge is empty, the owner is billed twice.
Data integrity is the reliability, or trustworthiness, of data across its whole life: from the moment it is captured, through every transfer and process, to the day it is archived or deleted.
Data with integrity is complete, consistent wherever it is held, and has not been changed except by authorised people following the rules.
The examiners are strict about the word: integrity is not the same as accuracy.
A database can hold an owner's phone number that passes every check, matches in every table and has never been tampered with, yet is wrong because the owner gave an old number.
Integrity is about whether the data can be relied on as a whole; accuracy is about whether one value is true.
- Data integrity: the reliability and trustworthiness of data throughout its life cycle, so that it stays complete, consistent and unchanged except by authorised, recorded processes.
- Data consistency: related data held in more than one place, or in related tables, is changed together, so every copy and every link agrees.
- Referential integrity: the database rule that a foreign key value must match an existing primary key value in the related table; Microsoft Access uses this name for the setting that keeps related tables consistent.
- Data governance: the overall management of an organisation's data: the policies, standards, roles and controls that decide who owns each data set, who may create, change or delete it, and how its quality is checked.
- Audit trail: a record of who changed which data, when, and what the old and new values were.
Why integrity matters
Organisations make decisions, send bills, treat patients and report to regulators using stored data.
When integrity fails, the damage spreads, because one bad value is copied, summed and acted on many times.
- Decisions: managers read reports built from the data. A stock report that double-counts deliveries leads to under-ordering.
- Customers and safety: a vet who sees the wrong allergy record may give a harmful drug.
- Money: duplicated or lost transactions mean wrong bills, refunds and chasing payments.
- Law and reputation: data protection law in many countries requires personal data to be accurate and, where necessary, kept up to date (in the UK this is one of the principles of the UK GDPR). A breach or a run of errors damages customers' trust.
- Efficiency: staff who do not trust a system keep their own spreadsheets, which creates more copies and more inconsistency.
How integrity is lost
Integrity can be damaged at every stage of the data's journey.
The examiners credit specific causes, not database vocabulary.
Stage
Cause of weak integrity
Example at Lark Hill
Input
human input or transcription errors
a receptionist types a microchip number with two digits swapped
Input
missing validation in the program
the system accepts a blank owner phone number
Transmission
errors in transfer that go undetected
a lab result file is corrupted on its way from the laboratory
Processing
bugs in programs or scripts
an update routine changes every dog's weight instead of one
Storage
hardware failure
a disk fails part-way through writing a record
Storage and access
malware, hacking, unauthorised changes
an attacker or a careless user edits or deletes records
Copies
the same data held in several places and updated in only one
the owner's new address changed in billing but not in reminders
How integrity is maintained: data governance
Data governance is the "who and how" behind the controls.
It is a management framework, not one piece of software.
A governance policy at Lark Hill would say:
1. Ownership and roles. Each data set has a data owner (the practice manager owns client records) and day-to-day data stewards who keep it in good order. Responsibility is clear, so errors are fixed, not passed around.
2. Standards. Agreed formats, field names and definitions, held in the data dictionary (next section), so every surgery records a weight in kilograms and a date the same way.
3. Access control. User accounts with permissions by role: receptionists can change contact details but not clinical notes; only vets can record a dose. Fewer people able to change data means fewer accidental or malicious changes.
4. Validation and verification at input. Validation rules reject impossible values; verification (double entry, or reading back to the owner) catches values that are possible but mistyped.
5. Database controls. Primary keys stop duplicate records; referential integrity stops a treatment record pointing at an animal that does not exist; normalisation stores each fact once, so one change updates it everywhere; transactions make a multi-step update all-or-nothing, so a stock reduction and a bill are both written or neither is.
6. Error detection in transfer. Checksums and other error checks with automatic resending.
7. Audit trails and monitoring. Every change logged with user, time and old value, so mistakes can be traced and reversed.
8. Backup and recovery. Regular, tested backups so data damaged by hardware failure or ransomware can be restored to a known good state.
9. Training and review. Staff taught the rules, with regular data-quality checks (reports of blank fields, duplicate owners) and a named person to act on them.
- Say integrity means the data is reliable and trustworthy; do not say it means "the data is accurate". The examiners repeat this in every series.
- Say consistency when related data is changed together in every place; do not use it as another word for accuracy.
- Say a cause (transcription error, a bug, malware, hardware failure, an undetected transmission error); do not answer with a database term such as "redundancy" or "replication" on its own. Those scored nothing.
Connecting it
Every other section of Topic 12 is a method of maintaining integrity: the data dictionary sets the standards, validation guards the input, normalisation removes the copies that drift apart, and Big Data raises the problem to a scale where veracity (trustworthiness) becomes one of its named issues.
Topic 13 adds file security, encryption and the DBMS functions (record locking, access control) that protect integrity in use, and Topic 14's disaster recovery and archiving keep it over time.
In WIT14 you will set referential integrity and validation rules in Access yourself.
Lark Hill's manager finds that some client records are wrong because receptionists typed them incorrectly.
Give two further reasons why the integrity of the records could be weakened.
Answer:
1. A bug in a program that updates the records (1)
2. Malware or a hacker altering or deleting data (1)
Also creditable: hardware failure while data is being written; errors in transmission that are not detected; no validation on input.
Not credited: typing errors (given in the stem); "redundancy" or "replication" (database terms, not causes).
Explain one reason why data integrity is important to Lark Hill Veterinary Group.
Top answer: Vets decide on treatment using the records in the database (1) so if an animal's allergy or dosage history cannot be relied on, a vet could give a drug that harms it (1).
Weak answer: "So that the data is correct."
This defines integrity as accuracy, which the examiners reject, and gives no consequence for the surgeries.
0 or at most 1.
Explain how an audit trail helps Lark Hill maintain the integrity of its client records.
Answer: The audit trail records which user changed a record, when, and the old value (1) so an incorrect or unauthorised change can be traced to the person and the original value restored (1).
Describe one feature of a data governance policy for the surgeries.
Answer: It names a data owner for each data set, such as the practice manager for client records (1), who is responsible for its quality and decides which staff roles may create, change or delete those records (1).
When a WIT13 item asks about integrity, use the word "reliable" or "trustworthy" in your first sentence, then link it to what goes wrong for the organisation in the stem.
A first sentence that says "accurate" puts you on the wrong definition from the start.
- Treating integrity and accuracy as the same. A value can pass every check and still be untrue; integrity is the wider idea of trustworthiness and consistency.
- Listing database words ("normalisation", "redundancy") when the question asks for a cause. Name what goes wrong: a person, a program, a device, an attacker.
- Writing "use validation" for every method. Validation checks input only; governance also covers ownership, access rights, audit trails, transactions and backups.
- Confusing Access's "referential integrity" setting with data integrity as a whole. Referential integrity is one method that supports consistency between related tables.
Check your understanding
1. A garden centre's sales records have errors because a till program has a bug. Give two other ways the integrity of the sales data could be weakened. (2) Total [2]
2. A student writes: "Data integrity means the data is accurate, so a record that passes validation has integrity." Explain why the student is wrong. (2) Total [2]
3. A dental practice gives every receptionist the same login, which can edit and delete patient records. Explain one way this weakens data integrity. (2) Total [2]
4. Describe how record-level access rights help maintain the integrity of a hospital's patient data. (2) Total [2]
5. State the meaning of the term data consistency. (1) Total [1]
6. Lark Hill Veterinary Group is writing a data governance policy for its four surgeries. Discuss why Lark Hill needs data governance and what its policy should contain. (6) Total [6]
Answers
1. Any two (1 each): human input or transcription errors; malware or hacking; hardware failure while writing data; undetected errors in transmission; missing or poor validation; unauthorised changes by staff. Do not award a program bug (given in the stem) or database terms alone.
2. Validation only checks that a value is reasonable and follows the rules, so a wrong but sensible value still passes (1); integrity is about whether the data as a whole is reliable and trustworthy, kept complete and consistent and changed only by authorised processes, which validation alone cannot guarantee (1).
3. Changes cannot be traced to a person, because every receptionist uses the same account (1) so a mistaken or deliberate deletion cannot be found and reversed, and nobody is accountable for the records (1). Also accept: anyone can edit clinical data they should not change, so errors enter records that should be locked to clinicians.
4. Each user is given rights by role, so only clinicians can change treatment notes while clerks can only read them (1), which reduces the number of people who can alter or delete clinical data by accident or on purpose (1).
5. Related data held in more than one place or table is changed together, so every copy agrees. (1)
6. Levels answer (6). Level 3 (5-6) covers both strands in context. Need: four surgeries share one database, so a careless change in one affects all; vets rely on the records for treatment; data protection law requires accurate personal data; without clear owners, errors are passed around and never fixed. Features: named data owners and stewards for client, clinical and stock data; standards in a data dictionary (weights in kg, one date format); access rights by role (only vets record doses); validation and verification at reception; audit trails reviewed by the manager; tested backups; training and regular data-quality reports. Level 2 (3-4): accurate points on need or features with some context, but one strand thin. Level 1 (1-2): generic statements about keeping data safe.