vaayaa Inclusive Education
Technical risk assessment

Data Analytics Applications Risk Assessment: Data Quality, Privacy and Model Risks

A worked example of how an analytics team finds, scores and controls the risks in a data project, from poor data quality and analytical errors to privacy, bias, access, supplier licensing and model drift.

Updated November 2026 8 hazards assessed Review every 6 months Highest risk: High to Medium
See the Data Analytics Applications course

Executive summary

This assessment shows how an analytics team can find, rate and control the risks that matter most when it turns data into decisions. Eight risks were chosen because they are plausible in everyday analytics work. Poor data quality and analytical errors are the most likely, because they come from how data is collected and handled. Privacy breaches and model bias are less likely but can harm people and the organization seriously. Each risk is scored before controls and again after the existing and added controls. Most risks fall to Medium or Low once added controls are in place, while bias and privacy stay at Medium because the impact on people stays serious. The first actions are data quality rules, reproducible analysis, a purpose record for each data set and a check of outcomes by group. The team is a teaching example, so every score illustrates method, not a real organization.

Focus area
Data analytics applications
Setting
Analytics team in a mid-sized organization that reports on customers, operations and models
Method
Five by five likelihood and impact risk register

Scope

This assessment covers an analytics team that collects, prepares, analyzes and reports data and runs a small number of predictive models. It looks at risks to accuracy, people, privacy, continuity and trust. It does not cover the design of a specific product or the finances of the organization. Data types, tools and jurisdiction are assumptions made for teaching.

Context

The team has an analytics manager, analysts, a data engineer, a data quality lead and a privacy officer. Data flows from internal systems, purchased data sets and public sources into a warehouse, where it is cleaned, analyzed and shown in dashboards and reports. A few models score customers or forecast demand. Each risk is scored before controls to show its natural level, then again after the existing controls and the controls that should be added.

Objectives, assets and stakeholders

The objective is to produce accurate, repeatable and lawful analysis that people can trust and act on, while protecting the individuals behind the data. The assets in scope are the data sets and the warehouse, the pipelines and dashboards, the models and their code, the analyst skills and the trust of customers and decision makers. Stakeholders include the analysts and engineers, managers who use the results, customers and staff whose data is used, data suppliers, regulators and auditors. Clear roles matter: every risk has one owner role, and the analytics manager keeps the register current and reports to senior management.

Risk matrix: before and after controls

Each cell shows hazard numbers placed by likelihood and impact. Compare the two grids to see how controls move each hazard.

Before controls

Inherent risk

Risk before controls: rows are likelihood from almost certain to rare, columns are impact from insignificant to catastrophic. Cells list hazard numbers.
1
Insignificant		 2
Minor		 3
Moderate		 4
Major		 5
Catastrophic
5 Almost certain
4 Likely Likelihood 4, impact 3, score 12, high: H06 Likelihood 4, impact 4, score 16, high: H01 H02 H05
3 Possible Likelihood 3, impact 3, score 9, medium: H07 Likelihood 3, impact 4, score 12, high: H08 Likelihood 3, impact 5, score 15, high: H03 H04
2 Unlikely
1 Rare

After controls

Residual risk

Risk after controls: rows are likelihood from almost certain to rare, columns are impact from insignificant to catastrophic. Cells list hazard numbers.
1
Insignificant		 2
Minor		 3
Moderate		 4
Major		 5
Catastrophic
5 Almost certain
4 Likely
3 Possible
2 Unlikely Likelihood 2, impact 2, score 4, low: H06 H07 Likelihood 2, impact 3, score 6, medium: H01 H02 H05 H08 Likelihood 2, impact 4, score 8, medium: H03 H04
1 Rare
  • Low 1 to 4
  • Medium 5 to 9
  • High 10 to 16
  • Critical 17 to 25

Hazard register

A summary first, then the detail for each hazard.

Summary of hazards with scores before and after controls
No.HazardBeforeAfterOwner
H01 Poor data quality and sampling bias 16 High 6 Medium Data quality lead
H02 Analytical errors and weak methodology 16 High 6 Medium Analytics manager
H03 Model bias and unfair outcomes 15 High 8 Medium Model owner
H04 Privacy and consent breaches 15 High 8 Medium Privacy officer
H05 Uncontrolled access, retention and deletion 16 High 6 Medium Data governance lead
H06 Pipeline and data source failure 12 High 4 Low Data engineer
H07 Third-party data licensing and supplier risk 9 Medium 4 Low Procurement lead
H08 Misread results and weak validation or monitoring 12 High 6 Medium Analytics manager
H01

Poor data quality and sampling bias

Technology and data Before 16 After 6

Missing values, duplicates, wrong units and samples that do not represent the group being studied produce results that look precise but describe the wrong thing.

Who or what is affected: Analysts, decision makers and customers

Causes

Data collected for another purpose, manual entry and samples chosen for convenience.

Consequences

Wrong conclusions, wasted effort and decisions that harm the people the data describes.

Existing controls

  • Basic checks for blanks and duplicates run on each load.
  • The analyst notes where the data came from.

Control gaps

Quality checks stop at format and do not test whether the sample is representative.

Additional controls

  • Prevent: Define data quality rules for completeness, range and format and check them automatically on every load.
  • Detect: Compare the sample with the wider group and record any gap before results are shared.
Likelihood 4 x impact 4 before; 2 x 3 after Owner: Data quality lead Target: Within 45 days Early warning: A rise in rejected or corrected records from one source.
H02

Analytical errors and weak methodology

Operational Before 16 After 6

Wrong joins, formula mistakes, wrong test choices and results that cannot be repeated mean that findings change when someone reruns the work.

Who or what is affected: Analysts, managers and clients

Causes

Time pressure, undocumented steps and analysis done by hand in spreadsheets.

Consequences

Incorrect findings, lost credibility, rework and wrong decisions.

Existing controls

  • A colleague reads the main figures before a report goes out.
  • Code and workbooks are saved in a shared folder.

Control gaps

Review checks the figures, not the steps that produced them.

Additional controls

  • Prevent: Keep analysis in version-controlled code or notebooks with documented steps and a peer review.
  • Detect: Rerun key results from raw data on a second machine and compare them.
Likelihood 4 x impact 4 before; 2 x 3 after Owner: Analytics manager Target: Within 45 days Early warning: A result that cannot be reproduced from stored data and code.
H03

Model bias and unfair outcomes

Reputational Before 15 After 8

A model trained on skewed or historic data can treat groups differently, for example in scoring, offers or screening, and the harm may stay hidden because overall accuracy looks good.

Who or what is affected: People affected by the model and the organization

Causes

Historic data that reflects past bias, missing groups in training data and no testing by group.

Consequences

Unfair treatment of individuals, complaints, regulator interest and loss of trust.

Existing controls

  • Overall accuracy is reported for each model.
  • A manager approves a model before it is used.

Control gaps

Fairness is discussed in principle but no measure or limit is set.

Additional controls

  • Prevent: Test results by group and set limits on acceptable difference before a model goes live.
  • Detect: Monitor outcomes by group each month and review any gap with the model owner.
Likelihood 3 x impact 5 before; 2 x 4 after Owner: Model owner Target: Within 60 days Early warning: A gap in outcomes between groups that is larger than the agreed limit.
H04

Privacy and consent breaches

Legal and compliance Before 15 After 8

Personal data is collected or used beyond what people agreed to, or is combined in a way that identifies them, which breaks privacy duties and trust.

Who or what is affected: Individuals whose data is used and the organization

Causes

Re-use of old data sets, unclear consent wording and joining data from several sources.

Consequences

Regulator action, fines, complaints and damage to trust.

Existing controls

  • A privacy notice is published on the website.
  • Data is used for projects the team has been told about.

Control gaps

Consent is recorded at collection but not checked when data is re-used.

Additional controls

  • Prevent: Record the purpose and legal basis for each data set and remove or mask identifiers that are not needed.
  • Detect: Review each new project for privacy impact before data is released to analysts.
Likelihood 3 x impact 5 before; 2 x 4 after Owner: Privacy officer Target: Within 45 days Early warning: A project that uses personal data without a recorded purpose.
H05

Uncontrolled access, retention and deletion

Technology and data Before 16 After 6

Too many people can see sensitive data, extracts sit on laptops and shared drives, and records are kept long after they are needed or are not deleted when they should be.

Who or what is affected: Data subjects, analysts and the organization

Causes

Convenience copies, rights that are never removed and no owner for deletion.

Consequences

Data leaks, breach notices and larger harm from any incident.

Existing controls

  • Access to the main database needs a login.
  • Old extracts are deleted when someone notices them.

Control gaps

There is no routine scan for extracts stored outside the managed database.

Additional controls

  • Prevent: Grant access by role with a review each quarter and set a retention period for each data set.
  • Detect: Log data exports and scan shared locations for unmanaged copies.
Likelihood 4 x impact 4 before; 2 x 3 after Owner: Data governance lead Target: Within 60 days Early warning: A data extract found outside the approved storage.
H06

Pipeline and data source failure

Operational Before 12 After 4

A feed breaks, a schema changes, a scheduled job fails or a sensor or system stops sending data, and dashboards show stale or partial numbers without any warning.

Who or what is affected: Analysts, managers and customers

Causes

Changes made upstream without notice, no checks on volume and a single person who knows the pipeline.

Consequences

Stale dashboards, wrong decisions, lost time and loss of trust in the data.

Existing controls

  • Jobs send an email when they fail.
  • The engineer restarts a failed job on request.

Control gaps

Failures are reported but a quiet drop in volume is not.

Additional controls

  • Detect: Add freshness and volume checks that alert the owner when data is late or unusually small.
  • Recover: Keep a runbook and a tested way to reload missing data from the source.
Likelihood 4 x impact 3 before; 2 x 2 after Owner: Data engineer Target: Within 45 days Early warning: A dashboard that shows data older than its agreed refresh time.
H07

Third-party data licensing and supplier risk

Supplier Before 9 After 4

Purchased or scraped data may be used outside its licence, the supplier may change terms or stop service, and a project can depend on a source that cannot be relied on.

Who or what is affected: The organization, suppliers and clients

Causes

Terms that are not read, data shared with partners and single-source dependence.

Consequences

Claims for misuse, forced removal of results and a stopped project.

Existing controls

  • Contracts for paid data are filed.
  • Public data sources are named in the project notes.

Control gaps

Licence terms are filed with the contract but not linked to how each data set is used.

Additional controls

  • Prevent: Keep a register of data sources with licence terms, permitted uses and renewal dates.
  • Respond: Identify a substitute source for each critical feed and test it once a year.
Likelihood 3 x impact 3 before; 2 x 2 after Owner: Procurement lead Target: Within 90 days Early warning: A data set in use with no recorded licence or expiry date.
H08

Misread results and weak validation or monitoring

Strategic Before 12 After 6

Results are shown without uncertainty, a model drifts as conditions change, and decision makers act on a chart or score that no longer holds.

Who or what is affected: Decision makers, customers and the organization

Causes

Pressure for a simple answer, no monitoring after launch and changing conditions.

Consequences

Poor decisions, financial loss and wasted effort.

Existing controls

  • Results are presented with a short written summary.
  • Models are retrained when someone requests it.

Control gaps

Performance is checked at launch and then only on request.

Additional controls

  • Prevent: Show uncertainty and limits with every result and agree what the result can and cannot support.
  • Detect: Track model performance against fresh data each month and set a trigger for review.
Likelihood 3 x impact 4 before; 2 x 3 after Owner: Analytics manager Target: Within 60 days Early warning: Model accuracy that falls below the agreed threshold on new data.

Rating method and evaluation

Each risk is rated on two five-point scales. Likelihood runs from Rare to Almost certain, and impact runs from Insignificant to Catastrophic. Multiplying the two gives a score from 1 to 25, which falls into one of four bands: Low from 1 to 4, Medium from 5 to 9, High from 10 to 16 and Critical from 17 to 25. The score is a guide for ordering work, not a measurement. Inherent risk is the score before any controls. Residual risk is the score after the existing and added controls. Impact is judged on harm to people first, then on decision quality, cost, legal effect and trust, so a privacy or fairness harm is not outweighed by a small gain in speed. Anything still High needs senior attention and a dated plan, anything Medium needs routine controls with a named owner, and anything Low is accepted and watched. The people who do the work should challenge the ratings, and they are reviewed whenever an indicator reaches its threshold.

Likelihood, impact and rating bands

Likelihood

  • 5 Almost certain: Expected to happen in most circumstances.
  • 4 Likely: Will probably happen in many circumstances.
  • 3 Possible: Could happen at some time.
  • 2 Unlikely: Not expected, but it could happen.
  • 1 Rare: Only in exceptional circumstances.

Impact

  • 1 Insignificant: No lasting effect. Handled within routine work.
  • 2 Minor: Small effect that is quickly corrected.
  • 3 Moderate: Noticeable effect that needs managed effort to correct.
  • 4 Major: Serious effect on people, delivery or reputation.
  • 5 Catastrophic: Severe or lasting effect that threatens the whole activity.

Bands (likelihood x impact)

  • Low (1 to 4): Accept and monitor.
  • Medium (5 to 9): Manage with routine controls.
  • High (10 to 16): Senior attention and a treatment plan.
  • Critical (17 to 25): Act first. Do not proceed until reduced.

Monitoring, escalation and review

Review this assessment every 6 months, and sooner when an indicator below reaches its threshold.

Early-warning indicators and escalation thresholds
Indicator to watchEscalate when
Share of records rejected or corrected on each data load, by source.Escalate to the data quality lead when a source rises above its normal level, and to the analytics manager if it persists for a week.
Key results that cannot be reproduced from stored data and code.Escalate to the analytics manager at once, and hold the report until it is resolved.
Outcome differences between groups for each live model.Escalate to the model owner when the gap passes the agreed limit, and to senior management if it is not fixed in a month.
Data sets in use with no recorded purpose, licence or retention period.Escalate to the privacy officer when any are found, and restrict access until the record is complete.
Data feeds that are late or smaller than normal.Escalate to the data engineer at once, and to the analytics manager if dashboards stay stale for a day.

Contingency arrangements

If a serious event occurs, such as a data breach, a faulty model that affects people or a wrong figure that has been published, the analytics manager pauses the affected report or model and informs senior management. The privacy officer assesses what must be reported under the rules that apply and tells affected people when required. The data engineer restores data from backup or the source and reruns the analysis. Corrected results are shared with a plain note of what changed, and the cause is reviewed before work resumes.

Prioritized action plan

  1. Set data quality rules and automatic checks on every load, owner data quality lead, within 45 days.
  2. Move key analysis into versioned and reviewed code with a rerun check, owner analytics manager, within 45 days.
  3. Record purpose and legal basis for each data set and mask unneeded identifiers, owner privacy officer, within 45 days.
  4. Test each model by group and set a limit on acceptable difference, owner model owner, within 60 days.
  5. Add freshness and volume alerts to every pipeline, owner data engineer, within 45 days.

Conclusion

Good analytics rests on a few steady habits: checks on data quality, documented and repeatable steps, a clear purpose for personal data, access by role and monitoring after launch. The highest scores sit with data quality, analytical errors, privacy and bias, and each has actions that bring the residual score down. Treat the register as a living record, review it on the stated cycle, and seek qualified advice for any real privacy, legal or model fairness question.

Assumptions and limitations

Assumptions

  • The organization holds personal data about customers and staff and is subject to privacy law.
  • Analysis is done by trained staff on managed computers and a central data warehouse.
  • Jurisdiction is not specified, so legal duties are described in general terms.

Limitations

  • The scores are examples for learning and do not describe any real team or data set.
  • Privacy, consent and data rules differ by country and need specialist review.
  • This is educational content and does not replace a formal assessment or professional advice.

Related frameworks to study

  • Data governance and data quality frameworks
  • Privacy by design principles
  • Risk register with likelihood and impact scoring

Technical Risk Assessment

China, India, United States, Indonesia, Pakistan, Nigeria, Brazil, Bangladesh, Russia, Japan, Mexico, Philippines, Vietnam, Egypt, Turkey, Germany, United Kingdom, France, Italy, Canada, South Korea, Australia, Spain.

For Learning Reference Only: This assessment demonstrates how risk assessment principles are applied in a practical setting, providing a useful reference for learning and understanding. Actual assessments may be adapted to suit specific site conditions and circumstances.

Keep learning

Study the subject behind this assessment.

Data Analytics Applications course Read the best practices for this course

More Technical risk assessments

Hourly Active: 40,959 Learners