DWP: Universal Credit Advances Model
Risk assessment of Universal Credit Advances for fraud prevention
1. Summary
1 - Name
Universal Credit Advances Model
2 - Description
If you need help to pay your bills or cover other costs before you get your first Universal Credit payment, you can apply to get an advance. The amount you can get depends on your circumstances. The most you can get is the amount of your first estimated payment. The UC Advances model is a supervised machine learning classifier designed to risk assess requests for Advances ahead of payment.
3 - Website URL
N/A
4 - Contact email
Tier 2 - Owner and Responsibility
1.1 - Organisation or department
Department for Work and Pensions
1.2 - Team
Integrated Risk and Intelligence Service
1.3 - Senior responsible owner
Head of Integrated Risk and Intelligence Service
1.4 - Third party involvement
No
1.4.1 - Third party
N/A
1.4.2 - Companies House Number
N/A
1.4.3 - Third party role
N/A
1.4.4 - Procurement procedure type
N/A
1.4.5 - Third party data access terms
N/A
Tier 2 - Description and Rationale
2.1 - Detailed description
The UC Advances model is designed to risk assess requests for Advances ahead of payment. This approach enables the Department to focus its efforts on reviewing and verifying UC Advances assessed to have the highest risk of fraud.
The tool is a supervised binary classification machine learning model.
A risk- based approach minimises impact on the customer experience of legitimate claimants; optimises the use of taxpayers’ money in delivering the fraud control; and optimises the effectiveness of the control at tackling fraud.
2.2 - Benefits
The Department provided 1.4 million UC advances to new UC claimants in 2024/2025, with a total value of £0.8bn. As set out in the Official Statistics, we estimate that for 2024/ 2025 the monetary value of fraud and error on UC advances lies between £0m and £60m.
https://www.gov.uk/government/statistics/fraud-and-error-in-the-benefit-system-financial-year-2024-to-2025-estimates
Advances fraud was the subject of a National Audit Office report in March 2020 setting out the challenge of UC Advances Fraud. The machine learning model has been part of the response to this type of fraud and contributed to more than halving the estimated scale of the problem. The persistent nature of the fraud threat requires appropriate controls to tackle Advances fraud risk, such as the Advances fraud prevention model.
The performance information for 2024/25 demonstrates the model is around 3 times more effective at identifying fraud risk than a randomised control group sample. The outcomes from human decision makers in this stratified random sample are compared to the outcomes from Advances identified as high risk by the model. This performance comparison for 2024/25 demonstrates the model is around 3 times more effective at identifying fraud risk than the random sample control group.
2.3 - Previous process
DWP mandated all UC claimants to attend face to face appointments where circumstances would be verified before an Advance is paid. However, the model and other data-driven fraud controls have enabled a more efficient targetted approach.
2.4 - Alternatives considered
There are multiple fraud controls for Advances and these are complementary rather than alternatives to the tool.
Tier 2 - Deployment Context
3.1 - Integration into broader operational process
Advance requests identified as high risk by the model are referred to a DWP employee, who reviews all available and relevant information, to decide whether to approve or decline the request. The ultimate safeguard in place is that there is always a human intervention and decision, with no automated decision making by the model.
A blend of high risk and control group referrals are sent for human intervention to mitigate against human bias. In addition, the DWP employee delivering the intervention does not receive a risk rating in the referral nor are they made aware the referral has been generated by the model.
A decision to decline an Advance request does not prevent the same claimant from making a further Advance request and does not automatically result in the associated new UC claim being refused. Decisions on eligibility and entitlement for a new UC claim are a separate consideration.
3.2 - Human review
The model sends referrals in real time to a human decision maker who reviews all available and relevant information in advance of payment. This prevents incorrect overpayment and the associated debt recovery from the claimant.
There is a suite of safeguards to minimise the risk of unfair treatment or detrimental impact on legitimate customers, including:
- A blend of high risk and control group referrals are sent for human intervention to mitigate against human bias. In addition, the DWP employee delivering the intervention does not receive a risk rating in the referral nor are they made aware the referral has been generated by the model;
- A decision to decline an Advance request does not prevent the same claimant from making a further Advance request and does not automatically result in the associated new UC claim being refused. Decisions on eligibility and entitlement for a new UC claim are a separate consideration;
- Regular fairness assessment, including statistical analysis, is conducted to identify any concerns of unfair treatment or detrimental impact on customers;
- Analysis of any impact on payment timeliness on legitimate Advances requests is also conducted.
3.3 - Frequency and scale of usage
Withheld under FOI Act - S31
3.4 - Required training
N/A
3.5 - Appeals and review
There is no automated decision making. All payment decisions are made by a human and can be appealed as standard, with no distinct process for the model.
Tier 2 - Tool Specification
4.1.1 - System architecture
Withheld under FOI Act - S31
4.1.2 - System-level input
Real-time UC claim data.
4.1.3 - System-level output
Probability of UC Advance being fraudulent.
4.1.4 - Maintenance
Regular fairness assessment, including statistical analysis, is conducted to identify any concerns of unfair treatment or detrimental impact on customers.
Retraining the model is an activity the Department undertakes in the normal course of maintaining and improving the model to ensure it remains optimised to identify fraud risk. The model is retrained if there is a significant reduction in performance or if a fairness assessment identifies statistical disparities.
4.1.5 - Models
UC Advances model - a supervised binary classifier.
Tier 2 - Model Specification
4.2.1. - Model name
Withheld under FOI Act - S31
4.2.2 - Model version
Withheld under FOI Act - S31
4.2.3 - Model task
The model is designed to classify UC Advances requests as high risk of fraud or not.
4.2.4 - Model input
Features using claim characteristics and target variable based on historic Advances outcome data.
4.2.5 - Model output
A blend of high risk and control group referrals are sent for human intervention to mitigate against human bias. In addition, the DWP employee delivering the intervention does not receive a risk rating in the referral nor are they made aware the referral has been generated by the model.
4.2.6 - Model architecture
Withheld under FOI Act - S31
4.2.7 - Model performance
Performance monitoring of the Advances model confirms it is an effective fraud prevention control and more efficient than an untargeted approach. The model is around 3 times more effective at identifying high risk advances than a control group sample. It has delivered and continues to deliver measurable savings. Therefore, the model enables the Department to reduce fraud and protect the public purse effectively.
Analysis confirms the payment of Advance requests predicted as high risk by the model and subsequently approved by a human decision maker are not unduly delayed. The median payment delay is 1 day longer compared to Advance requests that are approved automatically, which is in line with the delay experienced by Advance requests that are subject to other fraud controls that are distinct from the model.
4.2.8 - Datasets and their purposes
Withheld under FOI Act - S31
2.4.3. Development Data
4.3.1 - Development data description
Withheld under FOI Act - S31
4.3.2 - Data modality
Withheld under FOI Act - S31
4.3.3 - Data quantities
Withheld under FOI Act - S31
4.3.4 - Sensitive attributes
See fairness analysis - statistical analysis pages - https://www.gov.uk/government/publications/fairness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2024-to-31-march-2025#:~:text=This%20fairness%20assessment%20on%20the,statistical%20disparity%20may%20impact%20claimants.
4.3.5 - Data completeness and representativeness
Withheld under FOI Act - S31
4.3.6 - Data cleaning
Withheld under FOI Act - S31
4.3.7 - Data collection
Withheld under FOI Act - S31
4.3.8 - Data access and storage
Withheld under FOI Act - S31
4.3.9 - Data sharing agreements
Withheld under FOI Act - S31
Tier 2 - Operational Data Specification
4.4.1 - Data sources
Universal Credit provides the data as part of an API call to the model.
4.4.2 - Sensitive attributes
See fairness analysis - statistical analysis pages - https://www.gov.uk/government/publications/fairness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2024-to-31-march-2025#:~:text=This%20fairness%20assessment%20on%20the,statistical%20disparity%20may%20impact%20claimants.
4.4.3 - Data processing methods
Withheld under FOI Act - S31
4.4.4 - Data access and storage
Withheld under FOI Act - S31
4.4.5 - Data sharing agreements
Withheld under FOI Act - S31
Tier 2 - Risks, Mitigations and Impact Assessments
5.1 - Impact assessments
It is the Department’s assessment that there are minimal concerns of discrimination, unfair treatment or detrimental impact on legitimate claimants arising from the Advances model. Therefore, it remains reasonable and proportionate to continue operating the Advances model as a fraud prevention control. https://www.gov.uk/government/publications/fairness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2024-to-31-march-2025
5.2 - Risks and mitigations
It is the Department’s assessment that there are minimal concerns of discrimination, unfair treatment or detrimental impact on legitimate claimants arising from the Advances model. Therefore, it remains reasonable and proportionate to continue operating the Advances model as a fraud prevention control.