WGS1 Early Detection Pilot Scheme

Benefits Department

On this page, you will find information about the ‘WGS1 Early Warning Experiment’ algorithm. It implements the ministerial regulation under the Municipal Debt Assistance Act (Wgs). This is a joint experiment by the Tax and Customs Administration, the Benefits Service and ten local authorities to identify people with payment arrears at an earlier stage and provide them with assistance.

Last change on 19th of May 2026, at 13:02 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: DPIA, IAMA
• Status: In use

General information

Theme

Public finances

Begin date

6-2025

End date

7-2027

Contact information

[e-mail]

Link to publication website

https://www.toeslagen.nl/hulp-na-aanmaning

Responsible use

Goal and impact

The Tax and Customs Administration and the Benefits Service are working with ten local authorities to identify people with payment arrears and those at risk of developing problematic debt at an earlier stage, and to help prevent those arrears from increasing further. This is a two-year pilot scheme.

The pilot involves the sharing of information (subject to conditions) under the Wgs. The aim of the pilot is to reach citizens with (imminent) problematic debts who were not previously identified by the local authority and to offer them support. The pilot will investigate whether signs of payment arrears are a good indicator of (imminent) problematic debts.

For the implementation of the experiment, the WGS1 Early Warning Experiment algorithm is used to make the correct selections of payment arrears that meet the criteria. Municipalities have a maximum number of alerts they can process, which is why a random selection takes place in the final phase of the algorithm. The algorithm for the WGS1 Early Warning Experiment selects people with payment arrears exceeding a certain amount after a reminder has been sent. They live in one of the ten participating local authorities. These people are first offered assistance by the Tax and Customs Administration or the Benefits Service, and if that offer of assistance has no effect, their details may be shared with their local authority for further assistance.

The algorithm relates to the implementation of the ministerial regulation drawn up for this experiment under the Wgs. The link to this regulation:

https://zoek.officielebekendmakingen.nl/stcrt-2025-17636.html

Considerations

This algorithm is essential for the successful implementation of this experiment. It enables the targeted selection of the appropriate citizens.

The WGS1 Early Warning Experiment is important for citizens and business owners (natural persons) because offering timely assistance can resolve (impending) problematic debts and prevent the need for debt counselling.

The algorithm ensures that we can select people who are in arrears and have an outstanding reminder without a payment plan, meaning there is a risk that the arrears could lead to problematic debts.

A ministerial regulation has been drawn up for the experiment (see reference).

This algorithm has been developed within the framework of Early Warning. This framework is derived from the Municipal Debt Assistance Act.

Human intervention

No human intervention is involved in the operation of the algorithm when it comes to deciding whether or not a member of the public is included in this experiment. The algorithm assesses, on the basis of predefined objective selection criteria, whether a payment arrears is included in the selection. This has no legal consequences for the citizen. The selection process leads solely to a non-binding offer of assistance. First from the Tax and Customs Administration or the Benefits Service, and subsequently from the local authority where the citizen resides. The offer of assistance is made by the Tax and Customs Administration and the Benefits Service by sending a letter and telephoning the citizen. If the citizen does not wish to make use of the offer of assistance and/or does not wish for data to be provided to the local authority in this context, they may indicate this.

Risk management

The General Administrative Law Act (Awb) requires that the actions of the government be transparent and lawful. The Tax and Customs Administration observes the general principles of good governance when applying and developing algorithms.

The algorithm uses data collected under various tax laws. As required by the GDPR, no more data is used than is necessary. No special categories of personal data are used within the scope of Article 9 of the GDPR.

The selection rules are periodically reviewed and adjusted where necessary to ensure continued compliance with legislation and regulations.

Project staff undertake the national AI training.

Citizens can submit requests for access or rectification via the standard GDPR channels of the Tax and Customs Administration and the Benefits Service.

Legal basis

The legal basis for the use and collection of data is set out in the ministerial regulation drawn up under the experimental provision of the Municipal Debt Assistance Act https://zoek.officielebekendmakingen.nl/stcrt-2025-17636.html

Links to legal bases

• Ministerial Regulation under the Municipal Debt Assistance Act of the Ministry of Social Affairs and Employment.: https://zoek.officielebekendmakingen.nl/stcrt-2025-17636.html
• The local authority receives the data for the purposes of implementing Article 3(1)(b) of the Local Authority Debt Assistance Act.: https://zoek.officielebekendmakingen.nl/stcrt-2025-17636.html
• Article 2.5 of the MR establishes the legal obligation (Article 6(1)(f) of the GDPR) for the Tax and Customs Administration/Benefits Service to provide the data.: https://avgb.nl/art-6-avg/

Elaboration on impact assessments

Privacy and the GDPR

The use of data must be assessed against the General Data Protection Regulation (GDPR). By assessing personal data, any privacy risks are identified and appropriate measures can be taken.

The GDPR stipulates that no more data may be used than is necessary. This is known as data minimisation. The Tax and Customs Administration regularly assesses whether the data used is still necessary and may therefore be used.

The GDPR compliance document has been drawn up. GDPR-related risks have been identified and recommendations have been followed up with measures.

Equality and non-discrimination

The algorithm is assessed in line with applicable non-discrimination principles for direct and indirect discrimination. By processing as little personal data as possible, the risk of direct discrimination is reduced. Staff involved in the development and management of the algorithms receive training on data protection and bias.

Impact assessment

• Data Protection Impact Assessment (DPIA)
• Human Rights and Algorithms Impact Assessment (IAMA)

Operations

Data

1. Municipality: to determine whether someone resides in one of the ten participating municipalities (source: Personal Records Database (BRP))
2. The amount of the payment arrears from the early warning signal for taxes or benefits: method, amount and status (source: Tax and Customs Administration)
3. Opt-out registration, where a person has indicated that they do not wish their data to be shared with the local authorities (source: HM Revenue & Customs)
4. Availability of a telephone number (source: Tax and Customs Administration and Benefits Service)

Links to data sources

• Basic Register of Persons (BRP): https://www.rijksoverheid.nl/onderwerpen/privacy-en-persoonsgegevens/basisregistratie-personen-brp
• Tax and Customs Administration: www.belastingdienst.nl
• Tax and Customs Administration: www.belastingdienst.nl
• The Tax and Customs Administration and the Benefits Service: www.belastingdienst.nl, www.toeslagen.nl

Technical design

The algorithm was developed by staff at the Tax and Customs Administration and is maintained internally.

The algorithm consists of selection rules drawn up by subject matter experts based on legislation, regulations and their expertise.

The algorithm assesses, on the basis of the selection rules, whether or not the payment arrears are included in the selection. The algorithm is not self-learning. This means that the algorithm does not develop itself whilst in use.

The algorithm selects payment arrears that meet the criteria set out in the ministerial regulation, including: the citizen lives in one of the ten participating municipalities; in the case of taxes, this concerns turnover tax, payroll tax and income tax; the threshold amount for taxes is 600 euros; for benefits, this concerns repayments of all types of benefits (housing benefit, childcare allowance, child-related budget and healthcare allowance); the threshold amount for benefits is 500 euros; a reminder has been sent and no payment arrangement has been made.

Citizens who are registered with the local authority due to an ongoing debt assistance programme and whose details are known to the Tax and Customs Administration are not included in the trial.

Citizens for whom no telephone number is known are also excluded. Citizens who are still in the assessment process for the resolution of the childcare allowance affair will also not be selected for this experiment.

A citizen will be included in the selection with a maximum of one payment arrears. If they indicate that they do not wish to participate in this experiment, they will not be included in subsequent selections. Municipalities have a maximum number of alerts they can process; therefore, a random selection takes place in the final phase of the algorithm.

Following selection, citizens first receive a letter containing information from the Tax and Customs Administration or the Benefits Service and are contacted by telephone with an offer of assistance. Before data is provided to the local authority, the algorithm is used to check whether the payment arrears still meet the criteria. Arrears that no longer meet the criteria are removed from the selection. This may be because, for example, the person has since moved house, has since paid, has agreed a payment plan, or the telephone number is no longer in use.

A check is also carried out to see whether someone has indicated via an opt-out form that no data may be shared with the local authority, or whether someone has indicated during the outbound call campaign that they do not wish to participate; in these situations, no data is provided to the local authority as part of the experiment.