NorwayPublic sector2025 – ongoingNorwegian Tax Administration
Machine-learning scores in a tax-return reminder experiment
Norway's tax administration reported around a 10-percentage-point increase in filing probability from a model-stratified contact experiment
Problem
Following up late tax returns by telephone consumed staff time. The administration needed evidence about which businesses responded to contact.
Approach
A 2025 randomized experiment grouped sole proprietorships by scores from an existing filing model. Staff attempted calls and sent SMS messages when contact failed. Analysts compared filing outcomes within score groups to assess differences in response.
Results
Takeaway
The measured improvement came from attempted contact, including SMS. Savings from concentrating calls on selected score groups remained estimates.
Categories
- Industry
- Public sector
- Business function
- Finance, risk and legalCustomer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
The administration reported a randomized contact experiment analyzed using machine-learning scores. Filing probability increased by around 10 percentage points after adjustment, while response varied across score groups.
Full details are in the original sources.
