Key findings
The main source of information for this report is the annual quality evaluation conducted among all producers of official statistics. A self-assessment covering all statistics in the National Programme for Official Statistics 2024–2027 (Statistics Norway, 2024a), together with information from follow-up meetings with those responsible for the statistics, constitutes the factual basis for the evaluation.
The results of the quality evaluation show a good level of compliance with many of the quality requirements set out in the Norwegian Statistics Act and the European Statistics Code of Practice (Eurostat, 2017). In most areas, the results are broadly at the same level as last year. In some areas, improvements are evident. The share of statistics that now describe uncertainty and potential sources of error on the statistics’ website has increased from 75 to 90 per cent. Statistics Norway has shared its template and guidelines for the user-oriented documentation “About the statistics” with other producers, and descriptions of sources of error and uncertainty are included in this template. Many producers outside Statistics Norway are working to improve their user-oriented documentation and to adapt it to the template. Several are also considering adopting PxWeb as a dissemination platform, in line with the recommendations from last year’s report.
In other areas, there is still a need for stronger awareness and compliance. This applies, among other things, to the quality requirement on statistical confidentiality. Insufficient use of recognised methods and software to ensure statistical confidentiality entails a risk of either overly strict disclosure control, which may reduce the relevance of the statistics, or overly weak control, which may lead to disclosure of individual units in statistical dissemination.
The use of quality indicators to measure various aspects of compliance with quality requirements remains low. Only 27 per cent report that such indicators are used in the assessment of the production process and output quality, which is the same level as in previous years. Measures such as publishing a note on recommended indicators and offering courses have so far not had a visible effect. For Statistics Norway, the transition to the new cloud-based data platform (Dapla) has also created challenges, due to both time pressure and to immature functionality for quality measurements. At the same time, there are good examples of indicators being used effectively. Such experiences should be shared more widely. Producers recognise the value of quality indicators, but both awareness and competence vary considerably across environments. In the longer term, numerical quality indicators should constitute a larger part of the basis for quality reporting, as they enable more continuous, automated and precise monitoring of quality.
This year’s evaluation examines how the transition to Dapla affects compliance with quality requirements. The transition to Dapla represents a comprehensive modernisation of statistical production at Statistics Norway, with the aim of achieving more efficient processes, improved data retrievability and increased reuse of code. Approximately one third of the statistics have now been partially or fully migrated to the cloud platform, and several benefits have been reported, in particular, improved overview of production workflows, greater automation and improved code quality. Many also report better documentation and opportunities to clean up existing routines, as well as increased competence and opportunities to adopt new methods such as machine learning. At the same time, Statistics Norway’s statistical divisions point to several challenges, including missing or immature tools on Dapla, frequent changes in systems and standards, and reliance on legacy solutions that have not yet been phased out. Time pressure, shortages of competence and vulnerability when only a few individuals master new technologies add further strain. Although quality is largely maintained, some report that analytical work is being deprioritised and that the transition period is long and demanding.
Overarching recommendations and status of improvment measures
In addition to the need for continued work on statistical confidentiality and quality indicators, this report highlights a number of new recommendations. Many of these concern increased and improved logging, including corrections of errors in published statistics, to monitor developments over time. Another new recommendation relates to privacy and entails that statistics with personally identifiable information in the underlying data must either pseudonymise such information or document exemptions from the requirement to store such information separately from other data.
The implementation of improvement measures is a central part of the quality system, and statistical producers annually assess which new measures are necessary based on the recommendations. In March 2026, the status for 315 measures was reported, of which 57 per cent of the measures planned since 2022 have been completed. Although many measures have been implemented, there remain areas—particularly related to quality indicators and confidentiality—where further efforts are required. The measures vary in scope and impact, and counting measures alone does not provide a complete picture of quality improvement but must be supplemented by qualitative assessments.
Developing the quality system
The quality system for official statistics is continuously developed to adapt to technological and societal changes. Artificial intelligence (AI) is becoming increasingly important in the production of official statistics, and the current quality framework was developed before AI gained widespread application. Producers are now using AI in both production and support processes, which require new interpretations of quality requirements and increased emphasis on traceability, transparency and control. The AI Act (EU Artificial Intelligence Act, 2026) introduces a risk-based regulatory framework that will also apply in Norway. This implies that producers of official statistics must identify, document and classify the use of AI, particularly those that may be considered high risk. This requires a coherent framework for the use of AI in statistical production and updated guidelines that ensure quality, compliance and trust.
