Overview of Changes to the UK and Canada Models

The 2026 Canada release and the 2026 UK release – including scores for England, Wales, Northern Ireland, and Scotland – were revised in several important ways. 

These updates reflect methodological enhancements added to the UK and Canada crime risk model, improving geographic precision, score differentiation, and comparability across regions. These include:

1. Changes to aggregation levels. In the UK, we moved from the ward level to the lower statistical output area (LSOA) level, while Canada has moved from a mix of Census tracts and dissemination areas (DAs) to use only dissemination areas. In the UK, moving from wards – of which there were 8,600 in the country – to LSOAs – of which there are 43,000 – means a more granular evaluation of crime risk. Similarly in Canada, moving from census tracts to dissemination areas in metropolitan places means parsing data into 57,000 dissemination areas vs just 6,000 census tracts – this change leads to greater precision in the CAP Scores.
2. Reweighting the inner and outer rings to match the US model. With more weighting on the inner ring, there is greater internal consistency to the UK and Canada crime risk model and improved interpretability across markets.
3. Refining the population of the lowest geographic levels that affect the radii and contouring (gradient maps). This improvement makes the model more accurate in terms of its geographic computations for the distance-relative-to-population calculation (i.e., how far out each methodology goes until it hits a population cap) and the way the gradient maps are shading the scores on the backend. This change will affect some site scores in larger, populous, urban areas.

To illustrate the impact of these improvements, the image below shows the same site in London (UK) using the larger wards from the 2024 release (left image) as compared to the disaggregated LSOAs from the 2026 release (right image). Note that the volume of crime did not change much around this site – that is, that the level of crime stayed generally stable in terms of nearby incidents, but the units into which those incidents are tallied changed.

The key finding from this comparison is that the use of wards (2024) makes the CAP scores appear more middle-of-the-road than the same volume of crime using LSOAs (2026). In essence, aggregating smaller geographies that are both very safe and very risky into the same ward ends up “washing out” important underlying variation in those smaller geographies. For example, the black box in the upper-left corner of the first image shows a ward with a CAP score of 218 in 2024; however, the same area in 2026 displays a combination of low-risk neighborhoods (CAP scores between 70 and 100) and high-risk neighborhoods (CAP scores above 500). The same is shown for the box in the center of each image: a ward with a CAP score of 332 in 2024 contains smaller geographies with CAP scores between 124 and 1000 in 2026, the latter of which is the maximum score for the UK.

While the image comparison reveals the importance of geographic disaggregation, small changes to emphasize the weight of the inner ring relative to the outer ring are also at play here. For example, the overall CAP score in 2024 would include a highest ward score of 422 in the inner ring and outer ring to produce (after population weighting) an overall CAP score of 329. The same site in 2026 would include scores up to 1000 in the inner and outer ring to produce an overall CAP score of 697. Again, the same volume of crime each year yields very different CAP scores because there is much more variation in the underlying geographies (LSOAs) and because the rings have been reweighted.

Comparison of changes between Canada and the UK

Changes to the scores in the Canada model are slightly more pronounced than changes to scores in the UK model. This is mostly because (a) Canada has many geographic units with lower residential populations, especially in more rural places throughout the country; (b) there are more rural geographic units given the larger size of Canada overall; and (c) the disaggregation of units overwhelmingly affected metropolitan areas in Canada where each census tract was divided into several dissemination areas (vs the disaggregation in the UK that was more universal across all community types).

An examination of scores across methodologies reveals two different patterns of change for the two country models. In Canada, the number of client sites experiencing a CAP score change increases from Max-1 to Max-12 (i.e., with increasing distance) shows the number of sites with significant changes in scores increases as the methodology distance increases. In other words, there are more sites with large score differences at Max-12 than at Max-1. In contrast, in the UK, the opposite pattern emerges and there are fewer client sites with large score changes at Max-12 vs Max-1. The higher concentration of urban areas in the UK vs Canada is likely the main cause of this difference. Because the UK has more densely populated urban areas and fewer widely dispersed rural regions, geographic conditions tend to be more consistent across locations. As a result, increasing methodology distance smooths differences rather than amplifying them, leading to fewer large changes in CAP scores.

Summary

The 2026 updates to the UK and Canada crime risk model introduce finer geographic resolution and refined weighting, revealing meaningful variation in risk that was previously obscured by larger aggregation units.

Explore the updated scores to see how enhanced geographic detail impacts your sites.