Matt Ashby
LightNews LightNews
banner
mattashby.com
Matt Ashby
@mattashby.com
1.5K followers 330 following 1.2K posts
I help people use data to reduce crime. Associate Professor, Crime Science, UCL. Former police officer.

🌐: mattashby.com
Posts Replies Media Videos
mattashby.com
This is a slide I use quite a lot when doing POP training for police. Even when an agency or individual wants to follow SARA properly, there are so many institutional pressures on them to skip steps.
How not to do SARA Ignore the problem Try to respond without defining or understanding the problem Define the problem, but don’t take time to understand it Don’t assess whether or not the problem has been solved
November 5, 2025 at 10:58 PM
mattashby.com
I’m afraid at @waterstones.bsky.social Nottingham it’s even worse – it’s even upside down on the poster.
Reserve your copy now! THIS WAY UP Map Men Based on hit Youtube edutainment mini-series Map Men, This Way Up finds Foreman and Cooper-Jones taking a hilarious look at cartography's biggest blunders. 'A gloriously entertaining look at how maps can get things - sometimes very big things - badly wrong' Charlie Higson Published: 23/10/2025
October 31, 2025 at 12:36 PM
mattashby.com
Great afternoon talking with Durham Constabulary leaders about how to make police problem solving really work.

If you’d like me to come and talk about making problem-oriented policing effective and sustainable, I don’t charge for talks to police forces. So please get in touch!
Flowchart illustrating the importance of both short-term and long-term responses. The process begins with a leftward arrow labeled “scanning,” leading to two parallel arrows: “short-term response” above and “analysis” below. A red stop sign appears at the end of the short-term response path, indicating its limit. Analysis continues forward into a third arrow labeled “long-term response.” The upper caption reads, “Have a short-term and a long-term response.” “Priority: rapid implementation” is noted above the short-term response, while “Priority: sustainability” appears below the long-term response, emphasizing the balance between urgency and lasting solutions.
October 13, 2025 at 2:21 PM
mattashby.com
The rise in pedestrian fatalities in recent years seems to be specific to the United States – in other developed countries, pedestrian fatalities of either continued to fall or have flatlined.

usa.streetsblog.org/2020/10/10/e...
Bar chart showing cyclist fatality rates per 100 million km biked in five countries across three periods (2000–2002, 2008–2010, 2016–2018). The U.S. shows the highest fatality rates in all years and is the only country where fatalities rose in the latest period, increasing from 4.5 to 6. All other countries, including the Netherlands, Germany, the U.K., and Denmark, show steady declines or flat rates over time.
October 10, 2025 at 7:05 PM
mattashby.com
Two weeks after seeing last year’s graduates collect their degrees, tomorrow we start it all again with Welcome Week for our new undergraduates.

Going to uni is a daunting experience for many students, so we’ll be doing lots of work this week to help them get up to speed before teaching starts!
The skyline of the City of London has seen from Waterloo Bridge on a sunny afternoon. In the foreground are the waters of the river Thames. On the left is St Paul’s Cathedral and then across the rest of the skyline are the skyscrapers of the financial district.
September 21, 2025 at 5:03 PM
mattashby.com
It’s great to be spending the evening seeing this year’s cohort of BSc Crime and Security Science students graduate at the Royal Festival Hall.  it’s always so nice to be able to congratulate them individually afterwards, and to meet their proud families. 🎓👏
A university graduation ceremony in a large concert hall. The image is looking from the back of the stage. In the foreground a group of academics are sat with their backs to the cameras and wearing colourful academic robes. In the background are rows of new graduates in black formal academic dress. Sat behind them are their friends and relatives.
September 9, 2025 at 8:35 PM
mattashby.com
Today I spoke at #EuroCrim2025 about how to forecast the frequency of crime to support strategic decision making in policing.

Slides: lesscrime.info/slides/esc-2...

Working paper: doi.org/p4vn

Please get in touch if you’d like more information!
Conclusion Reasonably accurate forecasting is possible A combination of forecasts from multiple methods is typically most accurate Some specific methods (e.g. FASSTER) are wildly often inaccurate Remember: Forecasts are only useful as long as there is structural stability
September 6, 2025 at 5:23 PM
mattashby.com
Very nice #EuroCrim2025 presentation by @stijnruiter.bsky.social showing how burglars are much much more likely to offend on streets that they routinely travel on when moving between places they have some attachment to (eg home, friends’ houses, etc.)
A man pointing at a presentation slide showing a map showing how burglars move between places they know and offend on the way
September 4, 2025 at 3:06 PM
mattashby.com
Very engaging presentation at #EuroCrim2025 by @clanfear.bsky.social showing how homicide offenders during the 2021 homicide spike in the 🇺🇸 were very different to the offenders who killed during the homicide peak of the 1990s, with lots of policy implications.
A man pointing at a presentation slide that shows two charts. The left-hand chart shows that in 1995 a large proportion of homicide offenders in the US were adolescents, while in the 2021 homicide spike a much larger proportion of offenders were adults.
September 4, 2025 at 11:54 AM
mattashby.com
A reminder to my colleagues attending #EuroCrim2025: **a question is a single sentence that ends with a question mark**.

When invited to ask a question in a session, that is not an invitation to give your own talk on an unrelated topic.
The 'bad boyfriend' meme with a man looking at a pretty woman passing by on the street, while his girlfriend looks on in disgust. The man is marked 'random guy at a conference', the pretty woman is marked 'making a comment' and the girlfriend is marked 'asking a question'.
September 4, 2025 at 8:59 AM
mattashby.com
Fortunately the Aegean was like a lake – inside the ship you could hardly tell it was moving.
A view from the side of a large passenger ferry, showing a calm sea with rocks in the background. On the left is the end of the ship’s bridge, with a crew member silhouetted against the sunset.
August 30, 2025 at 10:08 PM
mattashby.com
I’m on my way to #EuroCrim2025

(I’m going on the big boat, not the small one)
A large passenger ferry in a harbour. The bottom half of the ferry is painted dark blue and the top is white. Above that is a very large white yellow funnel with a blue star. Blue Star Ferries is painted on the ferry in white. In the harbour is a small fishing boat. The view out of a large window on the side of a passenger ship. Through the window a mountainous island can be seen.
August 30, 2025 at 2:57 PM
mattashby.com
A useful guide for researchers on how to approach research collaborations with agencies such as police. Almost all the top researchers in our field are notably good at applying the tips listed here, especially the need to build long-term relationships.

doi.org/10.1080/2475...
We argue that researchers should: (1) approach collaborations with humility, valuing the insights of practitioners, (2) maintain objectivity by prioritizing evidence over personal (and even professional) biases, (3) align research priorities with agency goals to foster mutual respect and cooperation, (4) recommend incremental, practical changes rather than broad reforms, and (5) aim to foster long-term relationships.
August 28, 2025 at 2:02 PM
mattashby.com
New article finds districts (🇺🇸) with more domestic violence also have more stop and frisk. But the authors spend almost the whole article arguing this is probably because policing causes domestic violence. Only at the very end do they note that:

doi.org/10.1177/1077...
Findings from this exploratory study are suggestive, but must be interpreted with its limitations in mind. First, the cross-sectional, observational design means that causal conclusions cannot be drawn from the results of this analysis. In particular, as noted above, these data do not allow us to decisively distinguish whether SQF contacts exacerbate IPV, IPV exacerbates SQF contacts, or both. Future research to explore the connection between police uses of force and community IPV incidence might aim to exploit external shocks to policing and IPV and data sources that make it possible to establish the temporal ordering of independent and dependent variables.
August 25, 2025 at 2:00 PM
mattashby.com
Never having heard of this substance before, I initially misread this label and was quite surprised! 😳
A jar of clear crystals marked “CRYSTAL MENTHOL”
August 22, 2025 at 12:50 PM
mattashby.com
Use `hotspot_classify()` to identify emerging, persistent, intermittent and former hotspots (similar to the Emerging Hot Spot Analysis tool in ArcGIS Pro).
library(ggspatial) library(sfhotspot) library(ggplot2) memphis_grid <- hotspot_grid(memphis_precincts, quiet = TRUE) robbery_classify <- hotspot_classify(memphis_robberies, grid = memphis_grid) autoplot(robbery_classify) + geom_sf(data = memphis_precincts, colour = "grey20", fill = NA) + theme_void() A grid map of Memphis displaying robbery hotspot categories, with purple cells for emerging hotspots, yellow for intermittent hotspots, gray for former hotspots, red for persistent hotspots, and light gray for areas with no clear pattern. Hotspots are mainly clustered in central and southern parts of the city.
August 5, 2025 at 1:22 PM
mattashby.com
Map changes in crime patterns with `hotspot_change()`, e.g. to see how crime changed in different places after an event such as the start of a crime-prevention initiative.
library(ggspatial) library(sfhotspot) library(ggplot2) memphis_grid <- hotspot_grid(memphis_precincts, quiet = TRUE) robbery_change <- hotspot_change(memphis_robberies, grid = memphis_grid) ggplot() + annotation_map_tile(type = "cartolight", zoomin = 1, progress = "none") + geom_sf(aes(fill = change), data = robbery_change, alpha = 0.75, colour = NA) + geom_sf(data = memphis_precincts, colour = "grey20", fill = NA) + scale_fill_gradient2() + labs(fill = "change in\nrobberies") + theme_void() A grid map of Memphis showing changes in robbery counts, where blue squares indicate increases and red squares indicate decreases. Most cells are light-colored, with scattered pockets of change across the city.
August 5, 2025 at 1:22 PM
mattashby.com
Use `hotspot_gistar()` to find hotspots – places with more points (crimes) than expected by chance. Identifying hotspots is useful because it can help avoid wasting resources responding to random variations in crime.
library(ggspatial) library(sfhotspot) library(ggplot2) memphis_grid <- hotspot_grid(memphis_precincts, quiet = TRUE) robbery_gistar <- hotspot_gistar( st_transform_auto(memphis_robberies), grid = st_transform_auto(memphis_grid), bandwidth_adjust = 0.4 ) |> dplyr::filter(gistar > 0, pvalue < 0.05) ggplot() + annotation_map_tile(type = "cartolight", zoomin = 1, progress = "none") + geom_sf(aes(fill = kde), data = robbery_gistar, alpha = 0.75, colour = NA) + geom_sf(data = memphis_precincts, colour = "grey20", fill = NA) + scale_fill_distiller(direction = 1) + labs(fill = "density of\nrobberies") + theme_void() A grid heatmap of Memphis showing robbery density per cell, shaded from light to dark blue. Densest areas are west of downtown and in the southeastern neighborhoods.
August 5, 2025 at 1:22 PM
mattashby.com
Identify crime density with `hotspot_kde()`. Density maps are often easier to interpret than count maps while showing more detail than choropleth maps. This example shows how you can take more control with sfhotspot by over-riding the default values.
library(ggspatial) library(sfhotspot) library(ggplot2) memphis_grid <- hotspot_grid(memphis_precincts, quiet = TRUE) robbery_kde <- hotspot_kde( memphis_robberies, grid = memphis_grid, bandwidth_adjust = 0.4 ) ggplot() + annotation_map_tile(type = "cartolight", zoomin = 1, progress = "none") + autolayer(robbery_kde, alpha = 0.75) + geom_sf(data = memphis_precincts, colour = "grey20", fill = NA) + scale_fill_distiller(direction = 1) + labs(fill = "density of\nrobberies") + theme_void() A smoothed heatmap of robbery density in Memphis, shaded from light to dark blue, with higher-density areas concentrated downtown, in south-central neighborhoods, and to the east of the city center.
August 5, 2025 at 1:22 PM
mattashby.com
Count points with `hotspot_count()`. By default this creates a grid of cells and counts points (e.g. crimes) in each cell, but it can also be used to count points in polygons to make a targeting table or a choropleth/thematic map.
library(ggspatial) library(sfhotspot) library(ggplot2) memphis_grid <- hotspot_grid(memphis_precincts, quiet = TRUE) robbery_count <- hotspot_count(memphis_robberies, grid = memphis_grid) ggplot() + annotation_map_tile(type = "cartolight", zoomin = 1, progress = "none") + autolayer(robbery_count, alpha = 0.75) + geom_sf(data = memphis_precincts, colour = "grey20", fill = NA) + scale_fill_distiller(direction = 1) + labs(fill = "count of\npoints in\neach cell") + theme_void() A grid-based map of Memphis showing counts of robberies in each cell, shaded from light to dark blue. Most cells have low counts, with only a few small clusters reaching 20–25 points.
August 5, 2025 at 1:22 PM
mattashby.com
The goal of this package is to make hotspot analysis as simple as possible for novice users, by choosing reasonable default values and doing as much as possible for users behind the scenes. This can make quite a big difference even for simple maps:
# CREATING A BASIC KERNEL DENSITY MAP WITH SPATIALKDE library(SpatialKDE) library(ggplot2) memphis_kde <- sfhotspot::memphis_robberies |> # Need to transform data to use a projected CRS sf::st_transform("EPSG:32616") |> # Need to specify bandwidth and cell size kde(band_width = 5600, cell_size = 500) # Need to set up plot manually ggplot(memphis_kde) + aes(fill = kde_value) + geom_sf(colour = NA) + scale_fill_distiller(direction = 1) + labs(fill = "estimated\ndensity\nof points") + theme_minimal() # CREATING A BASIC KERNEL DENSITY MAP WITH SFHOTSPOT library(sfhotspot) library(ggplot2) memphis_robberies |> # Bandwidth/cell size set automatically based on data (can be set manually) hotspot_kde() |> # Basic plot produced automatically autoplot(memphis_kde)
August 5, 2025 at 1:22 PM
mattashby.com
Announcing sfhotspot, an #RStats package to make it as easy as possible to identify and map concentrations of crime (or concentrations of any other points in space).

Introduction: lesscrime.info/post/sfhotsp...
Package website: pkgs.lesscrime.info/sfhotspot/

🧵
A blue hexagon on a white background. Inside the hexagon is a grid of much smaller hexagons. Most of these are dark blue, but some of the small hexagons are lighter blue and spell out the letters SF. On top of the grid is the text sfhotspot.
August 5, 2025 at 1:22 PM
mattashby.com
I’ve come back from holiday to this envelope, which suggests that either I’ve been sacked and UCL does exit interviews by post, or that the alumni database doesn’t distinguish alumni who then became staff.

I guess I’ll find out which when I go back to work tomorrow?
An envelope bearing the UCL logo and the text “What stands out to you from your time at UCL?”
July 27, 2025 at 8:13 PM
mattashby.com
Some good recommendations resulting from a deliberate exercise with members of the public on how police should/shouldn’t use AI tools. It’s crucial police get this right from the start to avoid harming public confidence.

Short summary of the report: sciencewise.org.uk/wp-content/u...
A screenshot of a page from a report for the Home Office on police use of AI. Participants took part and in a deliberate exercise where they produced recommendations for police forces on how to use AI with public support. See page 8 of the linked PDF.
July 23, 2025 at 2:00 PM
mattashby.com
TfL (as London Transport) used to have its own factory for preparing food (including a butchery and a bakery) as well as its own cookery school.

www.ltmuseum.co.uk/collections/...
CROYDON FOOD PRODUCTION CENTRE Introduction London Transport's Catering Department operates 165 canteens in the Greater London area, providing meals and refreshments for 74,000 staff. It is the largest single catering unit in the country. To ensure complete control over quality and freshness, as well as the economy of centralized bulk-buying, London Transport opened in 1950, at Croydon, Surrey, its own Food Production Centre, which today makes (or receives and distributes in bulk everything which is sold at the indi- vidual canteens. Buying is done daily, by highly experienced staff, at the actual markets, at competitive prices. To distribute its products, the Food Production Centre maintains a highly organized road dispatch service and supplies of perishable foods are delivered to canteens on six days a week. Bakery The Bakery Manager has an assistant, two supervisors and a staff of 48. The most up-to-date equipment enables the Bakery to produce a total of some 120 different bakery lines, totalling some 100,000 cakes and a further 100,000 other items a week. Including meat items, such as pies and sausage rolls, the total is in the region of a quarter-million a week. Over the year there are other special items—100,000 Hot Cross Buns, 45,000 Christmas Puddings, 15,000 Christmas Cakes, and 1,000 or so Wedding and other special made-to-order cakes. B/W print; trainee cooks making pastry at the Baker Street training centre, by Topical Press, 1949
July 13, 2025 at 1:57 PM