☕ Buy me a coffee → Any donations go towards API costs and time spent on the website. Thank you!

Research Tools

Practical Summaries

Every article as a briefing for senior leaders — the problem, what the evidence says, and what you gain if you act. Each summary is free to read inside its article; sign in free to browse the whole library, filter by section and topic, and save summaries to your account.

25 summaries · updated with every article

Members · free

Sign in to unlock the summaries library

Sign-in takes under a minute: create a password or use Google, and you’re in. One free sign-up unlocks everything members get:

Save articles and practical summaries to your library
Download cited research briefings
Use every research tool on the site — free

Evidence & Practice · 15

Evidence & Practice

Self-Selection Policing: Catching Serious Criminals — and Bad Cops — Through the Small Offences They Can't Help Committing

01The problem

Serious offenders do not confine themselves to serious crime, yet policing treats minor offences as an end point rather than a lead — and applies the same blind spot to small rule-breaches by its own officers.

02What the evidence says

The founding Home Office study found about one in four vehicles parked illegally in disabled bays was linked to someone with a live warrant, a disqualification or active criminal involvement.
Chenery, Henshaw and Pease, Home Office (1999)
Roach's study of disqualified drivers caught behind the wheel found a large share had convictions for other, often serious, offences.
Roach, Policing: A Journal of Policy and Practice
The Casey Review found the Met dealing with behaviour that would mean instant dismissal in most organisations through light-touch "reflective practices", failing to spot repeated patterns in individual officers.
Casey Review / House of Commons Library (2023)

03If we act

If we treat minor offences committed in plain sight as the trigger for a closer look at who committed them,active serious criminals are caught at little extra cost to the force, because the stop is already happening.
If we pick trigger offences on evidence rather than gut feeling,checks stay aimed at offences that genuinely flag serious offending, protecting the public from the approach sliding into plain prejudice.
If we apply the same lens to patterns of small rule-breaches by officers,the force gets a chance to spot escalating misconduct before it does serious harm to the public.

self-selection-policing evidence-based-policing evidence-basedcrime-preventionpolicingpolicecriminologyprofessional-standards

Read the full article →

Evidence & Practice

Procedural Justice: A Guide for Police Leaders — and How to Explain What It Actually Is

01The problem

Public confidence in the police in England and Wales has fallen from 63% to 49% in under a decade, and neither crime rates nor officer numbers explain it.

02What the evidence says

In a multicity randomised trial at crime hot spots, procedurally trained officers made over 60% fewer arrests, crime in their patches fell 14%, and residents reported markedly less harassment.
Wood, Weisburd and Sherman, PNAS (2022)
In Queensland, a scripted procedurally just breath test led 17.4% of drivers to say the stop had improved their view of the police, against 12.5% in the control group.
Mazerolle et al., Criminology (2013)
Officers' sense that their authority is rightfully held is predicted most strongly by whether they see their own senior management as procedurally fair.
Bradford and Quinton (2014)

03If we act

If we train officers in voice, neutrality, respect and trustworthy motives as a sustained, scenario-based discipline,arrests fall, crime falls and the public reports less harassment, without the force giving up a single legal power.
If we model the same fairness toward our own supervisors and officers,officers extend it to the public, because internal fairness and street-level fairness are the same causal chain.
If we apply it hardest to stop and search, use of force and custody,the encounters most likely to damage trust do less harm to communities, and the crime-control case for the powers is protected.

procedural justice police legitimacy public confidence stop and search use of force police leadership evidence-based policingtrainingpolicingpolice

Read the full article →

Evidence & Practice

Why Medium Risk Is the Safe Option — and How Do We Learn to Accurately Grade Risk With Confidence?

01The problem

On domestic abuse calls officers default to grading risk medium because it feels safe, so the scale loses its bottom and the genuine medium-risk victim disappears into the noise.

02What the evidence says

The largest European study of police domestic abuse risk assessment found DASH only weakly predictive: close to nine in ten victims who suffered repeat violence had been graded standard or medium, not high.
Turner, Medina-Ariza and Brown (2019)
DARA's pilot reported a 38% increase in officers reaching the same risk decision as a domestic abuse expert — consistency is what the tool exists to buy.
College of Policing (2022)
In a survey of roughly 1,700 people, "likely" meant anything from 55% to 90% and "real possibility" from 20% to 80% — the same probability word carries wildly different odds.
Mauboussin and Mauboussin (2018)

03If we act

If we grade against the written definitions — high means serious harm could happen at any time, medium only if circumstances change,each grade carries real information, and the genuine medium-risk victim gets safeguarding instead of vanishing into the noise.
If we use standard honestly when the evidence does not indicate a likelihood of serious harm,medium regains its weight as a clear step above "nothing further", and downstream safeguarding effort goes to the victims who need it.
If we run each grading decision through the National Decision Model and record the reason,supervisors can test the reasoning rather than the box, and officers can defend "why medium and not high?" to a court or review years later.

domestic abuse safeguardingDARArisk assessmentviolence against womenNational Decision ModelCollege of Policingdecision-making

Read the full article →

Evidence & Practice

ASB Civil powers: A Frontline Guide to Warnings, Notices and Orders

01The problem

The Sentencing Act 2026 means more prolific and dangerous offenders are bailed while their cases run, yet the civil orders that could protect victims in the meantime are among the most powerful and least used tools frontline officers have.

02What the evidence says

The Sentencing Act 2026 requires courts to weigh whether there is "no real prospect of immediate custody" — the intended and likely effect is fewer remands and more bail.
Independent Sentencing Review (Gauke)
Forces are using Criminal Behaviour Orders to protect retail staff — a Clacton shoplifter was barred from named supermarket chains for five years, making the next breach a straightforward arrest.
Essex Police (2025)
The courts require order terms to be reasonable, proportionate, clear, enforceable and "policeable" — vague or unworkable terms are where applications most often come apart.
Boness [2005]; R v Maguire [2019]

03If we act

If we climb the ladder rung by rung — informal action, warning, notice, order — recording the evidence and reasoning at each step,orders survive challenge and victims get real, lawful protection that does not depend on a remand decision going the force's way.
If we apply for a CBO at the point of conviction for prolific offenders, with positive requirements attached,retail staff and communities get protection that outlasts the sentence, and the order addresses why the person keeps offending.
If we wrap a housing and safeguarding plan around any closure order on a cuckooed address,the dealing is broken and the vulnerable occupant is protected as the victim they usually are.

anti-social behaviourcivil orderscriminal behaviour ordercommunity protection noticeclosure ordersafeguardingretail crimecounty linesSentencing Act

Read the full article →

Evidence & Practice

A Police Leader's Guide to AI

01The problem

Police leaders are asked to approve, govern and publicly defend AI tools they were never trained to understand — and the accountability for what those tools do cannot be delegated to the supplier.

02What the evidence says

AI predicts rather than understands, so it produces confident, fluent, wrong answers — and a model trained on historic policing data reproduces historic disproportionality with an air of mathematical objectivity.
Samsung staff pasted confidential material into a public chatbot three times in 2023 and it left the company for good — what goes into a free AI tool may be stored abroad and used to train future models.
Samsung leaks (2023)
Using AI lawfully with personal data requires a Data Protection Impact Assessment and negotiated data-residency terms; the ICO's guidance is the authoritative UK reference.
Information Commissioner's Office

03If we act

If we use AI to narrow, draft and prioritise but never to decide, keeping a trained human able to overturn the output,the force gets the speed of AI on high-volume work without indefensible decisions reaching the public.
If we complete a DPIA and negotiate data-residency, retention and no-training terms before any tool is deployed,the force's data stays under contract rather than leaking for good, and leaders can defend the tool at a scrutiny panel.
If we brief every officer to treat a public AI tool like a postcard, not a sealed envelope,sensitive material such as a victim's account never leaves the building unrecoverably.

artificial intelligenceaimachine learningtechnologydatasurveillance leadershippolicingpolicepolicy

Read the full article →

Evidence & Practice

PoliceAI: What is it, why do we need it, and who runs it?

01The problem

AI tools are reaching the 43 forces of England and Wales faster than the rules governing them, and most leaders will meet them through a press release that answers none of the operational questions.

02What the evidence says

PoliceAI is not software but a national centre, hosted by the College of Policing and backed by £75m over three years, that tests, commissions and governs AI tools for all 43 forces.
GOV.UK launch announcement (2026)
Avon and Somerset's trial of the Söze platform across 27 cold cases did work estimated at 81 analyst-years in around 30 hours.
Avon and Somerset Police / Akkodis
West Midlands Police's use of Copilot generated false material that fed real operational decisions, yet around 21 forces were still using the tool afterwards.
Press reporting (2026)

03If we act

If we demand a tool's measured error rate and independent assessment before it goes live,the force can show a court exactly how the tool was tested and who is accountable when it is wrong.
If we audit the general-purpose AI staff already use and put training and policy around it,the force catches the quiet tools before they write false material that reaches operational decisions or a court.
If we start deliberate adoption with redaction,officers get large, immediate time savings on checkable output, where an error is a privacy slip rather than a wrongful lead.

policeaiartificial intelligence predictive-policing surveillance facial recognitiondatatechnologyleadershippolicingpolicepolicy

Read the full article →

Evidence & Practice

The Neighbourhood Policing Guarantee: Confidence or Crime?

01The problem

The Neighbourhood Policing Guarantee folds two different jobs under one banner — a named ward officer builds confidence, hotspot patrolling cuts crime — and the design risks judging each against the wrong outcome.

02What the evidence says

Around half of all crime concentrates in a few per cent of street segments and addresses, so concentrated patrol in those places is where the crime effect lives.
Weisburd, law of crime concentration
Patrol visits of around 15 minutes buy the most residual deterrence, and diluting the dose into ambient everyday presence buys no crime effect at all.
Koper (1995)
Neighbourhood policing lifts public confidence when targeted visible presence, community engagement and problem-solving run together.
College of Policing (2025)

03If we act

If we run the named-officer scheme and hotspot patrolling as separate programmes, each with its own outcome measures,the public gets both a confidence gain and a crime reduction, and neither programme is written off against the wrong yardstick.
If we keep dosage discipline — short, repeated, roughly 15-minute visits in the few street segments the data flags,crime and disorder fall measurably in the places that carry most of the harm.
If we judge the confidence programme on confidence measures, not crime figures it was never dosed to move,the force avoids banking a false failure and chiefs can defend their resourcing choices.

neighbourhood-policing hotspot-policing public-confidence evidence-based-policingwhite-paperwhat-workspolicypolice

Read the full article →

Evidence & Practice

Nitazenes - What This New drug Means For British Policing

01The problem

Nitazenes — synthetic opioids far more potent than the heroin they are displacing — are already killing at scale in the UK, and most forces are under-resourced even to detect the problem.

02What the evidence says

The National Crime Agency links nitazenes to roughly 1,000 UK deaths since June 2023, while heroin seizures have fallen 54% to their lowest level since 1989 — the conditions for a synthetic transition are in place.
National Crime Agency (2026)
Naloxone works but often needs multiple or prolonged doses, and apparent recovery can give way to re-narcotisation as the naloxone wears off before the opioid does.
Clinical Toxicology (2025)
Recorded deaths miss up to a third of nitazene fatalities because the drugs degrade in post-mortem blood samples, so a response funded on official figures is short by design.
King's College London / Copeland et al. (2026)

03If we act

If we plug the force into the national early-warning systems and contribute overdose, naloxone and drug-checking data,the force spots a cluster before the body count reveals it, and the public gets a response sized to the real threat.
If we recalibrate naloxone provision and custody monitoring for multiple doses and prolonged observation,victims and detainees are protected against the re-narcotisation and late deterioration that the single-dose heroin model misses.
If we fix officer-safety messaging with brief, evidence-based training on exposure risk,fear of the powder no longer delays rescue breathing or naloxone — one short intervention cut the false "touching is deadly" belief from 80% to 39%.

drugssynthetic-opioidsnitazenes harm-reductionpublic-healthcounty-linespolicingpolicepolicy

Read the full article →

Evidence & Practice

“Do you understand?” - Does this question serve a purpose?

01The problem

"Do you understand?" — asked after the caution, at booking-in and before a plea — works as a ritual rather than a test, inviting a yes that keeps the system moving while failing the people who most need clarity.

02What the evidence says

The Equal Treatment Bench Book warns that asking "do you understand?" is unreliable because people say yes to save face, and recommends asking the person to feed back what they have understood in their own words.
Equal Treatment Bench Book
PACE Code C already builds the safeguard in: the appropriate adult exists to facilitate communication and protect people who do not understand the significance of what is said to them.
PACE Code C
The Criminal Practice Directions require courts to take every reasonable step to enable a defendant to comprehend and participate effectively.
Criminal Practice Directions (2023)

03If we act

If we replace the checkbox question with teach-back, asking people to explain the caution or their conditions in their own words,officers hear where the misunderstanding sits and fix it, and interviews, custody decisions and conditions rest on real comprehension.
If we make "no" safe to say and treat clarity as the officer's responsibility, with an explicit escalation ladder when comprehension fails,the force gets fewer failed interviews, misunderstandings and complaints, and vulnerable people get a fairer process.

Read the full article →

Evidence & Practice

The Legal Aid Paradox - when "I don't need a solicitor, I've done nothing wrong" undermines procedural fairness

01The problem

Around 40% of adult suspects decline free legal advice in custody, and those who need it most — first-timers and people who believe they are innocent — decline most often, undoing a core PACE 1984 protection.

02What the evidence says

Adult take-up of custodial legal advice varied from 52.8% to 64.1% across forces in 2024, so a suspect's protection depends partly on which custody suite books him in.
Kemp (2024)
Only 6% of Scottish suspects tested on the Letter of Rights understood what the word "rights" meant, and the English and Welsh document has the same problem.
JUSTICE Scotland (2018)
Scotland has removed the right to waive altogether for children under 16 and vulnerable adults — the interview cannot proceed without a solicitor.
Criminal Justice (Scotland) Act 2016

03If we act

If we require a brief call with a duty solicitor before any waiver can be recorded,suspects hear what advice involves from a lawyer rather than a custody sergeant reading Code C, and fewer vulnerable people waive blind.
If we rewrite custody scripts to correct known misconceptions — innocent people benefit from advice, asking does not imply guilt, a caution creates a record,first-time and innocent suspects decide on accurate information, and the force's interviews and cautions stand up to later challenge.
If we collect systematic data on outcomes for advised versus unadvised suspects,the true cost of non-representation becomes visible to forces and policymakers, instead of remaining hidden.

crimeevidence-based-policing law leadership mental-health neurodiversitynewspolicepolicingpolitics

Read the full article →

Evidence & Practice

The Hidden Crisis: How Police Burnout Threatens Officer Safety and Public Trust

01The problem

Chronic burnout degrades the impulse control, judgement and threat assessment officers rely on, raising the risk of excessive force and eroding public trust — yet forces still treat it as ordinary tiredness.

02What the evidence says

Burnout correlates with more positive attitudes toward force and more frequent violence on duty, established with administrative and observational data on 422 Dutch officers rather than self-report.
Kop and Euwema (2001)
After 17 to 19 hours without sleep, performance matches a blood alcohol level of about 0.05%, and after 24 hours about 0.10% — a sleep-deprived officer can be functionally as impaired as a drunk one.
Williamson and Feyer (2000)
Early warning systems tracking complaints, force incidents and attendance cut citizen complaints by 62–67% in Minneapolis and New Orleans.
Department of Justice

03If we act

If we build an early-warning system from data the force already collects — complaints, use-of-force incidents, sick leave,struggling officers are reached before a crisis and citizen complaints fall, as they did in Minneapolis and New Orleans.
If we treat fatigue as impairment and review shift design,officers stop making critical decisions at drunk-equivalent levels of impairment, and schedule change alone has produced sustained gains in work-life balance and job satisfaction.
If we make brief trauma-focused CBT and EMDR genuinely accessible,officers get treatment that substantially reduces PTSD, depression and anxiety in six to eight sessions — interest already outstrips uptake.

crimeevidence-based-policing mental-health neurodiversitypolicepolicing

Read the full article →

Evidence & Practice

When Police Stop Investigating Crime: The Fuel Theft Crisis hurting British Policing

01The problem

Around 99% of fuel thefts end without prosecution even when retailers hand police clear CCTV, number plates and timestamps — an open abandonment of a solvable crime that is breaking retailer cooperation and eroding policing by consent.

02What the evidence says

Between 2020 and 2024 Sussex closed 94% of 2,892 fuel-theft cases with no suspect identified, Humberside 83% and South Yorkshire 91%, despite comprehensive CCTV, ANPR and payment evidence.
Forecourt Trader FOI investigation
Losses have reached £100m across 1.5 million incidents, and the industry now recovers £35m a year through private civil channels because criminal enforcement has withdrawn.
British Oil Security Syndicate
Non-investigation is not a staffing problem — officer numbers hit an all-time high of 147,745 in March 2024, making this a prioritisation decision, not a shortage.
Home Office police workforce statistics

03If we act

If we audit how making-off-without-payment reports are closed, checking that supplied CCTV is reviewed before a no-suspect message is sent,the force stops writing off solvable crimes, and victims who hand over evidence see it actually examined.
If we treat forecourts as islands of enquiry and retrieve footage within the 31-day retention window,investigators keep an evidence source that has already helped convict murderers, and serious cases get timeline evidence before it is overwritten.
If we state prioritisation decisions openly instead of dressing them up as evidential failure,the force protects its legitimacy with retailers and the low-trust communities who staff overnight shifts, instead of eroding policing by consent.

crimeevidence-based-policinghistorylawnewspolicepolicingpolitics

Read the full article →

Evidence & Practice

The Officers We’re Losing: Why British Policing Must Rethink Cognitive Diversity

01The problem

British policing invests heavily in recruitment and training, then loses officers whose cognitive profiles — ADHD, autism, dyslexia — do not fit institutional assumptions about what "good" looks like, discarding capabilities the job increasingly needs.

02What the evidence says

A screen of 302 wildland firefighters found 19.5% scored in the range suggestive of ADHD — roughly four times the general adult rate — consistent with high-stakes operational work suiting these attention systems.
US Forest Service
In a 2023 survey of UK police employees with ADHD, 78% said the trait brought benefits to their policing work, citing thriving in fast-paced environments.
UK police workforce survey (2023)
Organisations that over-invest in efficiency and process compliance at the expense of exploration become brittle — excellent until conditions change, then blindsided.
March (1991)

03If we act

If we audit performance and promotion criteria for which standards predict capability and which merely reward conformity,the force keeps officers with pattern recognition, crisis performance and deep community knowledge it would otherwise manage out.
If we develop supervisors to see that brilliant crisis performance and chaotic administration can sit in the same officer,neurodivergent officers are deployed as assets rather than quietly resigning, and the force keeps capability it paid to train.
If we judge officers on the strengths they bring, not only the boxes they tick,the force holds the exploratory thinkers it will need when standard approaches stop working in a more volatile operating environment.

adhdaibusinesscrimeevidence-based evidence-based-policing leadership mental-health neurodiversitypolicepolicingtechnology

Read the full article →

Evidence & Practice

Seeing Is No Longer Believing: How Synthetic Evidence Could Convict an Innocent Man

01The problem

Convincing video with synchronised audio can now be faked cheaply from a single still, while genuine footage can be waved away as "probably AI" — and neither investigators nor the courts can yet reliably tell the difference.

02What the evidence says

A California court threw out a case in 2025 after finding the claimants had submitted deepfake video, admitting it lacked the time, funding and expertise to authenticate what it was handed.
Mendones v. Cushman & Wakefield (2025)
Europol projected 8 million deepfakes in circulation in 2025, up from around 500,000 in 2023, and the Arup fraud took $25.6m in a single day on a deepfaked video call.
Europol Innovation Lab
England and Wales still presumes a computer was working correctly — the presumption behind the Post Office Horizon convictions — and reform only reached a call for evidence in January 2025.
Ministry of Justice call for evidence (2025)

03If we act

If we seize the native recording from the original device, hashed at seizure, before the roughly 28-day retention loop overwrites it,the court gets an original it can test, and a manipulated clip can be caught before it convicts an innocent person.
If we test the independence of every corroborating clip rather than counting angles that agree,juries are protected from manufactured certainty, because a second camera angle is now one more prompt away.
If we build access to forensic experts and press for reform of the computer-evidence presumption,disputed footage gets tested rather than believed or waved away, protecting the innocent from fakes and society from the liar's dividend.

artificial intelligencedeepfakesdigital evidencecriminal justiceCCTVdigital forensicsjury

Read the full article →

Evidence & Practice

Boosting Public Confidence Through Neighbourhood Policing

01The problem

Neighbourhood policing demonstrably builds public confidence, but delivery across the 43 forces is uneven and neighbourhood capacity is constantly abstracted onto response work.

02What the evidence says

Delivered to a high standard, neighbourhood policing produces quick, large and sustained gains in public confidence.
College of Policing systematic reviews
Crime concentrates — 50% of all crime in under 4% of addresses — and hot spots policing shows consistent benefits across more than 80 rigorous evaluations, with no evidence of displacement.
Sherman; Braga and colleagues
A Campbell Systematic Review of 34 studies found problem-oriented policing cuts targeted problems, with crime reduction diffusing to nearby areas rather than displacing.
Hinkle et al. (2020)

03If we act

If we protect neighbourhood capacity from abstraction onto response work,prevention actually happens, engagement promises are kept, and public confidence rises instead of going with the broken promises.
If we concentrate problem-solving on the small number of places producing most harm, funded with the analysis SARA needs,crime falls in those places and the reduction spreads to nearby areas rather than displacing.
If we measure harm reduction rather than activity,resources go where they do communities the most good, because what gets measured gets done.

crimeevidence-based evidence-based-policing mental-healthnewspolicepolicingpolitics

Read the full article →

Leadership & Culture · 3

Leadership & Culture

Martyn's Law: An overview for Police, Business owners and Local Authorities

01The problem

Before the Terrorism (Protection of Premises) Act 2025 no law required any UK venue to prepare for a terrorist attack — and its duties commence around spring 2027, giving venues, forces and local leaders a year of runway.

02What the evidence says

The Act applies where 200 or more people may reasonably be present, split into a standard tier (200–799) and an enhanced tier (800+).
ProtectUK
The standard tier requires no physical security at all — the duty is a plan for four scenarios (evacuate, invacuate, lock down, communicate) and staff who know it.
Home Office factsheet (2025)
Enforcement sits with the Security Industry Authority, not the police — no officer will inspect venues, and the SIA's stated posture is advice-first.
GOV.UK / SIA

03If we act

If we map the enhanced-tier venues and events in the force area and resource CTSAs for a year of heavy demand,businesses get accurate advice before commencement, and the force is ready when every sizeable venue wants an hour of a CTSA's time.
If we brief neighbourhood teams on what the law requires — and what it doesn't,officers save businesses money, the service earns goodwill, and myths about scanners and police inspections are killed early.
If we agree an interface with the SIA now,inspection findings that look like intelligence reach the force, while regulatory enforcement stays out of local policing relationships.

martyns-law counter-terrorismpublic-policypolicingpolicelawcrime-preventionleadershipbusiness

Read the full article →

Leadership & Culture

The Murder of Henry Nowak: What the Sentencing Remarks Reveal

01The problem

Much of what has been said about the murder of Henry Nowak — including about the officers who reached him first — differs sharply from what the court found in the judge's own sentencing remarks.

02What the evidence says

The pathologist found the chest wound unsurvivable — Henry would not have survived however quickly he received first aid, CPR or expert medical treatment.
Sentencing remarks, R v Digwa (2026)
The judge found the officers honestly believed there were grounds to arrest Henry, and listed Digwa's convincing but wholly false narrative to police as an aggravating feature of the murder.
Sentencing remarks, R v Digwa (2026)
The remarks are not the final word — the IOPC is examining whether the officers breached professional standards on use of force, duty of care and the respect and courtesy owed to Henry.
IOPC (2026)

03If we act

If we work from primary sources such as sentencing remarks when reviewing a contested incident,the force reports the court's measured findings rather than the headlines, and officers and the public get an accurate account.
If we ask in training and debriefs how an injured person's own words should be weighed at a chaotic scene,responders are better armed against a determined, plausible liar turning their instincts against a victim.
If we let the IOPC investigation and the commissioned review run their course,conclusions about officer conduct rest on evidence rather than a simpler, angrier story — fairer to the officers and to Henry's family.

crimecourtsevidence-based-policing lawnewspolicepolicingaccountability

Read the full article →

Leadership & Culture

What Palantir's Advance Means for British Policing — and Should We Trust Them?

01The problem

Palantir has moved from the edge of UK public-sector IT into a structural position inside defence, health and policing — including the national firearms-licensing database — faster than the transparency and procurement rules governing it.

02What the evidence says

Cumulative UK public-sector commitments to Palantir now exceed £900m, and a £9m contract gives the firm the national firearms-licensing database covering all 43 forces.
The Nerve and openDemocracy reporting
Bedfordshire's own impact assessment for the multi-force Nectar pilot disclosed processing of political opinions, sexual orientation, race and health data on witnesses, victims and children as well as suspects.
Liberty Investigates (2025)
The Mayor of London blocked a proposed £50m Met contract in May 2026, citing a clear and serious breach of procurement rules and the risk of being locked into Palantir's technology.
Mayor's Office for Policing and Crime (2026)

03If we act

If we publish contract values, data categories ingested and impact assessments voluntarily,the force gets ahead of FOI and ICO action, and the public sees what has been bought before the secrecy itself becomes the story.
If we require written exit and data-portability plans before any contract extension,the force avoids the lock-in the Mayor of London cited when blocking the £50m Met deal, and keeps the freedom to walk away.
If we commission independent legal advice on US CLOUD Act exposure for sensitive datasets,the force holds its own documented assessment rather than the supplier's account of a risk Parliament has called an unacceptable point of weakness.

palantir surveillancedataprivacy civil-liberties predictive-policingprocurementpolicingpolicepolicy

Read the full article →

Data Stories · 2

More or Less, and the Art of Not Being Fooled by Numbers

01The problem

Confident statistics — crime figures included — arrive in the news and harden into fact within a day, before anyone asks where the number came from or what it should be divided by.

02What the evidence says

A total means nothing on its own: the first question is what to divide by — the population, the budget, the number of years — before deciding whether to be impressed.
More or Less, BBC Radio 4
Half of all arguments about crime figures are really arguments about definitions — recorded crime and the crime survey are two different instruments measuring two different things.
More or Less
Harford's first rule is to watch your own feelings — the numbers that slip past unchecked are the ones you wanted to believe.
Tim Harford, How to Make the World Add Up (2020)

03If we act

If we ask for the denominator before reacting to any headline figure,the force judges whether a number is actually big for its population, budget or period before acting on it.
If we ask who counted, how and under what definition — starting with recorded crime versus the crime survey,most contested crime figures dissolve into clarity, and the force argues from what was actually measured.
If we treat the statistic that flatters our existing opinion with the most suspicion,checking starts at exactly the moment it usually stops, and decisions rest on tested numbers rather than wishful ones.

public-confidence evidence-based-policingstatisticsnumbersmediadata

Read the full article →

Yes, Burnham Is Eyeing North Sea Oil, But Not the Way You (or Trump) Think

01The problem

Donald Trump says opening the North Sea will make Britain rich, and the new government looks set to approve more drilling — but the claims about bills, reserves and national wealth do not survive contact with the data.

02What the evidence says

About 93% of the cumulative North Sea total has already been extracted, and production outpaced new discoveries by more than four to one in 2024.
North Sea Transition Authority / ECIU
Because oil and gas sell at global prices, maximum extraction would cut household bills by only £16 to £82 a year, against around £330 from decoupling electricity prices from gas.
Oxford Smith School
The oil and gas workforce is projected to fall from 115,000 in 2024 to between 57,000 and 71,000 by the early 2030s regardless of policy, because the decline is structural.
Robert Gordon University

03If we act

If we judge the "make Britain rich" claim by whether more drilling changes the global price the UK actually pays, and who owns the oil once it is pumped,the debate moves onto the measures that would cut bills, like decoupling electricity prices from gas, worth around £330 a year against £16 to £82 from maximum drilling.
If we treat the Jackdaw and Rosebank decisions as what they are — consent decisions on existing licences, bound by climate law and open to legal challenge — rather than a switch a Prime Minister can simply flip,expectations match what any government can actually deliver, and the real questions about jobs and transition get a fair hearing.
If we pursue energy security by needing less oil and gas, through electrified heating and transport backed by clean generation and storage,the country gains real insulation from volatile world markets, which chasing the last barrels of a depleted basin cannot provide.

Read the full article →

Technology & Innovation · 5

Technology & Innovation

Can an Electric Car Catch a Criminal? The Physics, the Payload and the Pursuit

01The problem

Policing has committed to electrifying a fleet that is 80% diesel, but the operational edge cases — pursuits, armed response, rural 24/7 response — are governed by physics and payload limits that current electric vehicles do not meet.

02What the evidence says

A 2021 UK emergency-services trial found a Tesla Model 3 managed over 200 miles of blue-light advanced driving on one charge, and that average UK blue-light runs last only 7–15 minutes.
Tesla UK trial findings (2021)
Aerodynamic drag power rises with the cube of speed: independent testing found a BMW i4 M50 used 40% more energy at 120 km/h than at 90, and a Porsche Taycan held at ~160 mph drained its 93 kWh pack in roughly 80 miles.
InsideEVs / Bjørn Nyland range tests (2022)
The BMW X5 armed response vehicle offers ~730 kg of payload; its closest electric equivalent offers ~540 kg, because the battery pack itself weighs about 700 kg — before crew, weapons or kit.
Manufacturer specifications

03If we act

If we electrify by role — beat, neighbourhood and most response cars first, with pursuit-capable and ARV fleets last,the force cuts fuel and maintenance cost on the miles that suit EVs today without gambling pursuit capability on physics that has not changed.
If we fund depot rapid charging before the vehicles arrive, not after,response cars keep double-crewed 24/7 availability, and officers stop losing patrol hours to public charge queues.
If we publish clear fleet criteria for when hybrid or diesel is retained,the public gets a credible net-zero plan rather than a target that quietly fails, and officers get vehicles specified for the job.

technologyfleetroads policingarmed policingnet zerodata

Read the full article →

Technology & Innovation

How Worthing Became 'The Shoplifting Capital of the UK' — And Why That Was Never the Real Story

01The problem

Commercial rankings branded Worthing the shoplifting capital of the UK when the rise in recorded theft was largely the force's own reporting scheme working as intended — and Sussex Police took fourteen months to say so.

02What the evidence says

One Touch Reporting made shop theft easier to report, so recorded crime rose because more theft was being recorded, not because more was happening.
Sussex Police (2024)
Worthing's recorded shoplifting fell year on year while it was being called the country's worst town — 1,856 incidents by September 2025, down from 2,596.
Sussex PCC figures (2025)
Four commercial studies used four different denominators and named four different worst towns from broadly the same data.
Accu Components, LEI Register, Avigilon, CrimeRate.co.uk

03If we act

If we publish plain-English reporting-rate context when we launch any reporting-improvement scheme,the force owns the story of rising figures from day one, and the public hears a good-news explanation instead of a shoplifting-capital headline.
If we brief comms to have the "why the numbers went up" answer ready at launch,journalists get the explanation first time, and the force is spared months of reactive correction as the same figure is recycled.
If we challenge a commercial ranking's methodology directly, starting with its choice of denominator,the public sees the figure in context, and the force's genuine successes lead the coverage rather than trailing it by a year.

More or Less Policing evidence-based-policing public-confidencestatisticsmediadata

Read the full article →

Technology & Innovation

The Perception of the Police in the British Press

01The problem

Surveys say whether public confidence in the police rose or fell, but not the language that ran alongside it — and a quarter-century of UK police headlines shows the two curves are not the same shape.

02What the evidence says

Public confidence climbed to a peak of around 62–63% in the mid-2010s and has fallen since, to 49% by the year ending March 2025 — its lowest in the series.
ONS Crime Survey for England and Wales
The negative share of police headlines climbed almost without pause, from a few per cent in the early 2000s to a peak of 23% in 2021 and 2022.
Project headline dataset (Guardian Open Platform)
For the first decade the two curves move in opposite directions — confidence rising while coverage darkened — so headline language is its own signal, not a mirror of public opinion.
Project headline dataset (Guardian Open Platform)

03If we act

If we track media tone as a distinct signal alongside survey measures, not as a proxy for public opinion,confidence work reads the information environment the public actually lives in, and the force is not misled by either line alone.
If we publish the limits with any text analysis — single-source scope, headlines only, normalised counts,the findings stay honest and defensible, and the public can trust what the force claims its data shows.
If we resist causal claims the data cannot support,analyses keep the discipline this dataset models, and the force's credibility is protected when the numbers are challenged.

public confidencemediatrustmetropolitan policestop and searchdata

Read the full article →

Technology & Innovation

Does Domestic Abuse Spike When England Play? Reading the Evidence Honestly

01The problem

The famous claim that domestic abuse rises 38% when England lose comes from a single-force study, and the most repeated version of the statistic is the least representative of what the evidence shows.

02What the evidence says

Across three World Cups in Lancashire, abuse reports rose 38% on days England lost — and 26% on days England won or drew.
Kirby, Francis & O'Flaherty (2014)
Ten years of West Midlands data show alcohol-related domestic abuse rising around 47% on the day England win, with no measurable increase in abuse not flagged as alcohol-related.
Trendl, Stewart & Mullett (2020)
The link is likely causal: fixture timing is effectively random, and the rise tracks the match hour by hour — beginning at kick-off, peaking after the final whistle, then fading.
LSE Centre for Economic Performance

03If we act

If we surge response and safeguarding capacity around England match days, timed to kick-off and the hours after the whistle,help reaches victims during the predictable window when the danger actually spikes.
If we target alcohol rather than the scoreline, planning for wins as seriously as losses,the response addresses the real lever, and victims are protected on the nights the data says are most dangerous.
If we quote the evidence accurately — abuse rises around England matches, driven by heavy drinking, win or lose,campaigns and briefings rest on solid ground, and the public gets a true warning rather than a slogan the evidence has moved past.

More or Less Policing safeguarding violence against women domestic abusestatisticsalcohol

Read the full article →

Technology & Innovation

More or Less Policing: how to read the numbers without being fooled

01The problem

Police statistics have never been more available or easier to misread, and quoting a figure without examining the method behind it misleads in either direction.

02What the evidence says

data.police.uk publishes millions of records a month under an open licence, which makes careful reading more necessary, not less.
data.police.uk
The familiar stop-and-search ratio divides searches by residential population, but searches happen where people are, not where they sleep, so the raw ratio cannot say how much is bias and how much is deployment.
A single figure is a dot, not an insight: a force's number means nothing until it is set beside last year, a similar force, or what you would expect.

03If we act

If we ask what exactly is counted on top and what is on the bottom before repeating any police rate,the force stops quoting figures the method cannot support, and the public gets claims that mean what they say.
If we set every figure against last year, a comparable force, or a sensible expectation,a dot becomes an insight, and leaders draw conclusions the public can actually weigh.
If we check the working in the underlying data, noting the month it describes and the day it was pulled,claims can be verified rather than taken on trust, by the force and the public alike.

More or Less Policing stop and searchstatisticstransparency

Read the full article →