Stolen vehicle recovery in NI: why PSNI said the answers would take 700 hours to retrieve

A police disclosure exposes the question motorists still cannot answer

How many stolen vehicles in Northern Ireland are eventually found? It sounds like the obvious second question after asking how many were taken. Yet an official police disclosure shows why the answer is much harder to establish than an annual theft total. In August 2025, the Police Service of Northern Ireland refused a freedom-of-information request covering stolen vehicles, their recovery, commonly targeted makes and models, the places where theft occurred and the postcodes with the highest numbers. The force said important parts of that information were held electronically but could not be extracted in the requested form without examining individual records.

The distinction matters to anyone trying to assess the real consequences of vehicle crime. A car that is recovered intact within hours, a vehicle found badly damaged weeks later and one that is never recovered are not equivalent experiences for an owner. A theft count records the offence; it does not tell the reader what happened afterwards. Without a clearly defined recovery measure, claims about the chances of getting a stolen car back in Northern Ireland should be approached with caution.

This is a PureCar examination of what the published records establish, what the PSNI has said it cannot readily supply and what motorists should know about the process if their vehicle is located. It does not suggest the force is failing to investigate individual offences, and it does not present a missing national recovery figure as proof that vehicles are not being found. Those would be conclusions the evidence cannot sustain.

The 5,669-offence request and the 700-hour estimate

The central evidence is PSNI freedom-of-information response FOI/15434, dated 22 August 2025. The request covered 1 January 2020 through the latest available point in 2025. It sought annual counts of offences classified as theft or unauthorised taking of a motor vehicle, numbers recovered, the five most frequently stolen makes and models, typical location types and the leading postcode districts. Police treated the request as exceeding the statutory cost limit and declined to provide it in full.

The PSNI said there were 5,669 recorded offences within the period considered, and estimated that retrieving the recovery and make-and-model information alone would take approximately 700 hours. That estimate should not be mistaken for an actual audit of the individual offences: it was the force's projection of the work needed to answer the request. Dividing 700 hours by 5,669 produces about 7.4 minutes per recorded offence, but that is only an arithmetic illustration of the scale of the estimate, not a reported case-processing time.

The refusal did not say that recovery information had never been entered into police systems. On the contrary, it explicitly stated that information sought in the recovery and model questions was held electronically. The barrier identified was how it would have to be retrieved. Nor did the refusal establish that every offence in the range concerned a stolen privately owned car. The recorded offence classification is broader than the everyday expression 'car theft'. These qualifications are essential when turning a technical disclosure into a headline.

One in three offence records lacked the vehicle link sought

One part of the answer goes directly to data quality and usability. The PSNI said that around one third of recorded offences under code 48, theft or unauthorised taking of a motor vehicle, did not have a linked vehicle in the vehicle table. To discover which makes and models were stolen, staff would therefore have to check offence records manually rather than simply export a complete list from the vehicle table.

That does not mean one third of the crimes had no vehicle, that police lost a third of the files or that a third of stolen vehicles were never recovered. A missing link between two database tables is not the same thing as missing knowledge about the real-world incident. Information may be stored in narrative notes or other records, or may not have been captured in a consistently structured field. The PSNI's disclosure did not provide enough detail to quantify each explanation.

For public statistics, the consequence is nevertheless important. If a newspaper were to rank the most stolen makes by counting only those offences successfully joined to the vehicle table, it could produce a distorted result. The missing links might not be distributed evenly across places, dates, reporting practices or vehicle types. Without knowing whether the linked cases represent the entire theft population, the ranking would be a selection from the records that are easiest to retrieve, not necessarily a ranking of the cars criminals most often take.

Why recovery is a more complicated measure than a theft count

Counting recorded theft offences is comparatively straightforward because each record has a crime classification and a reporting period. Recovery is a later event that must be associated with the correct incident and the correct vehicle. One offence may involve circumstances that do not fit the simple picture of one stolen car and one subsequent recovery. A vehicle may also be linked to another investigation, moved between records or recovered after the calendar year in which the theft was reported.

Those timing questions are not technical trivia. Suppose a vehicle is reported stolen late in December and found in January. A table of thefts by the reporting year and recoveries by the discovery year would not create a meaningful recovery percentage simply by dividing one year's numbers by the other. A more defensible measure would follow an identified group of stolen vehicles for an explicit period after the report, while explaining whether later recoveries can be counted and whether partial recovery, write-offs and abandoned vehicles are included.

The force's refusal did not provide such a longitudinal measure. Its answer instead explained that recovery status would require manual examination of each offence record. That is a reason to resist unsupported recovery-rate claims, not a reason to assume the recovery operation itself does not happen. A person who sees a precise Northern Ireland percentage should ask which offences and vehicles are in the calculation, when the outcomes were measured and how cases were linked.

A second refusal shows this was not a one-off difficulty

The August 2025 response was not the first occasion on which a vehicle-theft information request ran into the way police data were stored. In February 2025, PSNI response FOI/13641 dealt with a request for incident-level motor-vehicle theft data from 2020 to 2024. That request sought identifiers, dates, times and sufficiently detailed locations. The force said extracting the information would involve a manual check of roughly 6,200 relevant code 48 and aggravated-taking offences, with an estimated workload exceeding 1,000 hours.

The February response also explained a further complication: a vehicle marked as stolen in a vehicle table may have been stolen earlier and subsequently used in a different offence. A count based on that status could count the same vehicle's appearance in multiple records rather than count fresh theft events. It said around a quarter of vehicles marked stolen lacked a corresponding qualifying theft offence, or were linked to an occurrence with a different offence type. The response also identified missing vehicle links in roughly a third of relevant offence records.

The 6,200 offences and 1,000-hour estimate in February must not be compared directly with the 5,669 and 700 hours in August as though they were a measured improvement or deterioration. The requests differed in period, scope and requested output, and police were estimating the cost of different retrieval tasks. What can fairly be said is that two separate disclosures described manual reconciliation between offence and vehicle records as a substantial obstacle to detailed public reporting.

A newer theft total does not resolve the recovery question

In February 2026, a separate PSNI disclosure, FOI/16599, provided a national and district breakdown of offences under code 48 for calendar years 2023, 2024 and 2025. It reported 1,036 such offences in 2023, 1,015 in 2024 and 911 in 2025. The 2025 figure was described as provisional when published because some records were still subject to validation. That gives readers a clear, defined trend for one offence category: 104 fewer recorded offences in 2025 than in 2024, a fall of approximately 10.2%.

But that later publication did not answer the different question of how many stolen vehicles were located or returned. The existence of a newer annual crime total cannot be used to manufacture a recovery series. Neither should a decline in recorded thefts be advertised as proof that the prospects of recovery have improved. Incidence and recovery are different dimensions of the problem, and each requires its own evidence.

PureCar's Northern Ireland vehicle-crime explorer supplies another useful view: police-mapped records over time and across council areas. Those records are not a list of recoveries, and their broad 'vehicle crime' category also includes offences other than theft of a whole motor vehicle. The map cannot close this evidence gap. It is valuable for exploring where and when mapped reports were recorded, but not for telling an owner what proportion of stolen cars ultimately came home.

The 18-hour limit is a legal threshold, not a measure of public importance

Why would a request about stolen vehicles be refused if the information exists? Under section 12 of the Freedom of Information Act 2000, a public authority can decline a request when it reasonably estimates that identifying, locating, retrieving or extracting the information would exceed the applicable cost limit. In the August 2025 response the PSNI set out the threshold for authorities such as itself: £450, calculated at £25 per hour, or 18 hours of permitted work.

In practical terms, a projected 700-hour exercise is far beyond a threshold designed for ordinary requests. The cost exemption does not determine whether journalists, researchers or the public have a legitimate interest in recovery statistics. It determines whether that particular request can be compelled under the framework at the estimated extraction cost. It is possible for an important public-interest question to be difficult to answer within the legislation's routine limits.

The PSNI also offered a narrower route for some material. Its August response said it could address the requests for annual theft totals, location types and postcode areas within the appropriate cost limit, while offering no viable refinement to produce the recovery and model data under the same limit. It did not publish those revised answers in the refusal itself. That distinction gives any future requester a practical starting point: narrow the question and do not assume an all-in-one request must succeed merely because some of its fields exist in police systems.

What actually happens when the police find a stolen vehicle

The absence of a published national recovery percentage does not mean the public lacks information about the recovery process. In its guidance on recovering stolen vehicles, the PSNI explains that when officers locate a reported stolen vehicle, they arrange for a contracted recovery operator to take it to safekeeping on behalf of the owner or insurer, depending on the insurance arrangements. The stated reasons include preventing further theft or damage, avoiding risks to other road users and preserving the possibility of a forensic examination.

That sequence is easy to overlook because owners understandably focus on getting their car back as quickly as possible. But a recovered vehicle may be unsafe to drive, may need to be examined and may have been exposed to uses or damage that are not visible from the outside. A car that looks intact in a photograph is not necessarily roadworthy. The police describe removal as a way to protect the vehicle and reduce the risks associated with leaving it where it was found.

The PSNI states that owners are told where the vehicle is being held and should inform their insurer where their policy makes that appropriate. Charges for removal and storage may have to be paid to the recovery operator before collection; the guidance says storage charges become payable from the day after notification that the vehicle is available. The precise charge depends on the applicable regime and the circumstances. A headline suggesting every recovered stolen car is simply handed back without administrative or financial consequences would mislead owners.

The costs and choices facing an owner after recovery

Recovery can mark the beginning of another complicated phase rather than the end of the disruption. The owner may need to find out whether the vehicle is driveable, whether locks or keys need replacement, whether an insurer has already progressed a settlement and whether the recovered vehicle needs a further inspection. These are real decisions even if the public statistics cannot quantify how often each situation occurs.

The police guidance refers to the Road Traffic Regulation (Northern Ireland) Order 1997 as part of the legal basis for removing vehicles and recovering reasonable costs. A separate PSNI disclosure about roadside recovery fees, FOI/14433 from May 2025, explains that charges can vary with the size and weight of the vehicle and that PSNI does not receive the payments collected through these recoveries. It would be inaccurate to infer a universal fee from these statements, or to treat every police-directed recovery as one resulting from theft.

Drivers should obtain the recovery reference, the operator's location, the notified release date and an itemised explanation of applicable charges before making arrangements to collect. Where an insurer is involved, the owner should establish who is arranging transport and inspection, whether charges are covered and whether settlement or salvage arrangements have changed their rights to collect the vehicle. Those are practical questions; the answers depend on the specific claim and documentation, not on a generic public-crime table.

Finding the vehicle yourself is not the same as recovering it safely

The PSNI has a separate warning for people who discover their own stolen vehicle before police have arranged its removal. Its published guidance recommends leaving the vehicle where it is and informing police without delay. An owner who simply drives it away may not know whether it has been damaged, altered, mechanically compromised or involved in another incident. There is also a risk of interfering with evidence.

If an owner chooses to take charge of the vehicle, the PSNI advises arranging a comprehensive examination by a qualified garage or mechanic before it is driven, at the owner's expense. The force's explanation is framed around safety and the limits of police responsibility after the owner takes possession. The appropriate response will depend on the immediate situation, and a suspected crime in progress or threat to safety should be reported through emergency channels rather than approached personally.

This guidance shows why a public recovery statistic, even if available, would need more than a simple yes-or-no flag to describe the human outcome. A vehicle located but not yet fit to use has technically been found, while the household may still face hire-car costs, lost work and uncertainty. A useful assessment of recovery would ideally distinguish finding the vehicle, returning it to the owner and returning it in usable condition. Those are analytically different results.

What the published data cannot tell van owners or businesses

For a van-dependent business, the absence of a reliable public recovery-and-model breakdown has a particular consequence. A plumbing company, electrician or delivery operator cannot use the August disclosure to determine which van models were most frequently stolen or how often work vehicles were retrieved. The police identified a difficulty retrieving makes and models for the records requested, not a ranked list of the brands most at risk. Any article naming a model as Northern Ireland's most stolen on the strength of this refusal would be inventing the answer.

The difference between a stolen van and stolen tools inside a van is also important, but neither can be quantified from the refusal alone. A business can lose its working day even when the vehicle remains in place. Conversely, the vehicle may be recovered but equipment or stock may be missing. Those experiences are not interchangeable with offence code 48. For owners, the actionable response is to keep a verified list of vehicle identifiers and equipment details and to check that the relevant insurance policies cover the losses they actually face.

It would also be misleading to claim that the lack of a publicly supplied model ranking means criminals have no preference for particular vehicles. The disclosure is about accessible data, not criminal behaviour. Research into which makes or technologies are more often targeted would need validated vehicle-level information and, for meaningful risk comparisons, some measure of the number of those vehicles in use. Raw theft counts by make can be heavily influenced by how common a model is on the road.

Why the broad map cannot be used as a recovery dashboard

The PureCar crime map is designed to answer a different set of questions: how mapped vehicle-crime records are distributed over time and across Northern Ireland's council areas. It is not an index of specific seized vehicles, insurer settlements or stolen-property returns. Even when a public police-data system displays an outcome category, that information is about the recorded crime's status, not automatically a vehicle recovery or a return to the policyholder.

A case marked as having no identified suspect, for example, cannot be assumed to involve a vehicle that was never located. The reverse is also true: a case with a criminal justice outcome does not prove the property was returned intact. Recovery, conviction, charge, claim settlement and the condition of the vehicle are distinct variables. A rigorous newsroom should not convert one into another because it makes a cleaner chart.

Likewise, the approximate public mapping coordinates used to protect the location of crime reports do not identify storage yards, recovery sites or the place a stolen car was found. They are not evidence that a particular car park, street or business was responsible for an offence. PureCar's vehicle-crime report can help readers understand the published mapped series, but the question of the eventual fate of each stolen vehicle requires separate, verified police or recovery records.

The accountability test: which figures would make the system clearer?

The useful challenge for public reporting is not to demand that a complex police database produce a simple answer at any cost. It is to define a small set of questions that could eventually be reported consistently without manual reconstruction of thousands of cases. First is the number of qualifying theft offences in a clear calendar or financial year. Second is the number of distinct vehicles attached to those offences, with an explanation of cases lacking a usable vehicle link. Third is the number subsequently located, using a stated follow-up period.

A fourth measure would distinguish vehicles located but not returned from vehicles collected by the owner or insurer, without publishing personal data. A fifth could record recovery timing in broad bands: within 24 hours, within seven days, after seven days, or not recovered by the reporting cut-off. A sixth could distinguish recovery location, if disclosure could be made safely and at a sufficiently aggregated level. These are proposals for more informative statistics, not measures PureCar says are already being published by PSNI.

The missing-link figure in the 2025 disclosure is a strong reason to be cautious about immediate targets or league tables. Before any recovery rate could be trusted, data collectors would need to know how reliably offences and vehicles were associated and how updates were made over time. Better reporting begins with a consistent definition, audit of incomplete records and transparent footnotes; it cannot be achieved simply by attaching a percentage sign to the first extractable figure.

A narrower freedom-of-information question might be more useful

The PSNI said the August 2025 request could have been narrowed for some categories but not, within the cost limit, for the recovery and most-stolen-model questions. That should shape any follow-up. Rather than asking for five complex extracts covering many years, a reporter could first request the data dictionary or an explanation of the fields used for vehicle recovery, with sensitive information excluded. Another narrow question could ask whether a routinely generated annual recovery count exists, what it includes and how it is quality assured.

A more ambitious future request might seek a small aggregated breakdown from a defined period, with a clear distinction between code 48 offences and aggravated vehicle taking, and with the response limited to fields that already exist in a report. The force may still be unable to provide it. A refusal should be reported accurately rather than treated as proof that the underlying operation was ineffective. But an accessible explanation of the obstacles would help the public understand what the figures can and cannot establish.

There is a wider lesson for transparency in the design of public systems. Data originally collected for case management can be difficult to repurpose for statistical analysis. A record might be perfectly adequate for an officer to investigate a particular case and still be unsuitable for a clean, automatic aggregate. Conversely, the fact that statistics are hard to extract can justify questions about whether important public-interest measures should be structured more consistently in future. Both propositions can be true.

What an individual motorist should do after a theft

The practical advice in this investigation is deliberately narrower than a promise that a stolen vehicle can be traced or recovered. If a vehicle is missing and theft is suspected, report it promptly to PSNI, using emergency services if there is an immediate threat or a crime in progress. Record the registration, make, model, VIN if available, identifying features and any relevant tracker or camera information. Keep copies of original video evidence rather than relying on an edited clip or screenshot. Do not confront anyone suspected of taking the vehicle.

Contact the insurer in accordance with the policy terms and keep the crime reference, claim reference and details of any conversations about recovery. If police later say the vehicle has been found, ask where it is held, when storage charges begin, whether it has been released for collection, whether there is a safety or forensic restriction and whether the insurer or a recovery contractor is handling transport. Before returning it to use, arrange an appropriate mechanical and security assessment, especially if keys, locks, electronic systems or bodywork may have been affected.

The PSNI's prevention guidance emphasises locking vehicles and verifying that electronic locks have operated, keeping keys secure, removing visible valuables and using appropriate physical security. Those recommendations remain sensible without a localised list of the most stolen models or a national recovery rate. A statistic may help policymakers evaluate the system, but the first decisions facing a victim are about safety, evidence, insurance and returning the vehicle to a usable condition.

What has been established — and what remains open

The published evidence supports three clear conclusions. First, the PSNI said in August 2025 that extracting recovery and make-and-model details from the requested vehicle-theft records would require substantial manual work, estimating around 700 hours for those aspects alone. Second, the force identified a practical data-linkage limitation, reporting that roughly a third of relevant code 48 offences did not have a directly linked vehicle in the vehicle table. Third, police have published more recent offence totals, including the 2025 figure of 911, without those disclosures themselves supplying a matching national recovery percentage.

The evidence does not establish how many stolen vehicles were recovered in 2026, which makes or models were most frequently taken, what share of recorded thefts involved permanently missing vehicles or how effective individual investigations were. Nor does it show that those outcomes cannot be known by police officers managing their cases. The disclosed problem concerns producing the requested aggregated statistics within a particular legal cost threshold.

For Northern Ireland motorists, the missing recovery figure is not an abstract spreadsheet problem. It is the unanswered question that follows a stolen-car headline: what happened next? PureCar will keep its mapped crime figures clearly separated from theft offences and eventual recovery outcomes. Until a properly defined and verifiable recovery series is available, that separation is more informative than a confidently stated number that the public record does not support.

Sources and editorial limits

This investigation draws on PSNI FOI/15434 (22 August 2025), PSNI FOI/13641 (19 February 2025), PSNI FOI/16599 (4 February 2026), PSNI recovery and prevention guidance, and PSNI FOI/14433 on recovery fees. It distinguishes the publication dates of those documents from the time periods they describe. The first two refusals describe retrieval costs and database structure at the time they were issued; they are not evidence that no system or reporting improvement has happened since then.

Every reference to 5,669 offences, 700 hours, 6,200 offences or more than 1,000 hours is an attributed estimate or figure from the relevant PSNI response, not a new PureCar calculation of thefts or recoveries. The 911 figure relates to one specified recorded offence category in calendar year 2025 and was provisional when published. A completed theft enquiry, a found car, a repaired car and a returned car are different outcomes. This reporting deliberately does not manufacture a recovery rate from data that cannot support one.