Industries / Transport and vehicles / Trucking and freight / Fleet management
Maintenance
Maintenance is the work of keeping every vehicle in a fleet safe and working: drivers' daily checks, scheduled service, repairs, and the records that show what each vehicle costs to keep.
5
Milestones achieved, from 1968 to 2021.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1968
A licence that needs upkeep
Britain's Transport Act 1968 required anyone using goods vehicles for hire or in their own business to hold an operator's licence; small goods vehicles were exempt. Before granting one, the licensing authority had to be satisfied that there would be "satisfactory facilities and arrangements for maintaining the authorised vehicles in a fit and serviceable condition," and a conviction for poor maintenance was grounds to revoke, suspend or cut back the licence. The Driver and Vehicle Standards Agency's (DVSA) guide, updated on 28 April 2025, still sets out how operators meet this: daily walkaround checks by drivers, and safety inspections at the intervals each operator declares when it applies.
Credit
UK Parliament; the Traffic Commissioners, who serve as licensing authorities; DVSA
1970
A shared code for repairs
The American Trucking Associations (ATA) developed the Vehicle Maintenance Reporting Standards (VMRS), a coding system that gives each vehicle part, each kind of repair and each reason for failure its own number. TMC, ATA's Technology & Maintenance Council, dates VMRS to 1970 and calls it a "universal" language for fleets, truck makers and parts suppliers; a technician records a worn part as failure code 44, for example, so repair costs can be added up and compared across fleets. TMC keeps the codes up to date and has added codes for crash-avoidance systems, camera systems and electric vehicles.
Credit
American Trucking Associations; its Technology & Maintenance Council
2010
Old mail trucks cost more
The US Postal Service bought its Grumman Long Life Vehicles (LLVs) from 1987 to 1994 and expected each to last 24 years. In June 2010 its Office of Inspector General (OIG) found that repairing 19,257 of the oldest as they failed, at more than $5,600 a year each, would cost $342 million more than buying new vehicles. By the OIG's August 2020 count, the Postal Service spent $706.2 million in fiscal year 2019 maintaining 141,057 LLVs, about $5,000 each, and nearly 10,000 of them cost more than $12,000 each.
LessonRunning vehicles well past their planned life moves the cost from the purchase budget to the repair shop.
Credit
USPS Office of Inspector General
2015
Repair scores leave public view
FMCSA, the US truck and bus safety regulator, ranks carriers in its Safety Measurement System, part of a program called CSA (Compliance, Safety, Accountability). It uses roadside inspections and crash reports from the past two years, updates the results each month, and gives each carrier a percentile in seven categories, one of which, Vehicle Maintenance, counts defects and missed pre- and post-trip inspections. The Fixing America's Surface Transportation (FAST) Act, signed on 4 December 2015, barred FMCSA from showing the public those percentiles and alerts until a study by the National Research Council was done, its findings were acted on, and the Transportation Department's Inspector General certified that the problems named in a February 2014 report by the Government Accountability Office (GAO-14-114) had been fixed; the inspection and violation records themselves stayed public. FMCSA's site now shows complete results only to enforcement staff and to carriers logged in to their own profiles.
LessonA public safety score can be withdrawn when its link to crashes is in doubt: the Act ordered a study of how well the categories predict a carrier's future crashes before the rankings could be shown again.
Credit
FMCSA; the US Congress
2021
Lower upkeep for electric cars
In March 2019 New York City's fleet office reported that its battery-electric cars had the lowest maintenance costs of all the models it tracked in 2018, with the Chevrolet Bolt lowest among its sedans; it noted that the figures did not account for vehicle age. In April 2021 a study led by Argonne National Laboratory, with researchers from four other US national laboratories, put scheduled maintenance at 6.1 cents per mile for a battery-electric car against 10.1 cents for a gasoline car, 41% less, because oil changes and many filter and fluid changes drop away.
Credit
New York City's Department of Citywide Administrative Services, under chief fleet officer Keith Kerman; Andrew Burnham and colleagues at Argonne National Laboratory
Watch for
AI that sorts fault codes
A fault code is a numbered alert that a vehicle's computer stores when it detects a problem. Bosch announced in March 2026 that it plans to buy Uptake Technologies of Chicago, which makes AI software that predicts failures in fleet vehicles, and to pair it with Bosch's own vehicle data service. An Uptake data scientist told S&P Global Mobility, as PYMNTS reported in June 2026, that one fleet it studied produced nearly 8,000 fault codes per vehicle a year, which AI can cut to 5 to 10 issues worth acting on. Motive launched an AI maintenance product in August 2026 that explains fault codes in plain language, ranks them by severity and turns them into work orders.
Achieved whena named fleet or an outside tester publishes counts of roadside breakdowns, and of failures the software missed, before and after using one of these tools.
Claimed by
Bosch and Uptake Technologies, March and June 2026; Motive, August 2026
Open
01
A public repair-coding test
A public test set of at least 2,000 real repair-order lines, with names removed, each holding the technician's own words and the VMRS codes that two experienced coders agree on. Then score AI models and commercial coding software on how often they pick the right code at each level, from the vehicle system down to the single part.
The briefCredit · unclaimed02
Public repair costs by age
An open table of maintenance and repair cost per mile for government fleet vehicles, by vehicle type, age and power source (gasoline, diesel, hybrid, electric), built from the fleets' own work-order records, with a link to the record behind each row. A fleet could use it to judge when a vehicle has become cheaper to replace than to keep.
The briefCredit · unclaimed
Open 01 · the brief
A public repair-coding test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test set of at least 2,000 real repair-order lines, with names removed, each holding the technician's own words and the VMRS codes that two experienced coders agree on. Then score AI models and commercial coding software on how often they pick the right code at each level, from the vehicle system down to the single part.
Why it is still open
Software that assigns VMRS codes exists. Decisiv launched one in September 2021 and said fleets that checked it found correct system codes, the broadest level, on more than 75% of service jobs; one carrier, Howard Sheppard, reported 77%. Those figures come from the seller and its customers and were not measured on a shared test. A trade article of August 2026 says machine learning "can" suggest VMRS codes and warns that incomplete or inconsistent records weaken the results. This research found no public test set.
Built with what exists
TMC's VMRS code library (TMC Recommended Practice 802); repair orders donated by fleets or obtained from public fleets by records request; language models such as Claude and Gemini, including Google's Gemini Decision 1, which returns probabilities for up to 1,000 fixed-answer questions in one request (see the Note "A thousand questions, one pass"); Python for scoring.
First step
Ask TMC whether VMRS codes may be published alongside a test set, and request one year of repair orders from two city fleets.
Done looks like
The test set and scoring code published, with accuracy at each code level for at least five tools, and the rate at which the two human coders agreed with each other reported as the ceiling.
Open 02 · the brief
Public repair costs by age
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of maintenance and repair cost per mile for government fleet vehicles, by vehicle type, age and power source (gasoline, diesel, hybrid, electric), built from the fleets' own work-order records, with a link to the record behind each row. A fleet could use it to judge when a vehicle has become cheaper to replace than to keep.
Why it is still open
Published comparisons are partial. New York City's 2018 comparison of electric and gasoline cars did not account for vehicle age. The Postal Service figures above sit in audit reports. The American Transportation Research Institute's yearly cost study reports averages for heavy trucks, from data that carriers give it on more than 182,000 tractor-trailers. The only city work-order dataset this research found in the US federal open-data catalog, for Austin's emergency medical services fleet, counts services due and done each month, not costs.
Built with what exists
Public records requests and open-data portals; city and state fleet budget documents; an AI model to pull tables out of PDF reports, with a person checking each figure against its page; Python; a public GitHub repository.
First step
Send records requests to five large city fleets for five years of work orders giving vehicle, odometer reading, labor cost and parts cost.
Done looks like
At least ten public fleets in at least two countries, five years each and 20,000 vehicle-years or more in all, with cost-per-mile curves by age for each power source and the method and data public.