At Least One in Six Battery and Alternator Complaints to NHTSA Describe a Car That Died on the Road

In November 2024 the owner of a 2019 Kia Sorento filed a complaint with the federal government. On a busy highway at 70 miles per hour, they wrote, the car "lost power": "the battery light flashed twice and then the instrument cluster had all warning lights." Then the line this study is about: "I lost power steering and brakes." (ODI 11623788.)

That owner is not unusual. On 23 September 2026 we downloaded every vehicle complaint NHTSA's Office of Defects Investigation has published for complaints received since 1 January 2015, 780,071 distinct complaints, and read the owner's description in each one by published rules. 10,776 describe a 12-volt battery or charging system that failed. In 1,849 of them, 17.2 percent, the car died, stalled or lost power with the owner driving it. When the owner names the alternator, the share rises to 25.2 percent. And the winter spike that pre-season battery advice leads you to expect is not in this file.

I built PulsCar, a tool that listens to a 30-second recording of an engine, after a string of shop visits in Phuket in 2025 that never found the sound my car was making (the story is on the About page). That is my bias. It is also why every rule, count, script and file hash behind this page is published for anyone to re-run, and why the section on what our own product reads is as exact as we could make it.

The short version

What did PulsCar's analysis of NHTSA battery and alternator complaints find? That a failing 12-volt system is not only a cold-morning problem. Across 780,071 vehicle complaints received from January 2015 to 21 September 2026, 10,776 describe a battery or charging system that failed. In 17.2 percent of those, 1,849 complaints, the owner writes that the car died, stalled, shut off or lost power while it was being driven, and 15.3 percent of that group record a tow. When the alternator is named, the on-the-road share is 25.2 percent. 37.5 percent describe a car that would not start or a battery found dead when parked. The monthly rate shows no winter spike, and 199 of the 271 battery, alternator and starter recalls filed since 2010 name fire as the consequence.

How did PulsCar decide whether a complaint described a battery or charging failure on the road? By reading the owner's own text with published rules, in four steps. First, the description must mention a battery, an alternator, the charging system or a voltage regulator. Second, complaints about an electric car's high-voltage battery, a key fob battery or a motorcycle are dropped unless they also say 12 volt. Third, the text must describe a failure: a dead, weak or drained battery, an alternator that failed or was replaced, a battery light or charging message, low voltage or a jump start. Fourth, a complaint counts as on the road only if one sentence both places the car in use and says it died, stalled or lost power. Hand audits of random samples are reported in the method section.

Does PulsCar's study mean batteries fail more often on the road than in the driveway? No, and the file cannot show that. NHTSA collects safety complaints, and an owner whose old battery will not start the car on a cold morning usually calls roadside assistance, not the federal government. So the file over-represents failures that feel like safety problems, and the 17.2 percent figure describes what owners who filed wrote, not how often each kind of failure happens across all cars. What the file does show is narrower and still useful before winter: a real share of reported battery and charging failures happen at speed, in traffic or on a highway, often with the steering, brakes or lights affected, and when an owner mentions a warning, it often arrives in the same sentence as the failure.

Key findings at a glance

What we counted, and what we did not

The complaint source. NHTSA publishes its complaint database as flat files. We used the three that cover complaints received from 2015 onward, COMPLAINTS_RECEIVED_2015-2019, COMPLAINTS_RECEIVED_2020-2024 and COMPLAINTS_RECEIVED_2025-2026, downloaded on 23 September 2026: 1,105,443 rows. A complaint that names several components appears once per component, so we de-duplicated by ODINO, NHTSA's reference number, and kept product type V, vehicles. That left 780,071 complaints, received from 1 January 2015 to 21 September 2026. The SHA-256 hash of each file we used is in the published method file, so anyone can check they are running the same data.

The text. Every judgment is made on the description of the complaint, field 20, CDESCR: the owner's account in their own words, or NHTSA's standard summary of a phone call, which begins "The contact owns a". All matching is case-insensitive, because the file switched from capital letters to ordinary text in 2021 and a case-sensitive search silently loses most recent complaints.

Why not the component field. We expected to start from NHTSA's own component codes for the battery and the alternator. It turned out not to work: across 780,071 complaints, only 572 carry the 12-volt battery component and 370 the alternator component. Owners and call-center staff overwhelmingly file these problems under the generic electrical system entry, or under engine, power train or unknown. Of the 10,776 complaints in the failure set defined below, only 395 carry either code, while 7,991 carry the generic electrical system entry. A study built on the component field would have missed more than nine complaints in ten. The text is the only place the story is told.

Step 1, mention. The description contains battery or batteries, alternator, charging system or voltage regulator as whole words: 37,912 complaints.

Step 2, exclusions. We drop complaints whose text is about an electric car's high-voltage battery (hybrid battery, high voltage, traction battery, battery pack, state of charge, range, kWh, charging station, plug-in and similar), a key fob or remote battery, unless the text also says 12 volt, 12V, auxiliary battery or low-voltage battery. We drop motorcycle complaints. 33,196 remain.

Step 3, failure described. The text must describe the 12-volt system failing: a dead, weak, drained, discharged, failed or bad battery; a battery that died, went dead or keeps dying; an alternator that failed, went out, quit or was replaced; not charging; a charging system warning, fault or message; a battery light; low voltage or a voltage drop; or a jump start. A match is ignored when the words just before it are hypothetical (afraid the battery might die) or negate it (has never tested at low voltage). 10,776 complaints remain. This is the failure set, and every percentage below is a share of it unless stated otherwise.

Step 4, where it happened. Each complaint in the failure set lands in exactly one of three groups:

The on-the-road test is deliberately strict. A complaint that says "the car would not start" in one sentence and "while driving the radio flickers" in another does not count, because the rule requires the driving and the dying in the same sentence. Among complaints where the setting is clear, the on-the-road share is 31.4 percent, but the headline uses the whole failure set as the base, because an unclear complaint is not evidence either way.

The audits. Rules on free text make mistakes, so we read samples in full, drawn with a fixed random seed that is in the published script. Of 40 random complaints from the failure set, 32 describe a 12-volt battery or charging problem actually happening. Seven use the words around a different fault, most often a battery light that came on during an engine failure, once a jump start that did not help and once a car acting like the battery had died; one could not be judged from its text. So roughly one complaint in five in the failure set is about something else in which the battery shows up as a symptom, and the percentages below carry that noise. Of 40 random on-the-road complaints, 37 describe the car dying, stalling, shutting off or losing power while in use; three do not (an infotainment screen and a stereo switching off, and an owner writing that the car could have shut down), and in a few of the 37 the owner's account points to a cause other than the 12-volt system, such as an engine failure. In the other direction, of 40 random complaints outside the on-the-road group, 5 do describe the car dying or losing power while being driven, in words the rule does not catch, such as lost total battery power or suddenly stopped, or with the driving and the stall in different sentences. Scaled to the whole set, the misses outnumber the false matches, so the true on-the-road share of these complaints is probably nearer one in four than one in six. We report the rule's count because it is the one anyone can reproduce. One change was made after a first reading: the rule initially missed NHTSA's own call-center phrase lost motive power, so we added it, excluded lost power steering, and drew the final samples afterwards.

Recalls. From NHTSA's recall flat file FLAT_RCL_POST_2010, downloaded the same day, we kept vehicle campaigns with at least one record under the components 12V/24V/48V battery (including its cables), alternator/generator/regulator or starter assembly. The consequence class comes from the consequence summary the manufacturer filed: fire if it names fire, stall or loss of power if it names a stall, a shutdown or a loss of motive or electrical power, both if it names both. For the largest campaigns we opened the manufacturer's Part 573 Safety Recall Report and copied the vehicles and model years exactly as listed there. We did not use the recalls-by-vehicle lookup for years: it attributes a campaign's whole year span to every model in it, and we have caught it doing exactly that before.

What we did not do. We did not compute any failure rate. We did not rank models or brands. We did not estimate what any repair costs. We did not use any data from our own app or its users. We did not treat any complaint as a verified fact.

Finding 1: one in six describe the car dying on the road

The headline is 17.2 percent: in 1,849 of the 10,776 complaints that describe a failing 12-volt battery or charging system, the owner writes that the car died, stalled, shut off or lost power while it was being driven.

That is not the picture most pre-winter advice paints. The usual story is a car in a driveway at dawn, a slow crank, a click, a jump start from a neighbor. That story is in the file too, 4,040 times, 37.5 percent of the failure set. But one complaint in six describes something else: a car that was already moving when the electrical supply gave out.

The road group is towed four times as often as the no-start group. 15.3 percent of on-the-road complaints record a tow in the form's tow field, against 3.8 percent of no-starts and 5.1 percent of all complaints in the file. A car that dies in a driveway can often be jump-started where it stands. A car that dies in the left lane usually cannot.

The speeds are ordinary. Among on-the-road complaints that fill in the speed field, the median is 40 mph. These are ordinary driving speeds, not a highway-only problem and not a parking-lot one.

The headline is conservative on purpose. 45.4 percent of the failure set does not say where the failure happened, and we did not guess. If you count only complaints where the setting is clear, on the road versus no start, the on-the-road share is 31.4 percent. We report the lower figure because an unclear complaint is not evidence. Our hand audit, described in the method section, points the same way: the strict rule misses more road failures than it wrongly includes, so the true share among these complaints is probably higher than one in six, not lower.

Four more owners, in their own words, from complaints received from 2024 to 2026:

• ODI 11749425, a 2025 Hyundai Ioniq 5, filed 9 July 2026: on a family road trip the car "suddenly showed red 'Check Electric Vehicle System' and 12V battery warning lights and then completely lost power." The dealer diagnosed a failed integrated charging control unit, the part that charges the 12-volt battery.
• ODI 11731790, a 2020 Ford Explorer, filed 17 April 2026, in NHTSA's summary of the call: "while driving approximately 65 MPH, the vehicle lost automotive power and stalled." The messages "Charging System Fault" and "Service Steering" were displayed, and the owner reported "an abnormal whining sound" before the stall.
• ODI 11727324, a 2020 Ram 1500, filed 26 March 2026: "While driving my Ram Rebel 2020, the eTorque alternator failed, causing the vehicle to lose electrical power and forcing me to be towed."
• ODI 11589238, a 2021 Ford Bronco Sport, filed 17 May 2024: "while driving at undisclosed speeds, the vehicle stalled without warning," with the message "Charging System Service Now" displayed. A new battery did not fix it.

These are owner accounts, not findings. Each is included because it states what this page measures, not because the named model is worse than any other. Each quotation is the verbatim text of a single complaint, and the ODI number beside it pulls that exact record from NHTSA's public complaints API. None is a composite, and no name, mileage, date or outcome here was supplied by us.

Finding 2: when the owner names the alternator, it is one in four

Split the failure set by whether the owner names the alternator anywhere in the text.

25.2 percent of alternator complaints describe the car dying on the road, against 14.5 percent of the rest. The no-start share moves the other way, from 43.1 percent to 20.4 percent.

That split is what the physics of the 12-volt system predicts. A battery that has lost its capacity usually shows itself when the car has been sitting, because the starter motor is the single largest load it ever carries. An alternator that stops charging shows itself while the engine is running: the car keeps going on the battery alone, and every system that needs 12 volts, the ignition, the fuel pump, the engine computer, the electric power steering, the lights, draws the battery down until something drops out. The owner is behind the wheel when it happens.

Alternator complaints come from older cars, but not old ones. The median vehicle age at the time of the complaint is 6 years for alternator complaints, 5 years for the on-the-road group and 4 years for no-starts. Where the owner gives mileage, the medians are 80,000, 75,000 and 49,000 miles.

Some of the difference is recalls. Several large alternator campaigns, including two Chrysler recalls covering more than 400,000 vehicles each, sent owners to NHTSA's site, and a complaint written by someone who has just read a recall notice is more likely to name the part. We cannot separate that effect from the physics in this file; we can only say that both point in the same direction.

Finding 3: what else goes when the voltage goes

A car does not stop all at once when the charging system fails. Owners describe a sequence, and the order matters for anyone who has to get off a road. Among the 1,849 on-the-road complaints:

Read these as mentions, not diagnoses. An owner who writes brakes may mean the pedal went hard, the ABS light came on, or that they braked to stop. A steering mention is more specific: it almost always describes the wheel going heavy, which is what happens when an electric power steering system loses its supply. Either way, these are the systems owners bring up when they explain why the failure frightened them.

The order in the complaints is familiar. In the samples we read, the story usually starts with a light or a flicker, sometimes a battery light, often a cluster of unrelated warnings as the voltage sags below what the modules need. Then the radio or the screen resets, the steering goes heavy or the transmission stops shifting, and the engine stops. The order varies from car to car, and several owners describe losing the power steering or the hazard lights while still in traffic.

Newer cars add a twist. Cars with automatic stop-start shut the engine off at lights by design and restart it on a battery that is expected to be healthy. Ford's recall 25V-019 describes exactly that: a 12-volt battery that degrades suddenly while driving, leaving the vehicle unable to restart after an automatic stop or stalling as it comes to a stop at low speed. An owner experiences it as a stall at an intersection. See Finding 10.

Finding 4: the warning, and how long it gave

The obvious question is whether owners got a warning before the car died. The file answers part of it.

Fewer than half name a battery or charging warning at all. 42.9 percent do. Most of the rest mention some light or message, frequently a cluster of them, and 19.1 percent say in plain words that nothing warned them.

Almost nobody says how long they had. Only 27 of 1,849 on-the-road complaints both name a battery or charging warning and give any time or distance, ten minutes later, two miles down the road, and even those do not always measure from the warning to the stop. The file cannot tell you how long a car keeps running after the charging light comes on, and we have not found a public dataset that can. It depends on the battery's condition, the electrical load and the car.

When a warning appears, it often appears with the failure. In the samples we read, the battery light and the stall are frequently in the same sentence: the battery light came on and the car shut off. That is consistent with how charging warnings work on many cars: they react to the voltage or the alternator's own signal as it fails, not to the slow decline of a battery. A warning of this kind is information about the next few minutes, not the next few weeks.

Some warnings are not about the battery at all. A battery light that appears together with an oil light and a check engine light while an engine seizes is part of an engine failure, not a charging failure. Those complaints are in the failure set because the owner wrote battery light, and they are one reason the audit numbers in the method section matter.

Finding 5: winter is not the peak in this file

We went looking for the winter spike. Complaints of every kind are unevenly spread over the year, so each figure below is a rate: battery and charging failure complaints per 100 complaints of any kind with an incident in the same calendar month, incident years 2014 to 2026.

Per 100 complaints of any kind with an incident in the same month. Monthly counts behind each cell are in the downloadable table.

The whole failure set moves in a narrow band, from 1.26 per 100 in May to 1.49 per 100 in August. August, January, February and November are the highest months; May, June and March are the lowest. Winter is among the high months, but it shares that place with late summer.

No-starts dip in late spring rather than peak in winter. The no-start rate is highest in February, at 0.57 per 100, and lowest in June, at 0.44. December, January, April and August sit within a few hundredths of the top, so the cold months are among the highest without standing apart.

On-the-road failures lean toward summer. The on-the-road rate peaks in July, at 0.29 per 100, and is lowest in May, at 0.19. Monthly counts in this group are small, 127 to 213, so read the shape rather than any single month.

Cold states swing more, but not only in winter. We split complaints by the owner's state into a cold group and a hot group, both listed in the method file. In the cold states the rate runs from 1.11 in June to 1.49 in November; in the hot states from 1.27 in May to 1.53 in August. The cold states swing further, and their high months run from July through January rather than clustering in winter. If cold alone drove these complaints, the cold states would show a winter bulge that the hot states lack. What they show instead is a deeper early-summer low and a long high season from July to January.

What this does and does not mean. It does not mean batteries do not fail in winter; the federal file is a poor place to count cold-morning no-starts, for the reasons in the section below on why the file sees so few frozen driveways. It does mean that the failures owners report as safety problems are spread across the year, and that preparing a battery for winter is not the same as preparing a car for a charging failure on the road.

Finding 6: the voltages owners quote

Some owners measure before they write. 99 complaints in the failure set quote at least one voltage between 5 and 16.5 volts, 157 readings in all. That is a small sample, and each reading comes without the one fact that makes a voltage meaningful: whether the engine was running. We publish the distribution because it is the only place in the file where owners report a measurement rather than a feeling.

The largest group is below 11.8 volts. 41 complaints quote a reading under 11.8 volts. In the ones we read, most are batteries measured after the car would not start, several of them at 4 to 7 volts, far below anything a starter can use; the rest are low charging readings or a dashboard or phone notice of low voltage.

The next largest groups sit on either side of a healthy charging voltage. 32 complaints quote a reading between 13.2 and 15.0 volts, most often to show the charging system putting out a normal voltage while the battery kept dying anyway, or a reading that swung up and down; 26 quote a reading between 12.4 and 12.99 volts, most often a battery tested at rest, frequently after a shop called it fine.

Owners who quote numbers are not typical owners. They are the ones with a meter, a scan tool or a truck that shows voltage on its screen. The table says what kind of numbers people write down when something goes wrong; it does not say what voltage a failing battery or alternator usually shows.

Finding 7: electric cars have a 12-volt battery too

An electric car has no alternator and no starter, but it still runs its computers, locks, lights and contactors from a 12-volt battery, kept charged by a converter from the high-voltage pack. When that small battery or its charger fails, the car can refuse to start or shut down while driving, however full the big battery is.

Read the electric row with care. Complaints that use high-voltage words (traction battery, battery pack, range, kWh and similar) are excluded unless they also say 12 volt, so the 406 electric complaints here either name the 12-volt or auxiliary battery or use none of those words. Some of the second kind may still be about the big battery, and an owner who writes about the 12-volt battery in the same breath as the range is left out. The row is a pointer, not a measurement.

The recalls say it more clearly. Four campaigns in the recall file, 24V-204, 24V-868, 24V-200 and 24V-867, cover Hyundai, Genesis and Kia electric cars whose integrated charging control unit could stop charging the 12-volt battery, and each manufacturer's consequence summary reads "A loss of drive power increases the risk of a crash." The largest of them is in the table in Finding 10.

The powertrain label comes from the model name, because the fuel-type field is blank in most complaints. A car whose model name does not reveal that it is electric or hybrid is counted in the last row.

Finding 8: newer cars, older cars

Model years 2016 to 2020 are the no-start years in this file, at 47.0 percent. In the complaints we read from these years, a common story is a fairly new car whose battery went flat while parked, again and again, which is how a parasitic drain presents. The largest make in the failure set is Subaru, with 1,427 complaints, of which only 2.1 percent describe a failure on the road.

Model years 2011 to 2015 are the road years, at 21.7 percent on the road, the highest of any band. That band includes the Chrysler, Dodge and Jeep vehicles covered by the two large alternator recalls in Finding 10, and Dodge and Chrysler complaints name the alternator more often than any other make: 62.0 percent of Dodge failure complaints do.

By year received, the road share has fallen. In complaints received from 2015 to 2019 it ran between 20.9 and 21.8 percent; from 2020 on it has run between 11.6 and 17.6 percent. We do not know why. Newer cars, a different mix of owners filing, the end of the large alternator recalls and a change in how NHTSA's call center writes summaries are all possible, and the file cannot tell them apart.

This table is not a ranking. It lists every make with at least 150 failure complaints. Counts follow sales, fleet age, recalls and how motivated a given group of owners is to file. It is published because it shows the two patterns this page describes, parked-car drains and on-the-road charging failures, landing on different makes.

Finding 9: most battery, alternator and starter recalls are about fire

Complaints tell you what owners experienced. Recalls tell you what manufacturers reported to the government. NHTSA's recall file lists 271 vehicle campaigns from 2010 to 2026 with at least one record under the 12-volt battery, battery cable, alternator or starter components, covering 8,941,116 vehicles if each campaign's population is counted once. A car in two campaigns is counted twice, so that total is recall notices, not distinct cars.

199 of 271 campaigns, 73.4 percent, name fire. Battery cables that chafe, terminal covers that do not seat, alternator wiring that shorts, a battery that moves in its tray: the recall file's 12-volt problems are mostly about heat and current in the wrong place, not about a battery too weak to start a car. Only 3 campaigns name a no-start as the consequence without a stall or fire.

72 campaigns name a stall or a loss of power. Those are the recall-file version of Finding 1: a charging or battery defect that can stop the car while it is moving.

Many campaigns are trucks, buses and motor homes. Heavy vehicles run 24-volt systems and long battery cables, and their manufacturers file many small campaigns. The largest campaigns by population, below, are passenger vehicles.

Finding 10: the largest campaigns, as the manufacturers described them

The table lists the largest campaigns by population. The vehicles and model years are copied from each manufacturer's Part 573 Safety Recall Report, the document a manufacturer must file with NHTSA, not from NHTSA's recalls-by-vehicle lookup, which can attribute a campaign's full span of years to every model in it. Where a manufacturer amended its report, we used the latest amendment. If your car is listed, check your VIN at nhtsa.gov/recalls: a recall covers specific vehicles built in specific windows, not every car of that name and year.

Campaigns whose Part 573 report NHTSA received from 2015 onward, largest first. Vehicle counts are the manufacturer's population in its latest report; the consequence class is our rule applied to the consequence summary it filed. Every campaign in the study, with a link to its report, is in the downloadable table.

A few of these deserve a sentence each.

23V-734, the largest, is about the wrong battery. Toyota's report says the specified size for these RAV4s is SAE Group 35, and that the hold-down "may not sufficiently hold in place certain aftermarket SAE Group 35 batteries" with a smaller top, so a battery can move and short. It is a reminder that a replacement battery has to fit the tray and the clamp, not just the terminals.

17V-418 names road salt. Honda's report says that if "moisture containing conductive substances (such as road salt)" enters the battery sensor, it could short and cause a fire.

17V-435 is an alternator recall for Dodge, Chrysler and Jeep models; those three makes account for 1,264 of the 2,655 complaints in this study that name the alternator. FCA's report says a diode failure "could result in a rapid failure of the alternator with little or no warning; in some instances, a warning light may appear immediately prior to failure," and that "system voltage may drop to critical levels", disabling the anti-lock brakes and stability control or causing a loss of vehicle power.

24V-868 is one of four campaigns in the file about the unit that charges the 12-volt battery in Hyundai, Genesis and Kia electric cars. It is also the only report in this study that puts a time on the warning: the symptoms "may occur during 'fail-safe' mode driving over an approximate time period of 20-40 minutes." That figure is Hyundai's, for that fault in those cars. It does not transfer to an alternator failure in a gasoline car, and we do not use it that way.

19V-348 and 19V-805 are about connections, not batteries. Chrysler's report describes a loose battery ground that can cause an intermittent loss of power steering assist and a loss of motive power; Ford's describes a fastener on the F-150's power distribution box cable that can affect instrument displays, braking or steering assist, or stall the engine.

Finding 11: what the manufacturers said about warnings

The Part 573 report form has a field most readers never see: Identification of Any Warning that can Occur. It asks the manufacturer to identify any warning the driver might get before the defect shows itself. We read that field in every text-searchable report we could retrieve for the campaigns in this study.

We found a text-searchable Part 573 report for 193 of the 271 campaigns; the rest are older scanned letters or have no report on NHTSA's document server. Their answers to that field fall into four groups:

Only 23 of 193 reports name a warning on the dashboard. Nearly twice as many, 43, say plainly that the driver gets no warning at all, and 57 point only to signs a driver would have to notice with their eyes, nose or ears: smoke from under the hood, a burning smell, a flapping belt, a grinding starter, slow cranking. The classification rule is in the published reader script, and each report's own wording is in the downloadable table.

This is the manufacturers' own answer, not ours. It is the most direct statement in the public record of how much warning a given 12-volt defect gives.

Why the federal file sees so few frozen driveways

If you expected a winter spike, so did we. Here is why the file may not show one, and what that means for reading it. These are explanations, not measurements: none of them is tested by the data on this page.

NHTSA collects safety complaints, not service calls. The complaint form asks about a vehicle safety problem. A battery that is simply old and will not start the car on a cold morning is a maintenance event, and most owners who have one call roadside assistance or a parts store, not the federal government. The owners who do file tend to have a reason to think the car is defective: a new car, a battery that keeps dying, a failure on the road, or a recall they believe should cover them. That filter is exactly why the file is useful for this study, because it is weighted toward the failures that feel like safety problems. It is also why it cannot tell you how common winter no-starts are across all cars.

Heat and cold may do different things, and the file mixes them. A common explanation, which this file cannot test, is that heat wears a battery down and cold exposes the weakness, because a cold battery delivers less current at the moment the starter needs the most. If that is right, a battery damaged over a hot summer can fail in August or in January, and the complaint month tells you when the owner noticed, not when the damage was done. The on-the-road failures in this file lean toward summer; the no-starts dip in late spring and are otherwise nearly flat; neither shows a winter spike.

Incident dates are the owner's. The date of incident on the form is whatever the owner entered. Some round to the first of the month, some enter the date they noticed a pattern, some enter the date of the last failure in a long series. Across more than ten thousand complaints that noise mostly cancels out, which is why we show the month shares and not individual days.

The normalization matters. Complaints of every kind are not spread evenly over the year, so a raw count of battery complaints by month mostly tracks how many complaints NHTSA receives about everything. Every monthly figure on this page is therefore a rate: battery and charging complaints per 100 complaints with an incident in the same month.

The practical reading: the federal file does not tell you that your battery is safe until the first frost. It tells you that the failures owners report as safety problems are spread across the year, and that a real share of them happen with the car moving.

What this study does not show

It does not show how often batteries or alternators fail. Complaints are a self-selected sample of owners who chose to file. Nothing here is a failure rate, for any make, model or part.

It does not rank vehicles. Complaint counts follow how many cars were sold, how old they are, whether there was a recall that sent owners to NHTSA's site, and how motivated a given group of owners is. The make table is published because it shows two different patterns landing on different makes, not because a higher count means a worse car.

It does not show that a battery or alternator caused every failure described. The rules find complaints in which the owner describes a 12-volt battery or charging problem. In some of them the battery is a symptom of something else, a module that drains it, a belt that threw, a ground strap that corroded. The on-the-road group is defined by what the owner wrote happened, not by a diagnosis.

It does not measure time. Owners almost never write how long they drove between a warning and the car dying, and the handful who do cannot be generalized. Anyone who tells you that a charging light leaves you a fixed number of minutes is not quoting this file.

It does not use any data of ours. PulsCar has no fleet of cars whose batteries we have measured over a winter, and nothing on this page comes from our app or our users. Every number is from NHTSA's public files.

Some classifications are heuristics. The failure set, the on-the-road group and the warning groups are text rules, audited on samples, not codes in NHTSA's data. The electric and hybrid split uses model names, because the fuel-type field is mostly blank. The consequence classes for recalls are text rules on the consequence summary the manufacturer filed.

It is not verified data. Complaints are submitted by the public. NHTSA describes the database as a signal used with other data to identify possible safety problems, not as a record of confirmed defects. Recall populations are the manufacturer's estimate of vehicles potentially involved, and a vehicle can appear in more than one campaign.

Before winter: a checklist built from the data

The file cannot tell you when your battery will die. It can tell you which situations owners describe again and again, and that is what this list is built from. The three situations in the table below are composite scenarios, not individual customers: each is assembled from patterns that repeat across the complaints we read, and none describes a real person.

Test the battery at rest, not right after driving. A reading taken minutes after the engine stops still carries surface charge from the alternator and flatters the battery. Let the car sit for several hours, ideally overnight, then read the voltage across the terminals with everything off.

Test the charging voltage with the engine running. A battery can read fine at rest while the alternator is failing, and a healthy alternator reading can hide a battery that no longer holds a charge. The two readings answer two different questions, which is why they should never be judged against the same number.

Look at the terminals and the ground. Several of the largest recalls in this study are about cables, terminal covers and ground connections, not the battery itself. Corrosion, a loose clamp or a frayed cable can produce every symptom in the table above.

Check your VIN for open recalls. It takes a minute at nhtsa.gov/recalls, and the battery, alternator and starter recalls listed in this study are only the ones with those components in their NHTSA record. A recall filed under a module, a harness or software can still end in a dead battery.

If a new battery dies, do not just buy another. The largest single make in the failure set has only 2.1 percent of its complaints on the road; in the ones we read, the story is a battery that went flat while parked, again and again. A battery that keeps dying is often telling you about something else in the car.

If the battery light comes on while you are driving, treat it as a road problem, not a morning problem. One in six failure complaints in this file describe the car dying with the owner in it, and when the alternator is named it is one in four. The owners who wrote that the power steering went heavy or the lights went out did not have that failure in a driveway.

What PulsCar reads, and what it does not

A study like this is only honest if we are exact about our own product, so here is precisely what the PulsCar app does with the 12-volt system when it is paired with an OBDLink MX+ scanner, checked against the app's code on the day this page was published.

Voltage, labeled by engine state. The scanner reports the voltage it sees at the car's diagnostic port. The app reads it when it connects and about every ten seconds while it is recording a drive, and it labels each reading as engine off, battery at rest, or engine running, charging voltage. A reading taken when the app cannot tell whether the engine was running is not judged at all. In the app's health breakdown, a battery at rest counts as healthy at 12.4 volts or more, and a charging voltage counts as healthy between 13.2 and 15.0 volts. Those are the app's own thresholds, not a universal standard, and a reading at the diagnostic port can sit slightly below one taken across the battery terminals.

One alert for the road. If the voltage falls below 11.8 volts while the car is moving faster than 5 km/h, the app raises a full-screen alert headed Charging system failure, and a push notification unless you have switched critical notifications off. It shows the voltage it actually measured, not a stock number, and it fires once, then re-arms only after the voltage has recovered to 12.5 volts or more. This is the on-the-road failure from Finding 1, seen from the inside: by the time the voltage is that low while driving, the battery is carrying the car alone.

Stored trouble codes. The scanner reads the codes the engine computer has stored, including the system-voltage codes some cars set when the supply runs low or high.

What it does not do. It does not measure cranking voltage, the dip while the starter turns. It does not predict when a battery will fail, and it does not warn you weeks ahead. It does not know about a drain that happens while the car is parked and the app is not connected. It cannot see a failing alternator before the voltage shows it. The alert tells you that the failure has started so that you can get off the road, and it works only while the app is recording the drive. It does not prevent the failure.

What the free check does. The free check on our site listens to a 30-second recording of your engine and sends a report by email. It does not read voltage, and it cannot test a battery or an alternator. A whining alternator bearing or a squealing belt is a sound, and a sound is what the free check listens to; the Explorer owner quoted in Finding 1 reported a whine before the stall. It cannot confirm that the alternator is the cause. If the problem is electrical and silent, a battery and charging test at a parts store or a shop is the right first step.

Free, no card, no subscription. The report comes by email.

Reading a battery or alternator complaint like an investigator

Every complaint in this study is public, and anyone can read the ones about their own car.

Where to find them. NHTSA's recall and complaint lookup at nhtsa.gov lists complaints by make, model and year, with the owner's text, the date of the incident and flags for crashes, fires, injuries and tows. The flat files used here are free on NHTSA's static file server, and the rules on this page will run on them unchanged.

Do not filter by component alone. In this file, owners and call-center staff almost never file a battery or alternator problem under the battery or alternator component. Most go under the generic electrical system entry, and many under engine, power train or unknown. Search the text for battery, alternator and charging instead, and read the description.

Separate the three stories. A car that will not start in the morning, a car that dies on the road and a new battery that keeps going flat are three different problems with three different likely causes. Complaints that mix them are common, and a pattern only means something when the same story repeats.

Look for the recall number in the text. Owners and NHTSA's phone staff often write the campaign number into the complaint when the owner was told the car was covered, or was told it was not. Those complaints are the fastest route to finding out whether a remedy exists for your car.

How to write a useful complaint. Say whether the car would not start or died while moving, at what speed, whether a battery light or a charging message appeared and how long before, what else stopped working, and what the shop found. That last detail, stated plainly, is exactly what this study was built on.

Download the data

• Failure set funnel: from all complaints to on the road (CSV)
• Where it happened, by alternator mention, powertrain, model year and year received (CSV)
• Monthly rates by month of incident, all, on the road, no start, cold and hot states (CSV)
• Tows, crashes, fires and speeds by setting (CSV)
• Warnings, steering, brakes and lights in the on-the-road group (CSV)
• Voltages quoted by owners (CSV)
• Makes with at least 150 failure complaints (CSV)
• Vehicle age and mileage by setting (CSV)
• How NHTSA's component codes file these complaints (CSV)
• All 271 battery, alternator and starter recall campaigns with consequence class (CSV)
• Recalls by consequence (CSV) and by year the report was received (CSV)
• Part 573 reports: vehicles and model years as listed, safety risk and warning field (CSV)
• Exact rules, audits, file hashes and limits (TXT)
• Analysis script, Python standard library only (PY), Part 573 downloader (PY) and Part 573 reader (PY)

License. The tables, rules and scripts are published under CC BY 4.0. Please credit PulsCar analysis of NHTSA complaint and recall data and link to this page.

For reporters. Every figure can be re-run from NHTSA's public files with the script above; the method file lists the SHA-256 hash of each file we used, so you can tell whether you are running the same snapshot. NHTSA rebuilds its files every day, so a later download will contain newer complaints and give slightly different totals. If you need a cut we did not publish, a single make, a state, a model-year band, or the complaint numbers behind a table row, email [email protected] and we will send the counts and the rule used to produce them.

Related reading

For the problem that looks like this one from the driveway, see why a car battery keeps dying, which covers parasitic drains, and car won't start and car clicking when starting for the sound at the key. For the road, car stalls while driving. For the parts, alternator replacement, car battery replacement and starter replacement. Our earlier study of the same complaint files found that nearly three in four car noise complaints report no warning light.

Sources and documents

Primary sources only. Every complaint quoted is retrievable by its ODI number from NHTSA's public complaints API, and every recall report from NHTSA's document server.

1. NHTSA Office of Defects Investigation complaint flat files — COMPLAINTS_RECEIVED_2015-2019, COMPLAINTS_RECEIVED_2020-2024 and COMPLAINTS_RECEIVED_2025-2026, downloaded 23 September 2026; 1,105,443 rows, 780,071 distinct vehicle complaints by ODINO, received 1 January 2015 to 21 September 2026. File directory
2. ODI complaints data dictionary (CMPL.txt). Field 2 ODINO; field 6 YEARTXT, model year; field 8 FAILDATE, date of incident; field 12 COMPDESC, component description; field 14 STATE; field 17 LDATE, date received; field 18 MILES; field 20 CDESCR, description of the complaint; field 32 VEH_SPEED; field 46 PROD_TYPE, V for vehicle; field 49 VEHICLES_TOWED_YN. Data dictionary
3. NHTSA recall flat file FLAT_RCL_POST_2010 and its data dictionary (RCL.txt): field 2 CAMPNO, field 7 COMPNAME, field 11 RCLTYPECD, field 12 POTAFF, field 16 RCDATE, field 21 consequence summary. Recall file · Data dictionary
4. Part 573 Safety Recall Reports for the campaigns in Finding 10 and Finding 11, retrieved through NHTSA's safety-issues API and read with pdftotext -layout: 23V-734, 17V-418, 17V-435, 19V-888, 17V-114, 25V-019, 25V-636, 19V-348, 24V-868, 25V-644, 19V-805 and 25V-659 in the table; 24V-204, 24V-200 and 24V-867 in Finding 7; and every other campaign in the study with a text-searchable report, 193 of 271, listed with its document link in the published CSV. Example report, 25V-019 · Document lookup
5. 49 CFR 573.6, the information a manufacturer must put in a defect report: the vehicles by make, line, model year and dates of manufacture, paragraph (c)(2); the number of vehicles potentially containing the defect, paragraph (c)(3); and the risk to motor vehicle safety, paragraph (c)(5). Regulation text
6. Data.gov catalog entry for the ODI complaints dataset, describing complaint information as used with other data sources to identify safety issues that warrant investigation. Catalog entry
7. Quoted complaints — ODI 11623788, 11749425, 11731790, 11727324 and 11589238. Example query
8. PulsCar app thresholds — the 11.8-volt on-the-road alert, its 5 km/h speed floor and 12.5-volt re-arm, and the 12.4-volt and 13.2 to 15.0-volt health bands, as set in the app's code on 23 September 2026. These are our product's settings, listed so readers can check our claims against what we say elsewhere, not an external standard.
9. PulsCar study files — thirteen tables, the analysis script, the Part 573 downloader and reader, and the exact rules used on this page. Method and rules

How these numbers were built: one Python script over NHTSA's public complaint and recall flat files, run on 23 September 2026, and a second that reads the manufacturers' Part 573 reports; the sample, the four steps, the three settings, the warning rules, the audits and the limits are all described above and in the published method file. No failure rate, no model or brand ranking and no repair price is computed anywhere on this page.

Prices verified: not applicable, this study quotes no repair prices · Technical documents verified: 23 September 2026 · Data refreshed: 23 September 2026

Re-ran the rules and got a different number? Email [email protected] with your counts and the file hashes you used.