The short answer: 24 months. FMCSA's Safety Measurement System scores you on the previous 24 months of on-road data, and events older than that are, in FMCSA's own words, no longer used to assess a carrier's safety posture. But a violation does not sit at full strength for two years and then vanish. It fades in three steps: time weight 3 while it is within 6 months of the SMS run date, weight 2 from 6 to 12 months, weight 1 from 12 to 24 months, then zero. Nothing you do, pay, or file makes that clock run faster.
Below: when points actually fall off, why the score sometimes moves the wrong way while you wait, and which of the things people try change anything. For the system as a whole (all seven BASICs, how percentiles are built), start with our hub guide, CSA Scores Explained.
Twenty-four months, counted from the inspection date: not from the day the paperwork reached you, not from the day you paid a ticket, not from the day the state uploaded it. Two clocks get confused with this one:
Side by side, the clocks people mix up when they ask how long something "stays on the record":
| Record | Whose record | How long it counts | Counted from | Where to check |
|---|---|---|---|---|
| SMS BASIC measure (the CSA score) | Carrier | 24 months, weight 3 then 2 then 1 | Inspection date | ai.fmcsa.dot.gov/SMS, log in for percentiles |
| Percentile in an inspection-based BASIC | Carrier | Comes off after 12 months without a violation plus a clean latest inspection | Last violation in that BASIC | SMS, logged in |
| Acute or Critical Violation from an investigation | Carrier | 12 months | Date the violation was cited | SMS, Investigation Results tab |
| PSP report | Driver | 3 years inspections, 5 years crashes | Inspection or crash date | psp.fmcsa.dot.gov, $10 per report |
| DataQs filing window | Carrier or driver | 3 years inspections, 5 years crashes | Inspection or crash date | dataqs.fmcsa.dot.gov, free |
| CPDP crash preventability review | Carrier | Crashes up to 5 years old | Crash date | DataQs, about 90 days to process |
SMS does not treat last month's violation and last year's as equals. Every applicable violation, every crash, and every relevant inspection carries a time weight based on its age relative to the SMS run date:
| Age on the SMS run date | Time weight | Share of original pull | What it feels like |
|---|---|---|---|
| 0 to 6 months | 3 | 100% | Maximum damage. One fresh violation equals three old ones. |
| Over 6 and up to 12 months | 2 | 67% | First real relief, usually the month a stubborn BASIC starts moving. |
| Over 12 and up to 24 months | 1 | 33% | Background noise, but counted for a full extra year. |
| Over 24 months | 0 | 0% | Out of the calculation entirely. |
On top of that sits a severity weight from 1 to 10 from Appendix A of the SMS Methodology, plus an additional 2 in most BASICs if the violation triggered an out-of-service order (Unsafe Driving carries no OOS weight). Severity weights from one inspection in one BASIC are capped at 30, and the cap applies before the time weight multiplies in: one catastrophic inspection has a ceiling, a steady drip of small violations does not.
Relief therefore arrives as two step changes, not a smooth slope. Guessing when a BASIC drops back under its intervention threshold (65% for Unsafe Driving, Crash Indicator and HOS Compliance, 80% for Vehicle Maintenance, Driver Fitness and Controlled Substances at the general threshold)? Circle the 6 and 12-month anniversaries of your worst inspections. Full table: what counts as a good CSA score by BASIC.
You count 24 months, the date arrives, you open SMS, and the violation is still there. Nothing is broken. SMS is a monthly system with a publication delay built in: FMCSA takes a data snapshot on the third or last Friday of each month, then needs roughly 10 days to process and publish it, and your violation leaves the calculation on the first snapshot taken more than 24 months after the inspection date. Chained together: expect the record to disappear about ten days to six weeks after the 24-month anniversary, depending on where the anniversary falls relative to the monthly snapshot. An anniversary a day or two before a snapshot Friday clears on that snapshot; one that lands just after a snapshot waits a full cycle. FMCSA publishes the schedule as tentative with approximate release dates, so read the two lines on the SMS front page rather than counting days: Data current as of tells you the snapshot date, Next update tells you when the next one lands. Note that a freshly published snapshot is already two to three weeks behind, and it keeps ageing until the next release, so late in the cycle you may be looking at data five or six weeks old.
| Inspection date | Weight 3 until | Weight 2 until | Weight 1 until | First snapshot without it |
|---|---|---|---|---|
| 15 January 2026 | 15 July 2026 | 15 January 2027 | 15 January 2028 | Late Jan 2028 snapshot, on the site in February 2028 |
| 20 October 2026 | 20 April 2027 | 20 October 2027 | 20 October 2028 | Late Oct 2028 snapshot, on the site in November 2028 |
You do not have to guess the cycle. FMCSA lists the snapshot Fridays and the approximate release weeks in the SMS Help Center, and calls the schedule tentative:
| Release month | Data snapshot date (Friday) | Approximate release |
|---|---|---|
| June 2026 (early) | 15 May 2026 | Week of 25 May 2026 |
| July 2026 | 26 June 2026 | Week of 6 July 2026 |
| August 2026 | 31 July 2026 | Week of 10 August 2026 |
| September 2026 | 28 August 2026 | Week of 7 September 2026 |
| October 2026 | 25 September 2026 | Week of 5 October 2026 |
| November 2026 | 30 October 2026 | Week of 9 November 2026 |
Three lines on the SMS front page tell you where you stand: Data current as of, Website Updated on and Next update. Read on 2 September 2026 they said 31 July 2026, 10 August 2026 and week of 7 September 2026 respectively. Treat those three values as an illustration of the lag, not as a permanent state: the September release lands in the week of 7 September 2026 and rewrites all three lines, so always read them live. Read them together and the lag is obvious: the numbers on screen were assembled from data that stopped moving five weeks earlier. A violation whose 24-month anniversary fell in mid-August 2026 would therefore not vanish on the August release at all: it leaves on the 28 August snapshot, visible only in the September release.
The answer surprises everyone: clean inspections age out on exactly the same 3-2-1 schedule as dirty ones.
In the driver BASICs (HOS Compliance, Driver Fitness, Controlled Substances/Alcohol) and the vehicle BASICs (Vehicle Maintenance, HM Compliance) the denominator is your total time-weighted relevant inspections, and FMCSA counts as relevant any qualifying inspection including those that produced no violation in that BASIC. Your violation-free inspections are what holds the measure down, and they decay 3, then 2, then 1, then nothing. A fleet inspected heavily 18 to 24 months ago that has barely crossed a scale since watches those clean inspections shrink out of the denominator while an in-window violation still counts. Nothing happened on the road, the measure rose anyway.
Unsafe Driving and Crash Indicator have their own trap: they do not divide by inspections at all, but by Average Power Units multiplied by a Utilization Factor, where Average PUs is the mean of your fleet size now, six months ago and eighteen months ago. Sell two trucks or file an MCS-150 with a lower power unit count and the denominator drops while old violations still count. Which also means clean inspections do not help you at all in these two: the only levers are time and fleet size.
Third reason a number drifts with no event behind it: the percentile is relative. SMS groups you with carriers that have a similar number of inspections, violations or crashes, ranks you 0 to 100 inside that group, and re-sorts everybody every snapshot. Your measure can hold still while the percentile moves.
Ageing violations are only half the story. The SMS also applies data sufficiency standards: below a minimum number of events you get no percentile in a BASIC at all, and there are explicit rules that pull you out of the ranking even when old violations are still inside the 24-month window. The rules below come from the SMS Methodology (Version 3.21, revised June 2026). Read the middle column as a two-part test: the count is of all relevant inspections, clean ones included, and on top of that at least one of them has to carry a violation in that BASIC. Five relevant inspections with a violation in only one of them is enough to be ranked in Vehicle Maintenance.
| BASIC | Minimum before you are ranked at all | You drop out of the ranking when |
|---|---|---|
| Unsafe Driving | 3 or more inspections with a violation in this BASIC | Every violation in the BASIC is older than 12 months. Remaining carriers retain the previously calculated percentile. |
| Crash Indicator | 2 or more applicable crashes | No crash was recorded in the previous 12 months. |
| HOS Compliance | 3 relevant driver inspections in total, clean ones included, plus at least one inspection with a violation in this BASIC | No violation in the BASIC during the previous 12 months and no violation in the BASIC on the latest relevant inspection. |
| Vehicle Maintenance | 5 relevant vehicle inspections in total, clean ones included, plus at least one inspection with a violation in this BASIC | Same two-part test as HOS Compliance. |
| Driver Fitness | 5 relevant driver inspections in total, clean ones included, plus at least one inspection with a violation in this BASIC | Same two-part test as HOS Compliance. |
| HM Compliance | 5 relevant inspections in total, clean ones included, plus at least one inspection with a violation in this BASIC | Same two-part test as HOS Compliance. |
| Controlled Substances/Alcohol | 1 inspection with a violation (safety event group 1 is a single inspection) | Every violation in the BASIC is older than 12 months. |
Two practical consequences. First, in the inspection-based BASICs the fastest visible relief is not month 24 but 12 clean months plus one clean inspection: satisfy both halves of that test and the percentile comes off your profile long before the underlying violation expires. Second, in Controlled Substances/Alcohol a single inspection with a violation is enough to put you in a ranked group, which is why that BASIC lights up for carriers who consider themselves clean.
Investigation findings run on their own 12-month clock. Acute and Critical Violations, defined in the Safety Fitness Procedures at 49 CFR Part 385, Appendix B, prioritise the matching BASIC when found during an investigation in the previous 12 months from the snapshot date, whatever your percentile says.
Open any driver forum and someone states with confidence that CSA points "drop after a year". It is wrong, but it is wrong for three understandable reasons, and all three are real 12-month rules inside the same system:
None of the three deletes the roadside violation. It keeps feeding the measure until the first snapshot taken more than 24 months after the inspection date, and it stays on the driver's PSP report for 3 years. When a consultant promises removal "after 12 months", ask which of the three rules they mean, because the answer is never the fourth one.
No. No expedited removal, no appeal that shortens the window, no fee that clears the data: the 24-month window and the 3-2-1 weights are arithmetic applied identically to every carrier on the same monthly cycle. Anyone selling guaranteed CSA violation removal is selling something nobody can deliver: the DataQs filing itself is free and open to you, the outcome belongs to the state reviewer, and no filing shortens the 24-month window. What genuinely changes a record is narrow:
| Situation | Removes it? | Mechanism and deadline |
|---|---|---|
| Data is factually wrong (wrong carrier, wrong driver, wrong CDL, duplicate, wrong cite) | Yes, if proven | Request for Data Review in DataQs. States must review inspection requests filed within 3 years of the inspection. |
| Court dismissed the citation or found the driver not guilty | Yes | RDR with certified documentation of the judicial result. Applies to inspections on or after 23 August 2014. |
| Convicted of a different charge | Partially | Severity weight set to 1, out-of-service weight dropped. The violation stays, it just weighs almost nothing. |
| Crash was not preventable | Excluded from the measure | CPDP: flagged in SMS as Not Preventable and excluded from the Crash Indicator measure and percentile, though still listed on the site. CPDP covers only the 21 listed crash types, mostly cases where someone hit the truck. Crashes on or after 1 December 2024 follow the current eligibility list, earlier ones the previous standards, and FMCSA cannot review crashes older than 5 years. FMCSA warns that, because of submission volume, CPDP requests are averaging about 90 days to process. |
| You paid the fine | No | No effect on SMS whatsoever. |
DataQs itself is about to get faster. FMCSA published its final notice on 16 April 2026 and set the new requirements to take effect 150 days later, which lands on 13 September 2026, alongside the DataQs system release that supports them. From that date, states that take MCSAP grant funding owe you a three-stage review on the clock:
| Stage | Deadline for the state | Deadline for you |
|---|---|---|
| Opening the RDR | 7 days from submission | File within 3 years of an inspection, 5 years of a crash |
| Initial Review | Decision within 21 days of submission | Respond to a request for more information within 14 days |
| Reconsideration Review | Decision within 21 days of the request | Request it within 30 days of the previous decision |
| Final Review | Decision within 45 days of the request | Request it within 30 days of the previous decision |
Two details worth knowing before you file. If the state asks for additional information, its clock stops and the 14 days you take do not count against the 21 or 45-day deadline, so answer fast and the review stays inside its window. And a request closed as Closed, No Data Correction Made must now carry a written explanation: the decision-maker by name and title, the list of evidence reviewed, the decision, the specific reasons, and the next steps including how to appeal. Procedural closures such as Closed, No Requestor Response are exempt from that detail, which is one more reason not to miss the 14-day window.
Not a rare tool, and not a crowded one either. FMCSA publishes roughly 3 million inspections, 5 million violations and over 180,000 vehicles in federally reportable crashes every year; in 2024 DataQs received 63,548 requests concerning inspections and violations and 8,314 on crash data, about 71,900 in total. If your inspection report is wrong, you are competing for attention with a small fraction of the record. Our step-by-step filing guide: how to dispute a violation through DataQs.
Repeated in every driver chat, and backwards. A roadside violation and a state citation are two records in two jurisdictions. FMCSA is explicit: violations written on the inspection form during an enforcement stop do not always produce a ticket, but they are used in SMS whether or not a citation is issued. The violation lands in your BASIC because the officer wrote it on the report. Paying or not paying never decides whether it counts.
The court result only decides whether you have grounds to remove a record that already exists, and only two outcomes help: dismissed or not guilty removes it, conviction of a different charge knocks the severity weight down to 1. Pleading guilty, paying the fine, or any disposition ending in a conviction on the original charge leaves the violation where it is, at full weight, for the full 24 months. The tracks are also independent: winning in state court does not update federal data by itself, you still file the RDR with certified documentation.
What you do in the first 30 days after a bad inspection matters far more than anything available in month 20. If the violation is defensible, defend it. For what specific hours-of-service violations are worth, see HOS violations and what they cost in CSA points.
FMCSA put it this way in a March 2012 briefing on CSA: carriers improve their BASIC percentile ranks by demonstrating improved roadside performance, either through reduced crashes or clean roadside inspections. The wording is old, the mechanics are current. The methodology adds the two mechanical routes out described above: violations lose weight and then leave the 24-month window, and the data sufficiency rules pull you out of the ranking after 12 months without a violation in that BASIC plus a clean latest inspection. In practice:
We read your SMS results after every monthly snapshot, flag which violations hit their 6, 12 and 24-month marks next, file DataQs Requests for Data Review on the ones that are wrong, and tell you what to fix before the next inspection. Safety Compliance plans from $99/mo (BASIC), $189/mo (START), $499/mo (PREMIUM). Rehearsal first? Mock DOT Audit, $399. Starting from scratch? Authority Bundle, $499. Russian and English.
Ask us which plan fits your fleet →Or just call: (315) 871-0833 · WhatsApp
24 months from the inspection date, with three steps on the way out. A violation carries time weight 3 for the first six months, 2 from six to 12 months, 1 from 12 to 24 months, then zero. FMCSA states that safety events older than 24 months are no longer used to assess a carrier. The inspection report stays in the federal database and on the driver's PSP record, it simply stops moving your BASIC measure.
It depends whose record you mean. For the carrier, 24 months in the SMS. For the driver, the PSP report holds the most recent 3 years of roadside inspections and 5 years of crashes, so a violation follows a driver through hiring screens for a year longer than it affects the company. FMCSA does not assign scores to individual drivers: the PSP report lists the events without a score.
FMCSA takes the SMS snapshot on the third or last Friday of each month and publishes about 10 days later. The dates the agency lists for 2026 are Fridays 15 May, 26 June, 31 July, 28 August, 25 September and 30 October, each published the following week. FMCSA calls the schedule tentative, so confirm on the SMS front page: read on 2 September 2026 it showed data current as of 31 July 2026 and a next update in the week of 7 September 2026.
Once per monthly snapshot. Between snapshots your BASIC measures and percentiles are frozen, even if you were inspected yesterday. Percentiles can move without any new event on your side, because ageing inspections change your denominator and the whole safety event group is re-ranked each month. Judge the trend over three or four snapshots rather than month to month.
If the officer recorded the violation on the inspection report, yes. FMCSA uses roadside violations in the SMS whether or not a citation is issued, so a warning given instead of a ticket does not keep the violation out. If nothing was written on the report, there is nothing to score. Do not confuse this with an FMCSA warning letter, which is an intervention triggered when a BASIC crosses its intervention threshold.
Yes, on their own, at 24 months. A violation counts at time weight 3 for the first six months, 2 from six to 12 months, 1 from 12 to 24 months, then it stops feeding the BASIC measure on the next monthly snapshot. Nothing you file, pay or negotiate shortens that. The inspection report itself does not go away: it stays in FMCSA records and on the driver's PSP report long after the points stop counting.
Two clocks, two answers. For the carrier, the inspection feeds the SMS measures for 24 months from the inspection date. For the driver, the PSP report carries roadside inspections for 3 years and crashes for 5 years, and a PSP report costs $10. FMCSA keeps the underlying record beyond both. So an inspection can be gone from your CSA percentile and still show up when a prospective employer pulls the driver's PSP.
There is no reset button. FMCSA does not wipe a carrier's SMS history on request, and a fresh USDOT number taken to escape an old record is a reincarnation issue FMCSA investigates. Three things genuinely change the picture: the 24-month clock running out, a successful DataQs Request for Data Review, and a Not Preventable determination on a crash. A percentile also comes off your profile after 12 months with no violation in that BASIC plus a clean latest relevant inspection.
24 months in the Crash Indicator BASIC, on the same 3-2-1 time weights as violations. A crash stays 5 years on a driver's PSP report, and states must accept crash Requests for Data Review filed within 5 years of the crash. A crash found Not Preventable under the Crash Preventability Determination Program is excluded from the Crash Indicator measure and percentile, though it stays listed on the SMS website.
24 months from the inspection date. The SMS scores a carrier on the previous 24 months of on-road data, and FMCSA states that safety events older than 24 months are no longer used to assess a carrier's safety posture. The inspection report itself does not disappear: it stays in FMCSA records and on a driver's PSP report, it just stops feeding your BASIC measures.
On the first monthly SMS snapshot taken more than 24 months after the inspection date. FMCSA takes the snapshot on the third or last Friday of each month and needs about 10 days to process and publish it, so expect the violation to vanish about ten days to six weeks after the anniversary, depending on where that date falls relative to the snapshot. FMCSA calls the release schedule tentative, so go by the Next update line on the SMS front page rather than a fixed day of the month.
No, they only get lighter. Time weight is 3 for violations recorded in the past six months, 2 for those over six and up to 12 months old, and 1 for everything older than 12 months but within the past 24. At the 12-month mark a violation drops from weight 2 to weight 1, a third of what it carried when it was fresh. It leaves the calculation entirely only at 24 months.
No. Nothing shortens the 24-month window: the weights are arithmetic applied identically to every carrier on the same monthly cycle. The only lawful removal is a successful Request for Data Review in DataQs, and only when the data is factually wrong or a court dismissed the citation or found the driver not guilty. Filing in DataQs is free and open to you directly, and the decision belongs to the state reviewer.
Clean inspections age out on the same 3-2-1 schedule. In HOS Compliance, Vehicle Maintenance, Driver Fitness, Controlled Substances/Alcohol and HM Compliance the denominator is your time-weighted relevant inspections, which include inspections that produced no violation in that BASIC, so when old clean inspections drop out the measure rises by itself. Unsafe Driving and Crash Indicator divide by Average Power Units multiplied by a Utilization Factor instead, so a shrinking fleet raises the measure. Percentiles also move as the peer group is re-ranked every snapshot.
No, that is backwards. FMCSA states that violations reported on the inspection form during an enforcement stop do not always result in a citation, but are used in the SMS whether or not a citation is issued. The record exists from the inspection date regardless of the ticket. Only a dismissal or a not guilty finding removes it, filed through DataQs with certified documentation of the judicial result; conviction of a different charge sets the severity weight to 1.
On the PSP report a driver carries the most recent 3 years of roadside inspection data and the most recent 5 years of crash data, longer than the carrier's 24-month SMS window. A single PSP record costs $10 at psp.fmcsa.dot.gov, and free PSP Monitoring emails the driver every time the record changes. The PSP report contains no score, just the underlying inspections and crashes.
Once a month. FMCSA takes a data snapshot on the third or last Friday of each month and publishes the recalculated results about 10 days later. Nothing changes in between, so checking daily is pointless. Read on 2 September 2026, the SMS front page showed data current as of 31 July 2026, the website updated on 10 August 2026 and the next update in the week of 7 September 2026. That was the August release still on screen; the September one moves all three lines, so check them live rather than trusting a printed date.