How Amtrak On-Time Performance Is Measured: The 2020 Rule, the 80% Standard and the Delay Split
The FRA counts passengers, not trains, within 15 minutes at every station. Why customer OTP replaced endpoint OTP and what the host-railroad delay split means.
In the quarter from January 1 to March 31, 2026 (the FRA's FY26 Q2), 13 of the 47 Amtrak routes the Federal Railroad Administration reported met the federal on-time standard of 80 percent. None of the 14 long-distance routes did. The figures come from the FRA's quarterly filing under 49 CFR part 273, the 2020 rule that changed what "on time" means for a passenger train.
This guide explains that rule: how customer on-time performance is measured, why it replaced Amtrak's older endpoint measure, and what the split of delay minutes between host railroads, Amtrak and third parties tells you about a route. The route-by-route record is in the Amtrak On-Time Performance by Route pages.
The measure is customers, not trains
The rule defines customer on-time performance as the percentage of all customers on an intercity passenger rail train who arrive at their detraining point no later than 15 minutes after their published scheduled arrival time (49 CFR 273.5(a)(1)). Three things in that sentence do the work.
It counts people. A train carrying 300 passengers that arrives 20 minutes late at a busy station lowers the route's figure far more than a train carrying 30. Amtrak supplies the FRA with ridership by train, by day and by station for exactly this reason (the rule calls it ridership data, 49 CFR 273.3).
It counts every station, not just the last one. A customer is on time or late at the station where they got off. A train that runs 40 minutes late through the middle of its route but makes up the time by the terminal still records every passenger who got off in the middle as late.
The threshold is 15 minutes at every station on every route. From October 1, 2025 that includes the Acela, for which Amtrak had used a 10-minute maximum before FY2026 (FRA FY26 Q2 report, "New 15-minute maximum for Acela OTP"). It is the same 15-minute line the U.S. DOT uses for airline on-time reporting, so an Amtrak route figure and an airline route figure on this site mean the same thing.
Why it replaced endpoint on-time performance
Before the 2020 rule, the figure Amtrak published was endpoint on-time performance: a train was on time if it reached its final station within a tolerance that grew with route length, up to 30 minutes for the longest routes. That measure had two problems the rule set out to fix. It said nothing about the stations in between, where most passengers on a long route actually get off, and it treated a sold-out train and an empty one the same.
The Passenger Rail Investment and Improvement Act of 2008 told the FRA and Amtrak to define metrics and minimum standards jointly. An earlier version, issued in 2010, was vacated in litigation; the version in force was published on November 16, 2020 and took effect in 2021 (49 CFR part 273). The FRA's first quarterly report with route-level customer OTP for every route covers October to December 2021 (FY22 Q1), which is where the series on this site starts.
Because the two measures differ in threshold, weighting and the stations they count, an on-time figure Amtrak published for 2019 is not comparable to a customer OTP figure from 2024. This site shows only the FRA's customer measure.
The 80 percent standard
The minimum standard is 80 percent customer on-time performance for any two consecutive calendar quarters (49 CFR 273.5(a)(2)). Two details follow from the wording.
A single quarter at or above 80 percent meets the metric for that quarter, but the standard is judged over two quarters together. The route pages mark both: whether each quarter met 80 percent, and whether the last two reported quarters both did. In FY26 Q2, 13 routes met the quarterly metric and 10 of those had also met it the quarter before.
Missing the standard does not fine anyone by itself. It gives Amtrak standing to ask the Surface Transportation Board to investigate whether a host railroad is denying passenger trains the preference federal law requires. That is why the rule also requires the delay split below: the standard only matters with the reasons attached.
Worked example: the Keystone, FY25 Q3 to FY26 Q2
The Keystone (Harrisburg to Philadelphia and New York) is a state-supported route that runs almost entirely on track Amtrak owns. Its last four reported quarters, from the FRA's Customer OTP Metrics files:
| Quarter | Period | Customer OTP | 80% metric |
|---|---|---|---|
| FY25 Q3 | Apr 1 – Jun 30, 2025 | 89% | met |
| FY25 Q4 | Jul 1 – Sep 30, 2025 | 92% | met |
| FY26 Q1 | Oct 1 – Dec 31, 2025 | 90% | met |
| FY26 Q2 | Jan 1 – Mar 31, 2026 | 86% | met |
Four consecutive quarters at or above 80 percent, so the two-quarter standard was met in each of the last three. In FY26 Q2 its delay minutes per 10,000 train miles were 620 charged to Amtrak, 340 to host railroads and 316 to third parties. On a route where Amtrak is its own host for most of the distance, the Amtrak-responsible share leading is what you would expect, and the figure is small either way.
Compare the Empire Builder (Chicago to Seattle and Portland) in the same quarter: 57 percent of customers on time, with 849 host-responsible delay minutes per 10,000 train miles against 387 Amtrak-responsible and 429 third party. Of its 400,756 train miles in the quarter, 336,222 were on BNSF Railway track. The route missed the metric in every one of the 18 quarters the FRA has reported, and the host line is where most of its delay minutes sit.
What the delay split means for a traveller
Outside the Northeast Corridor, Amtrak owns almost none of the track it runs on. The freight railroad that owns a line also dispatches it, deciding which train waits for which. The rule therefore requires every delay minute to be coded to a responsible party (49 CFR 273.3):
- Host-responsible: freight train interference, slow orders, signals, routing, maintenance of way, commuter and other passenger train interference, catenary or wayside power failure, detours.
- Amtrak-responsible: passenger-related delays at stations, equipment failures, holding for connections, injuries, initial terminal delays, servicing, crew and system delays.
- Third party: bridge, debris and vehicle strikes, customs, drawbridge openings, police activity, trespassers, utility delays, weather (heat and cold orders, storms, floods, slippery rail), acts of God.
The minutes are reported per 10,000 train miles so a 2,200-mile route can be read beside a 100-mile one. In FY26 Q2, host railroads accounted for at least half of all coded delay minutes on 27 of the 47 reported routes. On the five Northeast Corridor sub-services, where Amtrak is the host for most of the distance, the host share was between 18 and 43 percent.
For a traveller the split answers a practical question: is this route late because of something that could change next quarter, or because of the track it runs on? Weather and trespasser delays move with the season. Equipment and crew delays move with Amtrak's fleet and staffing. Freight interference on a single-track line moves only when the host railroad changes how it dispatches, which is the slowest of the three to change. A route whose delay minutes are mostly host-responsible has been late for the same reason for years, and the quarterly series on its page usually shows it.
Two cautions. The coding is Amtrak's, made under Amtrak procedures, and host railroads dispute some of it; the FRA reports the disputed minutes separately and this site does not adjust for them. And the per-10,000-mile rate excludes delay minutes on hosts for which the FRA files no train miles (Canadian territory on the Maple Leaf and Adirondack, some terminal railroads), though those minutes still count in a route's totals.
Reading a route page
Each page in the Amtrak On-Time Performance by Route section shows the latest quarter's customer OTP with a met-or-missed mark against 80 percent, whether the last two quarters together met the standard, the full quarterly series as bars with the standard drawn as a line, and the three-way delay split. Every figure names the FRA quarter it came from. These are the federal record of past quarters, not a forecast for a trip, and not live train status.
Informational only, not professional advice.
Sources
- Federal Railroad Administration, Intercity Passenger Rail Service Quality and Performance Reports (quarterly reports and data files, FY21 Q4 onward): https://railroads.dot.gov/rail-network-development/passenger-rail/amtrak/intercity-passenger-rail-service-quality-and
- 49 CFR 273.5, Metrics and minimum standards (customer on-time performance, train delays, delays per 10,000 train miles): https://www.law.cornell.edu/cfr/text/49/273.5
- 49 CFR 273.3, Definitions (Amtrak-responsible, host-responsible and third party delays): https://www.law.cornell.edu/cfr/text/49/273.3
- Metrics and Minimum Standards for Intercity Passenger Rail Service, final rule, 85 FR 72971 (November 16, 2020): https://www.federalregister.gov/documents/2020/11/16/2020-25212/metrics-and-minimum-standards-for-intercity-passenger-rail-service
Sources
- Federal Railroad Administration, Intercity Passenger Rail Service Quality and Performance Reports (quarterly, 49 CFR part 273)
- 49 CFR 273.5 — Metrics and minimum standards
- 49 CFR 273.3 — Definitions
- Metrics and Minimum Standards for Intercity Passenger Rail Service, final rule, 85 FR 72971 (2020-11-16)
Informational only — every figure here comes from historical federal filings, not a prediction of any future flight, and nothing here is professional travel advice.