Were trains on time in France in September 2026?
In September 2026, 79.9% of the 90,228 trains with data in France ended with no reported delay of 5 min or more (the source only reports delays from 5 min), out of 94,043 trains tracked (96% with data). Figure closed on 21 Nov 2026: it no longer changes.
- No reported delay ≥ 5 min79.9 %
+0.8 pp vs August 2026 · 95 % CI: 79.7–80.2 %
- Trains with data90,228
of 94,043 tracked · coverage 95.9%
- Average delay3.0 min
last known delay
Source of these cards and rankings: train_runs and line_daily_stats for the month (last delay seen for each train, over the trains seen); the Wilson CIs are for that figure. The «over scheduled trains» table uses another base (arrival at destination over scheduled trains) and may give another figure.
Most punctual operators
at least 3,000 trains with a known delayThere is not enough data for this month yet.
Most punctual lines
at least 1,000 trains with a known delay| # | Line | On time |
|---|---|---|
| 1 | L4 🇫🇷 FR | 97.7 % 96.6–98.4 |
| 2 | P14 🇫🇷 FR | 90.5 % 88.9–91.9 |
| 3 | C1 🇫🇷 FR | 89.4 % 87.6–91.0 |
| 4 | C2 🇫🇷 FR | 87.6 % 85.7–89.2 |
| 5 | C6 🇫🇷 FR | 86.1 % 83.9–88.0 |
| 6 | C10 🇫🇷 FR | 85.4 % 83.2–87.4 |
| 7 | P10 🇫🇷 FR | 84.4 % 82.3–86.3 |
| 8 | C7 🇫🇷 FR | 82.2 % 79.9–84.2 |
| 9 | K16 🇫🇷 FR | 77.0 % 74.5–79.3 |
| 10 | C13 🇫🇷 FR | 74.1 % 72.2–75.9 |
| 11 | K4 🇫🇷 FR | 72.9 % 70.6–75.2 |
| 12 | C3 🇫🇷 FR | 60.4 % 58.4–62.4 |
By type of train (September 2026)
category assigned by TrainsRadar, not the commercial brand; with 1,000 or more trains with data| Type of train | With data | No delay ≥ 5 min | 95% CI |
|---|---|---|---|
| Regional | 90,228 | 79.9% | 79.7%–80.2% |
Our threshold and the official one
This country’s source only reports delays of 5 min or more: a train with no notice counts as “no delay reported”, which is not the same as “on time”. That is why its figure is not compared with other countries at 3 min.
We do not yet have enough arrivals at destination for this country to compute punctuality at 5, 6 and 15 minutes.
Official thresholds by country
SNCF TER (régularité mensuelle)
Threshold: < 5 min
Our column with the same threshold: < 5 min at destination.
- Their page does not say how cancellations are treated.
la régularité est calculée lors de l’arrivée du train à la dernière gare de son parcours (terminus)
Official source · checked on 21 November 2026SNCF TGV (régularité mensuelle AQST)
Threshold: < 5 min (< 1h30), < 10 min (1h30–3h), < 15 min (> 3h)
No column in this report uses exactly this threshold.
- Their page does not say how cancellations are treated.
considéré à l’heure si son retard au terminus est inférieur à 5min pour un parcours inférieur à 1h30
Official source · checked on 21 November 2026
Our figure also only includes trains that appear in open data and whose arrival we saw, not every train the operator runs. How we measure
How to cite this report
How to cite this report
In a news story
Source: TrainsRadar, punctuality report for September 2026 (https://trainsradar.com/reports/2026-09)
In academic work
TrainsRadar team (2026). Monthly rail punctuality report: September 2026 [data set]. TrainsRadar. https://trainsradar.com/reports/2026-09
Reusing the data
This report combines sources whose licence we have not confirmed yet, so it carries no open licence: you may quote its figures citing the source.
The licence for reusing the press version is still to be confirmed. Meanwhile, cite the source; to reuse the data, write to us.
Press version
Only countries whose sources allow commercial use, one file per licence:
How it is calculated
- This country’s source only reports delays of 5 min or more: a train with no notice counts as “no delay reported”, which is not the same as “on time”. That is why its figure is not compared with other countries at 3 min.
- We give a month’s figure when we know the delay of at least 3,000 trains and 30% of the trains tracked. Below that, no figure is published and the page is not offered to search engines.
- These figures do not say why trains are late: TrainsRadar does not attribute causes (strikes, storms, engineering works) that the data does not show.
- Full methodology