GTFS Scorecard
← All agencies

Feed status · checked 2026-10-16

Trains régionaux ZOU ! — Transdev Rail Sud

Based on the official feed source on file

Service mode Rail

65.9 / 100

down 4.3 since 2026-10-15

Cataloged in Provence-Alpes-Côte-d’Azur, France.

Covers 189 daysAccessibility gaps

Source feed: https://www.data.gouv.fr/api/1/datasets/r/764d6d5b-d04e-4aa7-94b7-f8b2274d2964

A data-quality and completeness lens to help an agency improve its GTFS feed. Not an official compliance determination from any transit program. New to this? How to read your scorecard. Interactive view of this scorecard. Rubric v1.4, validator 8.0.1.

Checked for changes 37 minutes ago; last changed 10 days ago.

Measured 4 of 4 score categories from the official feed URL on file.

How we measured this

Confidence in this measurement: high.

  • All four score categories were measured this run.
  • Realtime was sampled in one bounded window of 3 snapshots.
  • The feed was downloaded from the official feed URL on file.

Confidence describes how much the pipeline could measure this run, not the feed itself. It never changes the grade.

About this feed

Trains régionaux ZOU ! — Transdev Rail Sud’s published feed lists 1 rail route, 5,486 scheduled trips and 12 stops. Its service dates run from 2026-10-04 through 2027-04-23. It publishes realtime trip updates; when we checked, it answered. The published file last changed on 2026-10-06.

Top things to fix

Fix 01

Review the rule documentation for 'foreign_key_violation' at https://gtfs-validator.mobilitydata.org/rules.html and check the flagged rows in your feed.

Foreign key violation (flagged by the MobilityData validator). See the linked rule for what this affects.

This is a validator notice, and the validator decides what each instance covers, so no share of the network can be read from it here.

⏱ Varies.worth about +12 points in its category

Fix 02

Ask your AVL vendor to send updates at least every 20 seconds. That is the Caltrans guideline.

Realtime data was up to 3055 seconds old when sampled. Stale positions and predictions are worse than none: riders watch a transit vehicle that already left.

This was measured over a realtime sampling window, so it has no feed-wide share.

⏱ A vendor configuration question.worth about +29 points in its category

Fix 03

Set wheelchair_boarding to 1 (accessible) or 2 (not accessible) for every stop. A field survey can start with the busiest stops.Likely your team

12 of 12 stops don't say whether a wheelchair user can board there. Riders who use wheelchairs can't plan a trip when accessibility is marked 'unknown'; apps show no information at all.

Fixing this covers all 12 stops in the feed.

⏱ A column in stops.txt; your scheduling software likely has it.worth about +25 points in its category

Riders and need behind these fixes

  • Rider-trip figures come from the United States National Transit Database, which does not cover this feed's country.
  • The served-area need overlays cover the United States and Canada, so none of them applies to this feed.

These describe this feed only. They do not rank agencies, and they do not change the grade or the order of the fixes.

Finding handoff

Move one finding to a recheck

Select one finding. Copy the request, make the change in the feed-producing tool, then compare the next complete run.

Finding
Fix 01 in Top things to fix: its feed evidence and the requested change.
Recheck
Publish the changed feed at the same URL. On the next complete, comparable scorecard run, confirm that this finding is no longer reported.
Copy handoff text
How to make and check these changes

Check the result on the published feed. Read the guide, make the change in your tool, and use the next comparable run to see whether the finding is still reported. Only an action or ticket record can attribute who made the change.

  1. Fix 01

    Make the change. Make this change in whatever tool produces your feed, then re-export.

    Check the result. The next scorecard run checks this finding again. If a comparable check no longer reports it, the feed's clearance log records that result. This confirms feed state, not who made the change. See this feed's finding-clearance log.

  2. Fix 02

    Make the change. Make this change in whatever tool produces your feed, then re-export.

    Check the result. The next scorecard run checks this finding again. If a comparable check no longer reports it, the feed's clearance log records that result. This confirms feed state, not who made the change. See this feed's finding-clearance log.

  3. Fix 03 · Read the fix guide for scorecard_wheelchair_boarding_unknown

    Make the change. Make this change in whatever tool produces your feed, then re-export.

    Check the result. The next scorecard run checks this finding again. If a comparable check no longer reports it, the feed's clearance log records that result. This confirms feed state, not who made the change. See this feed's finding-clearance log.

Rider view: what this feed publishes

A quick read of rider-facing information in this feed.

Schedule visibility
The feed's last published service date is in 189 days.
Published accessibility data
Accessibility information is stated for 0% of stops and 100% of trips. This measures published data, not whether stops or vehicles are physically usable.
Fare information
No fare information is published in the feed.
Realtime information
Live-arrival data covered 100% of scheduled trips in the sampled window.

Important: This does not rate service reliability. Riders should confirm current service alerts, fares, and accessibility accommodations with the transit operator before traveling.

Send your vendor a fix request

You may not control the GTFS export yourself. Copy this and send it to whoever runs your scheduling software export. It names each fix with the validator notice and a guide link.

Score by category

Correctness64.0 / 100

The MobilityData validator flagged 6 kinds of issue across 45997 instances (1 error, 45996 warning, 0 informational).

Freshness100.0 / 100

Service data covers the next 189 days.

Rider experience37.5 / 100

0% of stops state wheelchair accessibility (0% marked accessible, 0% marked not accessible). This measures what the feed publishes, not whether a stop is physically usable. Fare data is not published.

Accessibility37.5 / 100

0% of stops state accessibility (0% marked accessible). Reflects what the feed states, not verified physical usability.

1 accessibility depth signal

  • 1 route badge(s) pair a color and text color below the WCAG 4.5:1 contrast bar (SUD_IV15).

    Consider: Adjust route_color or route_text_color so each pair clears 4.5:1. Often switching the text between black and white is enough.

Opportunities to strengthen the data, not deductions from the sub-score above. States what the second accessibility lens can check from the feed, not verified physical usability.

Realtime quality70.6 / 100

Sampled 3 times: 1 of 1 configured feed healthy; 100.0% of scheduled trips had live predictions; predictions ran a median of 0s behind schedule.

Routes and stops

This feed has no route shapes, so the map shows its stops only.

This feed has 12 stops.

Skip to route and stop data

The route and stop data is ready below. Load the map only when you want the geographic view. It uses additional data.

Basemap: OpenFreeMap, © OpenStreetMap contributors. Routes and stops: this agency's GTFS feed.

Routes in Trains régionaux ZOU ! — Transdev Rail Sud's feed
RouteTypeLine color
SUD_IV15 Nice - MarseilleRailgray (no shape in feed)
List every stop
  • Marseille-St-Charles
  • Marseille-Blancarde
  • Aubagne
  • Cassis
  • Toulon
  • Carnoules
  • Les Arcs-Draguignan
  • Nice-Ville
  • Nice-St-Augustin
  • St-Raphaël-Valescure
  • Cannes
  • Antibes

Over time

Overall score across the last 15 checks — down 4.3 since 2026-10-15.

2026-10-02: 65.92026-10-03: 65.92026-10-04: 65.92026-10-05: 65.92026-10-06: 65.92026-10-07: 65.92026-10-08: 65.92026-10-09: 67.02026-10-10: 65.92026-10-11: 66.32026-10-12: 65.92026-10-13: 65.92026-10-14: 65.92026-10-15: 70.22026-10-16: 65.9
Show the numbers
CheckScoreChange
2026-10-0265.9 first check
2026-10-0365.9 no change
2026-10-0465.9 no change
2026-10-0565.9 no change
2026-10-0665.9 no change
2026-10-0765.9 no change
2026-10-0865.9 no change
2026-10-0967.0 up 1.1
2026-10-1065.9 down 1.1
2026-10-1166.3 up 0.4
2026-10-1265.9 down 0.4
2026-10-1365.9 no change
2026-10-1465.9 no change
2026-10-1570.2 up 4.3
2026-10-1665.9 down 4.3

What changed since your last check

  • Correctness no change
  • Freshness no change
  • Rider experience no change
  • Realtime quality down 21.6

Feed history

Every check on record for this feed, grouped by the published file, newest first. A new file means the published zip changed. Grades from earlier rubric versions are shown with their rubric and are not comparable with today’s grade or the trend above.

CheckedPublished fileService dates throughGrade at the time
2026-10-06 to 2026-10-16 (11 checks)f98975c699ab2027-04-23Current rubric: see the trend above
2026-10-050e18b418f8b72027-04-22Current rubric: see the trend above
2026-10-042e5597d734492027-04-21Current rubric: see the trend above
2026-10-03d74443437ec32027-04-20Current rubric: see the trend above
2026-10-025918e2db43cc2027-04-19Current rubric: see the trend above
2026-10-0168b2ee26292c2027-04-18D, 65.9 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-30568192c4cee82027-04-17D, 64.5 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-29901ca95491f12027-04-16D, 61.8 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-281cc1e70f22142027-04-15D, 61.8 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-27a2c7fd99684f2027-04-14D, 61.8 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-266ee8c389354b2027-04-13D, 65.9 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade
2026-09-19 to 2026-09-25 (7 checks)8a2159a8ce2e2026-12-11C, 70.1 under rubric 1.3, an earlier scoring contract; not comparable with today’s grade

Showing the 12 most recent of 21 periods.

What changed in this feed

Grade dropped from C to D since 2026-10-15.

Same feed file as 2026-10-15; the published zip did not change.

No longer reported since 2026-10-15 (1 finding)

  • Some live predictions are more than 30 minutes away from the schedule. (scorecard_rt_predictions_implausible)

This records the later feed state. It does not establish who made a change or why.

Subscribe to this feed’s changes (Atom) to hear about grade drops and export changes in a reader, with no sign-up.

What changed over time

On 2026-10-02 this feed started at grade D. On 2026-10-15 the grade moved from D to C. On 2026-10-16 the grade moved from C to D. On 2026-10-16 the check no longer reported: some live predictions are more than 30 minutes away from the schedule. As of 2026-10-16 it holds grade D.

A plain-language history of this feed, newest first.

  • 2026-10-16 Grade went C to D, realtime fell 22 points.
  • 2026-10-15 Grade went D to C, realtime rose 22 points.

Everything we checked

11 findings, ordered by severity.

Show every finding
  • Error1 instance

    Foreign key violation (flagged by the MobilityData validator).

    See the linked rule for what this affects.

    Fix: Review the rule documentation for 'foreign_key_violation' at https://gtfs-validator.mobilitydata.org/rules.html and check the flagged rows in your feed. (Varies.)

    Finding code: foreign_key_violation · See MobilityData GTFS Validator rules

  • Warning45992 instances

    Some rows in stop_times.txt leave the timepoint column blank.

    That column says whether a time is a real checkpoint or an estimate. Blank leaves apps guessing, and they may show your estimates as promises.

    Fix: Set timepoint=1 on stops that are real time checks and 0 on the rest. If every published time is a scheduled one, mark them all 1. (Usually one export setting.)

    Finding code: missing_timepoint_value · Read the fix guide for missing_timepoint_value · See MobilityData GTFS Validator rules

  • Warning5486 instances

    5486 of 5486 trips have no headsign, the text that tells riders where the transit vehicle is going. A trip counts when it has trip_headsign or a stop_headsign at every stop.

    When a route has multiple directions or patterns, the route name alone may not tell riders which service is coming.

    Fix: Add the destination, direction, or 'via' label riders actually see to trip_headsign. Do not copy the route name. If the label changes during the trip, use stop_headsign. (Usually one value per route pattern in your scheduling source.)

    Finding code: scorecard_missing_headsigns · Read the fix guide for scorecard_missing_headsigns · See GTFS Best Practices

  • Warning12 instances

    12 of 12 stops don't say whether a wheelchair user can board there.

    Riders who use wheelchairs can't plan a trip when accessibility is marked 'unknown'; apps show no information at all.

    Fix: Set wheelchair_boarding to 1 (accessible) or 2 (not accessible) for every stop. A field survey can start with the busiest stops. (A column in stops.txt; your scheduling software likely has it.)

    Finding code: scorecard_wheelchair_boarding_unknown · Read the fix guide for scorecard_wheelchair_boarding_unknown · See GTFS Schedule reference

  • Warning1 instance

    feed_info.txt lists no contact email and no contact URL.

    Without a contact, apps like Google Maps can't reach you when they find a problem in the feed.

    Fix: Add feed_contact_email or feed_contact_url to feed_info.txt. (One field, set once in export settings.)

    Finding code: missing_feed_contact_email_and_url · Read the fix guide for missing_feed_contact_email_and_url · See MobilityData GTFS Validator rules

  • Warning1 instance

    Some files leave out fields that GTFS asks for but does not require.

    Recommended fields like agency_phone or stop descriptions make the feed more useful to riders and trip planners.

    Fix: Review the flagged fields and fill in the ones your riders would use. (A field at a time; not urgent.)

    Finding code: missing_recommended_field · Read the fix guide for missing_recommended_field · See MobilityData GTFS Validator rules

  • Warning1 instance

    Some rider-facing names are in ALL CAPS or all lowercase.

    ALL-CAPS stop and headsign names are harder to read in apps and are read awkwardly by screen readers.

    Fix: Use mixed case for stop names and headsigns (e.g. 'Main St & 2nd Ave', not 'MAIN ST & 2ND AVE'). (Often a bulk fix in your scheduling software.)

    Finding code: mixed_case_recommended_field · Read the fix guide for mixed_case_recommended_field · See MobilityData GTFS Validator rules

  • Warning1 instance

    Unexpected enum value (flagged by the MobilityData validator).

    See the linked rule for what this affects.

    Fix: Review the rule documentation for 'unexpected_enum_value' at https://gtfs-validator.mobilitydata.org/rules.html and check the flagged rows in your feed. (Varies.)

    Finding code: unexpected_enum_value · See MobilityData GTFS Validator rules

  • Warning1 instance

    The feed contains no fare information.

    Riders see 'fare unknown' in trip planners and can't budget their trip; visitors are most affected.

    Fix: Add fare_attributes.txt (or Fares v2 files) with your fare structure. If your service is fare-free, ask to have it marked fare-free instead. (A small file for most flat-fare systems.)

    Finding code: scorecard_no_fare_data · Read the fix guide for scorecard_no_fare_data · See GTFS Best Practices

  • Warning1 instance

    Realtime data was up to 3055 seconds old when sampled.

    Stale positions and predictions are worse than none: riders watch a transit vehicle that already left.

    Fix: Ask your AVL vendor to send updates at least every 20 seconds. That is the Caltrans guideline. (A vendor configuration question.)

    Finding code: scorecard_rt_stale

  • Info1 instance

    feed_info.txt has no technical contact (feed_contact_email or feed_contact_url).

    App makers and state data programs have nobody to email when they spot a problem. Small errors then sit in the feed for months.

    Fix: Add feed_contact_email to feed_info.txt. (One field.)

    Finding code: scorecard_no_feed_contact · Read the fix guide for scorecard_no_feed_contact · See MobilityData GTFS Validator rules

Conformance mark Not yet

Two requirements remain for this feed to earn the conformance mark. 1 validator error to resolve. States wheelchair access on 0% of stops and 100% of trips; the mark needs 90% of each.

Valid Not yet
1 validator error to resolve.
Current Met
Service data covers the next 189 days.
Accessible Not yet
States wheelchair access on 0% of stops and 100% of trips; the mark needs 90% of each.

In plain words: earn this mark when your feed passes validation, has not expired, and says whether nearly every stop and trip is wheelchair accessible.

A pass credential for a feed that is valid, current, and states wheelchair access on nearly every stop and trip. Accessibility here measures what the feed publishes, not whether a stop is physically usable. How the conformance mark works.

Realtime reliability

The realtime feed responded on 100.0% of 200 checks in the 25 days to 2026-10-16, with 3279s median lag.

Sampled on a schedule between full scores, so this tracks uptime and freshness over time rather than at a single moment. The window ends at the newest observation, which is not always today: if the monitor stops, this window stops with it.

Live predictions vs schedule

Arrival predictions ran right on the schedule, and stayed within 0s nine times in ten. They were on time (about a minute early to five late) 96.8% of the time.

From the last full realtime sample: how far live arrival predictions sat from the schedule, and whether vehicle positions fell on the route. These feed the realtime score; they change no other category.

Clears the Google and Apple Maps four-week coverage bar. This feed has 189 days of service ahead, clearing the four-week (28-day) window Maps asks for. The feed also carries 1 validator error, the other thing Maps checks at onboarding; the findings below name each fix.

How this agency maps to the standards

The scorecard is a data-quality lens, not a compliance determination. Its universal references describe good GTFS and useful rider information. Useful references here are GTFS Schedule Best Practices, GTFS-Realtime Best Practices, MobilityData grading scheme, Google Transit publication guidance. Read the full standards crosswalk.

Correctness 64 / 100
GTFS Schedule best practices, checked by the MobilityData validator. MobilityData grading covers stop locations, route names, and colors. Google Transit requires a feed to pass validation for publication.
Freshness 100 / 100
GTFS Schedule best practices call for a dataset that stays current. An expired calendar can remove service from Google Transit and other rider trip planners.
Rider experience 38 / 100
GTFS Best Practices for rider-facing fields. MobilityData grading covers stop names and headsigns.
Realtime quality 71 / 100
GTFS-Realtime best practices: a stable URL, high uptime, and frequent updates.

Part of the France portfolio: the same scorecards for every feed we track in that group, on one free page.

Similar feeds in Provence-Alpes-Côte-d’Azur

Other scorecards in Provence-Alpes-Côte-d’Azur, starting with the same kind of service at a similar size. Listed for comparison, not ranked.

Show your grade

Put this badge on your agency site or feed README. It updates after each completed scoring check, shows the date of that check, and links back to this scorecard. When there is no current score it says so instead of showing an old one.

Trains régionaux ZOU ! — Transdev Rail Sud: GTFS feed quality grade D, 65.9 of 100, checked 2026-10-16. GTFS Scorecard.

Prefer a shields.io style? Point a dynamic endpoint badge at the published badge.json.

Cite this record

This page updates on every check. The record below does not: it is the dated file this grade came from, published at https://gtfsscorecard.org/data/artifacts/fr-pan-83448/2026-10-16.json and never overwritten, pinning the grade, category scores, rubric version, validator version, reader archive profile, and the scored feed's sha256 as they stood on 2026-10-16. Use it in a board packet, a regulatory filing, or a research citation instead of linking the live page, whose content will differ on your next visit.

Citing the tool itself rather than one agency's record? Use the repo's CITATION.cff.