Penalties
Financial audit
We track the “Invisible Undelivered Service”: ghost trains, route limitations, late cancellations.
We quantify the precise amount of unreceived savings (beyond penalty caps).
Transport Authorities Take back control of your railway data
The independent, AI-augmented client-side advisory (AMO) for the authorities that fund a delegated service. Steer, audit and negotiate your contracts on facts — not on the operator's declarations.
Contractual advisory • Data audit • Competition readiness
Request a diagnosticRecord · France, eleven regions · September 2026
246,259trains in the transport plan
from 1 to 30 September
239,448ran97.2% of the plan
6,753with no recorded run2.7% of the plan, broken down below
And 58 trains removed from the plan before their day of service, yet seen running.
- 970Cancelled removed from the plan before 5 pm the day before
- 478Cancelled late removed from the plan between 5 pm and midnight the day before
- 4,830Not observed still in the plan at midnight, no stop observed
- 211Replaced by a coach replacement declared by the carrier
- 264Unqualified real-time feed silent that day
Audit turnaround
D+1
Audit turnaround, vs. D+30 for traditional methods.
Markellia
D+1
Traditional methodsD+30
DD+10D+20D+30
01The finding
Breaking the information asymmetry
Transport authorities fund a delegated service based on the operator's declarations — with no independent way to verify them. To steer their contracts, audit penalties and prepare for competition, they lack a trusted third party, independent of the operator.
Observed in France, eleven regions · data published by the operator
Self-reported figures
The figures used to calculate your penalties are supplied by the operator.
Published by the operator · August 2026
14rows for all of France: four counts and one rate per network
- 213,574trains scheduled
- 208,830ran
- 6,121cancelled
- 21,186late on arrival
Sum of the 14 rows, data.sncf.com. As of 17 October, August is the latest month published.
Recorded by Markellia · August 2026
219,511rows, one per planned train, ready the next day
The operator only publishes the trains it runs; Markellia records every train serving a station in the eleven regions. The black box
A large share of incidents is classified by the operator itself, often under exonerating causes (weather, infrastructure, third parties).
5,498delays of more than 5 min with no identified reason in September, out of 20,640
- 5,498 no reason at all
- 3,741 one of the three phrases below
- 11,401 another reason
The three most published reasons
- “Reuse of a train”1,566
- “Traffic management difficulty on the rail network”1,344
- “Incident while preparing the train”831
Always the same phrase, with no place or detail. Every published explanation comes from the operator itself. Undelivered service
Are you paying for trains that don't run, or that stop before their terminus?
4,830trains in the day’s plan, never seen running, in September
From 39 trains on Saturday 19 to 310 on Friday 4. Shaded: Saturdays and Sundays.
Measured since 1 October
589trains stopped short of their terminus from 1 to 13 October, and 2,796 stops not served
One square per train
Not observed: in the day’s plan and never seen running, while real-time data was being received.
The Markellia mission
Markellia is the independent client-side advisory (AMO) for transport authorities. We support you across the full lifecycle of your contracts — diagnosis, negotiation, execution, renewal — leveraging open data, AI and our railway expertise. We are paid only by the authority, never by the operator.
02The solution
From audit to steering, all the way to competition
We reconstruct the reality of the service delivered from open data, then equip your contractual steering: grey-zone reclassification and support for drafting your future tenders.
| Number | What Markellia does | PenaltiesFinancial audit | Grey ZoneContract steering | AutonomyData strategy |
|---|---|---|---|---|
| 1 | Verify service reality: did each billed train actually run? | yes | yes | no |
| 2 | Measure punctuality: delays by train, station and line. | yes | yes | yes |
| 3 | Detect unexpected cancellations and ghost trains. | yes | no | no |
| 4 | Distinguish internal causes (penalisable) from external causes (force majeure). | yes | yes | no |
| 5 | Identify causes via AI: strikes, weather, construction, equipment failures. | no | yes | no |
| 6 | Produce automated analysis reports (daily or weekly). | no | yes | yes |
| 7 | Reduce the grey zone: reclassify delays as internal causes, hence penalties. | yes | yes | no |
Seven capabilities, three audits
Verify service reality: did each billed train actually run?
PenaltiesGrey Zone
Measure punctuality: delays by train, station and line.
PenaltiesGrey ZoneAutonomy
Detect unexpected cancellations and ghost trains.
Penalties
Distinguish internal causes (penalisable) from external causes (force majeure).
PenaltiesGrey Zone
Identify causes via AI: strikes, weather, construction, equipment failures.
Grey Zone
Produce automated analysis reports (daily or weekly).
Grey ZoneAutonomy
Reduce the grey zone: reclassify delays as internal causes, hence penalties.
PenaltiesGrey Zone
Grey Zone
Contract steering
We reclassify root causes day after day. A rolling stock failure hidden behind an infrastructure incident? Staff shortage masked by weather?
We restore contractual truth, month after month.
Autonomy
Data strategy
We help transport authorities build the data clauses of their contracts: data access requirements, auditable indicators, governance.
03The platform
A sovereign data platform, hosted in France
Markellia continuously ingests the operator's public data and autonomously computes the contractual gaps owed to the authority.
- 1
Data sources
Public operator data, field sensors, reference catalogs and AI models.
GTFS · GTFS-RT · SIRI ET
- 2
Ingestion & Storage
A sovereign lakehouse hosted on OVHcloud, in France.
OVHcloud · France
- 3
Analysis & Reconciliation
Cross-referencing planned service with actual delivery, line by line, train by train.
- 4
Delivery
Reports by email and an audit space on app.markellia.com
The same vocabulary, from the website to the platform: each authority finds its planned trains, those that ran, and the others.
See the audit space →Illustration · one train, one state
6,753 with no recorded run2.7% of the plan
04Our commitments
Autonomy accelerator for transport authorities
Sovereignty
Data stored and processed in France (OVHcloud)
Transparency
Rebalances the relationship with the operator
Independence
Paid by the authority alone, never by the operator.
Pooled cost
Shared infrastructure across all regions: 10–50× cheaper than a tool built alone.
05Our fields
Rail & urban transport
Our core business: auditing TER conventions. The same approach extends to urban transport contracts.
TER conventions & urban contracts
Auditing of TER conventions — and, by extension, urban transport contracts: penalties, grey zone, competition readiness. Delivery on app.markellia.com.
06The team
Expertise from the inside
Markellia is not an external audit firm: we come from the data leadership of the incumbent operator, whose information systems we designed and ran.
Arnaud Lutellier
Founder & CEO · 25 years leading data (Canal+, SNCF, Lacoste)
He turns technical information into a financial performance lever for Transport Authorities.
“We don’t guess how your data is produced.
We know.”
07The offer
A risk-free experimentation
One month to discover, then an analysis that grows richer at each step.
One month to discover
One month of access to the audit space, on your region’s trains.
Analyses grounded in public data
Transport plan, real-time data and reasons published by the operator, cross-checked every night and delivered at D+1.
Indicators tailored to your contract
Thresholds, scope and indicators taken from your contract, such as the delay threshold it sets.
Additional data, including adjustments
The analysis draws on additional data, starting with adjustments.
Perceived quality data
Passenger surveys and perceived-quality indicators set by your contract, alongside observed quality.
Request a diagnostic
Contractual advisory • Data audit • Competition readiness
08Questions
Frequently Asked Questions
What is a railway data audit for a Transport Authority?
A railway data audit is an independent analysis of production data (traffic, incidents). Markellia performs this work by cross-referencing its own open data with data provided by the operator (when available). This approach identifies gaps between reported figures and operational reality, revealing unapplied penalties and misclassified incident causes.
How does a Markellia audit work?
An audit covers a raw data sample spanning one quarter of operations — past (if data is available) or upcoming. It produces a detailed map of financial and operational gaps.
What is the "grey zone" in railway data?
The "grey zone" refers to incidents whose root cause is ambiguous or misclassified. A large share of incidents is classified by the operator itself, often under exonerating causes (weather, infrastructure, third parties). Markellia reclassifies these causes to restore contractual truth.
How does Markellia fund its services?
Markellia’s engagements are often self-financed: savings identified from the first detected anomalies typically cover Markellia’s fees. Additionally, engagements under €100,000 can be carried out under innovative procurement rules, without a formal tender process.
What is the expertise of the Markellia team?
Markellia was founded by Arnaud Lutellier (Founder & CEO), a CentraleSupélec engineer and data leader for 25 years (Canal+, SNCF, Lacoste). He led the data and customer platforms of SNCF Gares & Connexions for more than ten years.
Backed by:
