The dispatch log has four owners and none of them is responsible

Mobility LabJorge Aguilar SalasPublished 2026-09-2018 min read

German law requires a private hire vehicle to return to its base after every trip unless a new order arrives before or during the ride. The duty has said the same thing since 1983; what changed is what answers it. Whether a fleet can show it complied now turns on the dispatch log, which is also the platform's optimisation tool, the regulator's file and, in one 2026 Federal Court of Justice case, the competitor's evidence. In a Bavarian licence case that year, the missing part of that record was enough to cost an operator its licence, in an interim decision under appeal. Following the log through four pairs of hands shows a process where whoever decides does not answer, whoever answers cannot watch, whoever produces the data does not file it, and whoever files it does so by hand on someone else's numbers.

Declaration of interest. The author is co-founder of Nexio Fleet, a company that operates in fleet compliance, the same space this note describes, and previously worked at Bolt and Bliq, which operate in the German ride-hailing market this note examines. Every claim here rests on a public source the reader can open. The note would stand unchanged if Nexio did not exist.

OIM / FN01 / 01

1. 10:10Passenger dropped off. Trip ended.
2. 10:13Test order placed through Uber
3. 10:13Accepted. Confirmation names the operator.
4. 10:22Driver logs out. Car had not moved.

12 minutes

Breslauer Platz, Cologne

On 19 January 2023, a private-hire car stopped at Breslauer Platz in Cologne after dropping off a passenger.³ Then it did something remarkably ordinary.

Nothing.

The car stayed there. At 10:13, a board member of the taxi cooperative that would later sue (standing there, watching the car) placed a test order through Uber. It was accepted at once. The confirmation told the customer the ride would be carried out by the company that operated the vehicle. The order was cancelled. The car stayed where it was until 10:22, when the driver logged out.³

Twelve minutes of waiting. More than three years later, those twelve minutes were part of a case before Germany's Federal Court of Justice.

The case had been brought by a Cologne taxi cooperative against a competing private-hire operator. The cooperative had obtained an injunction in the lower courts over violations of Germany's Rückkehrpflicht: the rule requiring private-hire vehicles to return to their operating base after completing a trip unless another order arrives.³¹ In June 2026, the Federal Court of Justice upheld that result.³

That makes the twelve minutes interesting. But what makes them much more interesting is how the violation was demonstrated.

There was no inspector standing on Breslauer Platz with a stopwatch. No police officer had to follow the vehicle. Instead, a competitor placed an order through the platform and the platform answered. The acceptance proved the car was available on the street; the confirmation named the company responsible for it.³

A competitor had used the dispatch system to document what a rival's vehicle was doing. The system designed to assign rides had become a way of documenting whether another operator was following the rules. And that exposes an unusual chain of responsibility.

There are four actors in this story. The driver controls what the vehicle actually does. The platform controls the dispatch system and generates much of the trip data. The fleet operator carries the legal obligation to maintain the relevant record.¹ The authority decides how that record will be inspected and what evidence it expects.⁶⁴ They all touch the same operational process.

None controls the whole thing.

OIM / FN01 / 02

TripOrderMovementRecordReview

1. 01DriverOperates the vehicle and accepts orders³Acts
2. 02PlatformDispatches orders and provides data³Assigns
3. 03OperatorHolds the licence and keeps records¹Records
4. 04AuthorityReviews compliance and can initiate proceedings⁶⁴Audits

EvidenceA shared operational reality, assembled from different systems.

A rule designed before Uber existed

Germany's return obligation isn't new. Its modern form comes from the fifth amendment to the Passenger Transport Act, passed on 25 February 1983 and published in the federal gazette on 1 March. The idea was straightforward: a Mietwagen, a private-hire vehicle, wasn't supposed to behave like a taxi.¹²

A taxi could wait for customers in public. A private-hire vehicle couldn't. Once a trip ended, it had to return to its operating base unless another booking had already arrived.¹

When Germany's Constitutional Court examined the rule in 1989, it made an important distinction.

"Der Zweck des Rückkehrgebots liegt nicht in der Rückkehr selbst."²

Or, roughly: the purpose of the return obligation is not the return itself. The purpose, the court explained, was to prevent private-hire cars from standing around waiting for passengers like taxis.²

That meant a vehicle didn't necessarily have to make it all the way back to base. If another radio-dispatched order arrived while it was returning, the driver could turn around and take it.² So the rule created a peculiar window: the car can receive another trip while returning. It just can't sit somewhere waiting for one.² That distinction became much more important once dispatch moved onto smartphones.

The rule stayed. Dispatch changed.

Digital dispatch changed how the next trip reaches the driver. An order can now arrive through an app while the vehicle is close to base, halfway back or still near the previous passenger's destination. And the moment that order arrives isn't a fact of nature.

It's a design decision.

A platform can dispatch each request the instant it appears, or hold requests for a few seconds and solve a batch. Research on ride-hailing compares those strategies on a simulator calibrated with New York trip records: letting a learned policy decide when to trigger the match cut total waiting time by 3.1% in ride-hailing and 20.1% in pooling.⁷

Those numbers are about waiting time, not about law. The paper doesn't mention Germany, and it models a simulator rather than a deployed system.⁷ But the choice it describes is the same one that decides, in Germany, whether a stationary vehicle is merely idle or already in breach. Trigger the match a few seconds later and the car is still sitting at the drop-off point with no follow-on order to justify it.¹

Germany eventually acknowledged that technological reality. The 2021 reform of the Passenger Transport Act explicitly recognised electronic order records, including app-based systems.¹ But the return obligation remained.¹

Today, two systems therefore operate alongside each other. The dispatch system asks: which vehicle should get the next passenger? The legal system asks: what was that vehicle doing between the previous passenger and the next one?

Most of the time, the distinction is invisible. Until there is a gap.

Imagine a passenger gets out at 10:10. No new order exists. The vehicle should start returning. At 10:13, another order arrives. Those few minutes suddenly matter.

Did the driver start returning? Did the car remain parked? Had another order already been assigned? Where was the vehicle when it arrived?

The platform records when the next order enters the dispatch flow. The vehicle's movement shows what happened in between. But the operator may later have to account for both.¹⁴ That's where a traffic rule becomes an operations problem.

The company responsible doesn't control the whole record

The federal rule is unusually clear about one thing. The record-keeping obligation in § 49(4) sits with the Mietwagenunternehmer, the operator.¹ Not the platform.¹ But reconstructing what happened between two trips can depend on data generated elsewhere.⁴ That creates an awkward arrangement: the company responsible for the record does not necessarily generate the data needed to build it.

And the federal law is only half of the story. The return obligation is the same across Germany.¹ The audit process isn't. Berlin provides a documented example.

In a 2023 Senate response, the city's licensing authority said that operators using app-based dispatch must submit the ride data for every licensed vehicle as unaltered original electronic files.⁶ The authority uses those records to check compliance with the return obligation: through sampling, or when discrepancies are apparent.⁶

So this isn't simply a fleet owner keeping a file somewhere in case somebody asks. Platform-generated operational data can enter a regulatory audit.⁶⁴ And another authority can implement that audit differently. A 2026 case from Bavaria shows how far that can go.

When 210 trips become an investigation

In Ansbach, the licensing authority reconstructed individual trips.⁴ For fifteen vehicles on one day, it analysed 210 assessable trips semi-automatically and identified 67 cases it considered breaches of the return obligation.⁴

To turn the law into something a dataset could test, the authority had to translate a legal requirement into operational rules. For example, it treated a vehicle as having returned once it came within 250 metres of its operating base.⁴

OIM / FN01 / 03

• 15vehicles
• 210assessable trips
• 67flagged breaches

250 mtreated as returned

Threshold applied by this authority. Not defined by federal statute.

That 250-metre threshold isn't written into the federal statute.¹⁴ It was how this particular authority operationalised the rule. Then another problem appeared.

The operator's licence required it to report either the time a vehicle returned to base or the identifier of the follow-on order explaining why it didn't.⁴ That information was missing.⁴ And that mattered.

The case wasn't only about whether drivers had violated the return obligation. It was also about whether the operator could document what its vehicles had done. When the required records were incomplete, the evidentiary consequences fell on the operator.⁴

There are important limits to what this case tells us. It involved interim relief and a summary judicial review, not a final judgment, and an appeal was pending. It also reflects one authority's approach, not a national audit standard.⁴

But operationally, it exposes the mechanism clearly. A driver finishes a trip. A platform records dispatch events. A vehicle moves. Or doesn't. The operator has to turn those events into the record its licence requires.¹⁴ Months later, an authority may try to reconstruct those few minutes from the data left behind.⁴

The driver decides. The operator answers.

There is an uncomfortable asymmetry here. The decision to wait or return happens inside the car. German law can treat a breach of § 49(4) as an administrative offence by whoever intentionally or negligently violates it.⁵ But the consequences don't necessarily stop with the driver.

In the Bavarian case, the court attributed breaches committed by individual drivers to the managing director of the operating company unless they would have happened even with adequate supervision and organisation.⁴ In other words, saying "the driver did it" doesn't automatically solve the operator's problem. At company level, § 25 of the Passenger Transport Act also makes reliability relevant to whether an operator keeps its licence when obligations continue to be breached despite written warning.⁵

So the fleet owner faces a difficult operational question: How do you supervise a rule that is kept or broken vehicle by vehicle, minute by minute?

Watching every driver isn't realistic. The alternative is to make the operation observable through data. And that makes the quality of the record much more important than it first appears.

How expensive is the rule?

Operators have long argued that forcing vehicles to return creates unnecessary empty kilometres. A 2024 Fraunhofer FOKUS study commissioned by the operators' association wirfahren estimated that 28% of rental-car distance was caused by the return obligation, and extrapolated a saving of exactly 194,064 kilometres per day for Berlin.¹⁰

Those are the figures in the study. The figures in circulation are different: the press releases and the association's own interview say 30% and "almost 200,000".¹⁴

Those are striking numbers. They also need context, and the context starts with who asked for them.

wirfahren's registered speaker in the Bundestag lobby register is Thomas Mohnke.¹¹ Mohnke is the managing director of SafeDriver Group,¹² the group whose Ennoo subsidiary took on the role of Uber's general contractor in Germany when that structure was set up in 2019.¹³

So the figure that argues against the rule was commissioned by an association speaking for the operators the rule constrains. That doesn't make it wrong. It makes it a position, and a position is worth reading closely.

The study attributes the 28% figure to research it describes as still unpublished. More importantly, its alternative scenario assumes that drivers remain at the drop-off location waiting for another order.¹⁰ That is precisely the behaviour the current rule prohibits.¹²

So the estimate is useful for asking what might happen if the rule changed. It shouldn't be confused with a measurement of what compliance currently costs across the entire industry. And when the association's own speaker cited the figure in a newspaper interview, it had grown to 30%.¹⁴

We don't actually have that measurement. Official German passenger-transport statistics don't provide national trip data for taxi and Mietwagen operations in the same way they do for rail, tram and bus.⁹ Berlin publishes licence counts, but licences aren't trips.⁸ And Berlin's licensing authority has said it doesn't statistically record the results of these audits or the reasons licences are revoked.⁶

That leaves a strange evidence gap. Individual cases can be reconstructed with extraordinary precision: down to minutes, locations and individual orders.⁴ But the public data can't tell us how common the problem is across the market.⁸⁹

The dispatch log has changed jobs

This is where the technology story becomes more interesting. The dispatch log began as an operational tool.

A passenger requests a ride. A vehicle gets an assignment. The trip starts. The trip ends. Another assignment eventually arrives.

Those records help the marketplace run. But the Cologne and Bavarian cases show that the same data can serve another purpose.

Evidence.

In Cologne, a competitor placed an order through the dispatch system and used the platform's own confirmation to establish both that the vehicle was available and whose vehicle it was.³ In Bavaria, trip data was analysed semi-automatically to reconstruct whether vehicles had complied with the return obligation.⁴

The data didn't fundamentally change. Its job did.

And once operational data becomes evidence, mundane questions become important. Who generated this timestamp? Can the operator retrieve it months later? What happens when data is missing? Does the platform's export contain what the operator's licence requires? Can a follow-on order be linked reliably to the vehicle that didn't return? And what happens if the flow of data stops?

The Ansbach case gives one indication of how serious that last question can become. After 1 December 2025, relevant data was no longer transmitted, a fact recorded in that same interim decision, on summary review and under appeal.⁴ The obligation didn't disappear with the data. It remained with the licence holder.¹⁴

The missing field

The Ansbach case also exposes how small the gap can be. The operator's licence required a return timestamp or the identifier of the follow-on order.⁴ Those aren't sophisticated data points. But without them, the authority couldn't reconstruct the operation in the same way.⁴

That matters because the information needed to explain a vehicle's behaviour can be distributed across systems. The platform produces dispatch data. The vehicle and driver produce the movement. The operator is responsible for the record.¹ And the licensing authority determines what additional information its audit requires.⁴⁶

Uber's public German page for prospective rental-car operators illustrates part of that disconnect. It explains signup, vehicle requirements and the supplier portal. It doesn't explain the return obligation, the order book, what compliance data an operator will receive or how that data can be used to document compliance.¹⁵

That doesn't prove those tools or data don't exist. The proceedings themselves show that platform data can become part of an investigation.⁴ The narrower point is more useful: data created to operate the marketplace can acquire a second job: proving that the operation complied with rules the platform itself doesn't carry.¹

The duty itself has said the same thing since 1983.¹² What changed is what answers it: whether a fleet can show it complied now turns on a record the dispatch system produces, and in Ansbach the part of that record that was missing was enough to decide the case.⁴ Once that is true, managing the fleet starts to mean something different.

System check / 01

What breaks when compliance depends on operational data?

1. Auditabilitychecked

   Can the operation be reconstructed after the fact?

   In both Cologne and Ansbach, digital records made vehicle behaviour reconstructable. But that reconstruction depends on whether the relevant events (trip end, follow-on order, return or movement) are recorded and can later be connected.³⁴

2. Vendor dependencychecked

   What happens when evidence lives outside your system?

   The operator carries the record-keeping obligation while some of the data needed to reconstruct the operation can originate in platform systems it doesn't control.¹⁴

3. Failure modechecked

   What happens when one part of the evidence chain disappears?

   A missing timestamp, an unlinked follow-on order or interrupted data transmission (as recorded in the Ansbach proceedings, on summary review and under appeal) can turn a routine trip into an exception that requires manual reconstruction.⁴

4. Data minimisationopen question

   How much evidence is enough?

   § 49(4) requires incoming orders to be recorded and retained for one year. Modern systems can produce far richer operational data. The design problem is determining what is actually necessary to demonstrate compliance, not simply collecting everything available.¹

When operations becomes system design

The process itself is simple enough to describe. A trip ends. If no follow-on order exists, the vehicle should return. Another order may arrive while it is returning.¹² The operator may later need to demonstrate what happened in between.⁴

At small scale, you can imagine doing that manually. Open the trip. Check the timestamps. Look at the vehicle movement. Ask the driver. Correct the record.

But the Bavarian authority analysed 210 trips for just fifteen vehicles on a single day.⁴ Scale that across a fleet, every day, and manual supervision stops being an operating model.

No Operations Manager can watch every vehicle between every two trips.

OIM / FN01 / 04

Before

Manage the process

Reactive. Manual. Case by case.

1. TripA trip is completed
2. DriverThe driver continues working
3. CheckSomeone reviews the case, if flagged
4. CorrectAction is taken manually

• Time dependent
• Fragmented
• Not scalable

Works at low volume. Breaks at scale.

Same reality. Different job.

Now

Supervise the system

Proactive. Rule-based. Exception-driven.

1. EventAn operational event is recorded
2. RuleThe system checks it against compliance and business rules
3. DetectionPotential issues are flagged automatically
4. ExceptionHumans review what matters
5. HumanFocus on judgment, not monitoring every trip

• Systematic
• Scalable
• Consistent
• Auditable

From individual cases to an operational system of record.

So the work moves one level up. Instead of asking: Did this driver return to base after this trip?

Operations has to build a system capable of answering: Which vehicles should have returned, which appear not to have done so, and which cases actually require human attention?

That requires more than a dashboard. The system needs to know when a trip ended, whether a valid follow-on order existed, what the vehicle did next, whether the required evidence is complete and when a missing or contradictory record should become an exception for someone to investigate.

The Operations Manager's role changes with it. Less time reconstructing individual trips. More time defining what the system should detect. Which data is required. Which rules can be checked automatically. Which exceptions matter. And when a human should intervene.

The twelve minutes in Cologne show why. A competitor could test a vehicle's behaviour through the dispatch system. A court could later use the resulting record to assess what happened.³ In Bavaria, an authority could take the same basic class of operational data and analyse hundreds of trips semi-automatically.⁴

Technology didn't remove the operational responsibility. It changed the way that responsibility has to be managed. The work moves from trying to watch the process happen to designing a system that can tell you where the process needs attention.

Operator's take

What strikes me about this case is how far the risk travels from the decision. A driver decides to wait; § 25 answers by putting the operator's licence at stake.⁵ In between sit a platform that generates much of the data and an authority that decides what it wants to see when something is investigated. That is my reading, not a finding in the case: nothing here measures what each actor stands to lose.

If I were running that fleet, my first move would not be hiring someone to check what sixty or seventy drivers are doing. One authority needed a semi-automated review to get through 210 trips from fifteen vehicles on a single day.⁴ At that scale, manual checking is already too late. What I would want first is one place where the operation can be reconstructed: when a trip ended, whether a follow-on order existed, what the vehicle did next, and whether that sequence holds together. Not more data. Visibility.

Once that exists, I don't think the point is watching every driver all the time. The system should surface the exceptions that need attention. If a driver appears not to return once, the response is probably a notification: this happened, this is the rule, correct it. If it happens again, I want to know. If it becomes a pattern, I definitely want to know. And I want that exchange documented: partly because the operator is the one who has to produce the record,¹ partly because the driver should find out what happened from the fleet before finding out from an audit.

That means being explicit with drivers about what the system does. Not "we are tracking you", but: these are the rules the fleet operates under, this is what we have to be able to show, and the same system is there to tell you when something is going wrong while it can still be corrected.

I can disagree with the logic of the rule. I can think it produces kilometres nobody needs, though after reading how that cost is actually estimated, I wouldn't put a number on it.¹⁰ What I can't do, if I am responsible for the operation, is work around the regulation that exists today. So the question becomes how to turn something nobody can supervise by hand into a workflow: collect the right information, detect the exceptions, tell the people involved, document the response, escalate when it matters. Building that means understanding enough regulation to know what I answer for, enough of the platforms because part of my operation runs inside them, enough data to reconstruct what happened and enough technology to do it at scale. Not a lawyer, not a data engineer. Enough to connect the pieces into something that works.

What this case doesn't answer is the platform's part. If a platform generates information that later becomes evidence while the operator carries the record-keeping duty,¹ it seems fair to ask what an operator should be able to access, for how long, in what format, and what happens when the data stops arriving, which is what the Ansbach file records happening, on a summary review that is under appeal.⁴ I don't have the answer and neither does anything I read for this note: no source here establishes an obligation of that kind on the platform, and Uber's own page for prospective operators doesn't mention the duty at all.¹⁵ The same gap runs the other way. Two authorities turned the same duty into two different audits, one of them drawing a 250-metre line the federal statute doesn't contain.⁴⁶ I would want to know what an operator is actually expected to demonstrate. The operator is expected to control an operation that is partly happening inside systems it does not control.

One limit on where I'm standing. What I've seen first-hand is Berlin and Brandenburg, and none of this note rests on it: the argument is built on statutes, court files and official answers anyone can open. That costs the note something specific: I can't tell you how the audit actually runs in the districts where I haven't worked, and this note argues that practice is exactly where the variation lives.⁴⁶

You cannot control every driver. You cannot control the platform. You cannot control the regulation. What you can control is how the operation is designed around them: what becomes visible, what gets detected, what gets communicated, what gets documented, and when a human has to step in.

Sources, with appraisal

1. Personenbeförderungsgesetz § 49 — Verkehr mit Mietomnibussen und mit Mietwagen. Bundesrepublik Deutschland, 2021-08.Statute or rulingAppraisal: The statute itself: after completing a trip a Mietwagen must return to its base unless a new order arrived before departure or during the trip, orders may only be executed if received at the base, and the record may be kept electronically including via an app-based system. This is the rule, not its enforcement: it does not settle how local authorities read 'during the trip' or what happens in practice when an order lands mid-route.
2. Rückkehrgebot für Mietwagen, BVerfGE 81, 70 (1 BvL 14/85, 1 BvR 1276/84). Bundesverfassungsgericht, Erster Senat, 1989-11.Statute or rulingAppraisal: Holds the return obligation compatible with the constitution only as construed in its reasons, and imposes the reading that a rental car may accept radio-dispatched orders even during the return leg and break off that return to serve them, so the window stays open all the way back. It also states the purpose is not the return itself but preventing taxi-like standing, and that merely remaining parked already indicates a breach. Its technical premise is radio dispatch in 1989, so it does not settle what counts as 'during the trip' inside an automated assignment system. It is also the source for this note's dates: the decision records that the Fifth Act amending the Passenger Transport Act of 25 February 1983 (BGBl. I p. 196) appeared in issue 8 of the 1983 Bundesgesetzblatt Part I, whose stated day of issue is 1 March 1983, and that the act entered into force on 1 October 1983.
3. Urteil vom 3. Juni 2026, I ZR 123/25 (ECLI:DE:BGH:2026:030626UIZR123.25.0). Bundesgerichtshof, I. Zivilsenat, 2026-06.Statute or rulingAppraisal: The full judgment, not the press release. It records that at 10:13 a board member of the plaintiff cooperative who was watching the car placed a test order through Uber, that it was accepted at once, and that the confirmation named the defendant as the carrier, which is what ties the waiting vehicle to the company sued. It confirms the injunction for breach of the return obligation and holds that in a purely national case the duty is not measured against EU law, while stating that the appeal court's reasoning on freedom of establishment does not survive review. It decides this dispute on these facts: it does not establish how often the duty is breached, and it leaves open how the duty would fare in a case with a cross-border element.
4. Beschluss AN 10 S 26.536 — Widerruf der Mietwagengenehmigung wegen systematischer Verstöße gegen Rückkehr-, Vermittlungs- und Mitwirkungspflichten. Verwaltungsgericht Ansbach, 10. Kammer, 2026-07.Statute or rulingAppraisal: Attributes breaches committed by individual drivers to the operating company's managing director unless they would have occurred even with adequate supervision and organisation, and records how the authority counted them: 67 breaches across 210 assessable trips by fifteen vehicles on one day, semi-automated from the trip data supplied, treating the duty as met within 250 metres of the base. What lost the operator its licence was failing to report the return timestamp and the follow-on order identifier, not the return itself, and the evidential consequences of missing data fell on the operator. It is interim relief on a summary review and is under appeal in a decision that is not published, so it cannot be cited as settled law.
5. Personenbeförderungsgesetz § 61 (Ordnungswidrigkeiten) und § 25 (Widerruf der Genehmigung). Bundesrepublik Deutschland, 2021-08.Statute or rulingAppraisal: Section 61(1)(3)(h) makes breaching the rental-car rules of section 49(4) an administrative offence, worded against whoever breaches it intentionally or negligently rather than against a named role, while section 25(1) obliges the authority to revoke the licence and states that the operator's required reliability lapses where obligations are breached in his undertaking despite written warning. Together they establish that the severe consequence attaches to the operator rather than the driver; they are the provisions themselves and do not settle who is liable in a concrete case.
6. Antwort auf die Schriftliche Anfrage Nr. 19/16179: Konzessionierung im Berliner Mietwagengewerbe (II). Senatsverwaltung für Mobilität, Verkehr, Klimaschutz und Umwelt, Berlin, 2023-08.Statute or rulingAppraisal: Official Senate answer documenting how Berlin enforces the return obligation: operators using app-based dispatch must hand over the ride data as unaltered original electronic files for every licensed vehicle, and the authority checks compliance with the return obligation and shift lengths against those records, by sampling or where discrepancies are obvious. It also states that results of these audits and the grounds for licence revocations are not recorded statistically and that the Senate does not intend to start recording them. It covers Berlin in 2023 and its audit counts are aggregated across taxi, private hire and patient transport, so it cannot quantify how often the obligation is actually breached.
7. A timely match for ride-hailing and ride-pooling services using a deep reinforcement learning approach. Bao, Y.; Gao, J.; He, J.; Oliehoek, F. A.; Cats, O., 2026-01.Peer-reviewed researchAppraisal: Models when a platform should trigger matching, comparing learned timing against fixed-interval batching and first dispatch on a simulator calibrated with New York trip records: reported gains are 3.1% less waiting in ride-hailing and 20.1% in pooling. It is calibrated simulation, not a deployed system, and it optimises waiting and detour only. It says nothing about legal constraints on when a vehicle may take the next order, which is what this note argues about.
8. Übersicht über die Anzahl der konzessionierten Unternehmen und Fahrzeuge im Gelegenheitsverkehr mit Taxen und Mietwagen. Landesamt für Bürger- und Ordnungsangelegenheiten Berlin, 2026-03.Official statisticsAppraisal: Monthly count of licensed companies and vehicles in Berlin: between March 2025 and March 2026 licensed taxi vehicles rose from 5,680 to 6,941 while Mietwagen vehicles fell from 2,335 to 1,656 and Mietwagen companies from 436 to 304. It counts licences, not activity or trips, and it states no cause whatsoever: reading the decline as market contraction or as administrative enforcement would be an inference this table does not support.
9. Personenverkehr — amtliche Statistik. Statistisches Bundesamt (Destatis), 2026-09.Official statisticsAppraisal: Cited for an absence: German official passenger-transport statistics cover rail, tram and bus, and do not include taxi or Mietwagen, which is why national figures for the sector come from trade associations or data aggregators instead. It establishes that this particular statistic excludes the sector; it does not establish that no German official data on the trade exists through other collections such as service or VAT statistics.
10. Transformation des Taxi- und Mietwagengewerbes zur Emissionsfreiheit. Tschinibaew, I.; Grote, R.; Kwella, B.; Maiwald, F.; Pontow, J.; Schrab, K.; Schweppenhäuser, M.; Protzmann, R.; Massow, K.; Radusch, I. (Fraunhofer FOKUS), 2024-05.Industry reportAppraisal: Commissioned by wirfahren, the operators' association whose speaker is the managing director of SafeDriver Group, the group whose Ennoo subsidiary took on the role of Uber's general contractor in Germany, and it is the origin of the figures used to argue against the return obligation. Read in the original it reports 28% of distance caused by the obligation and 194,064 kilometres a day for Berlin, not the 30% and 200,000 that circulate in press releases. Its simulation covers 50 vehicles on two typical days extrapolated to the city fleet, it attributes the 28% to a study it describes as still unpublished, and its counterfactual assumes drivers wait for the next order at the drop-off point, which is the very conduct the statute forbids.
11. Registereintrag Bundesverband wirfahren, R003822. Lobbyregister beim Deutschen Bundestag, 2026-08.Statute or rulingAppraisal: Official lobby-register entry naming Thomas Mohnke as speaker of the association and declaring annual interest-representation spending of 1 to 10,000 euros with 0.20 full-time equivalents, funded by economic activity and not exercised on behalf of third parties. Neither Uber nor SafeDriver appears anywhere in the entry, so it establishes who speaks for the association and what it declares, and cannot by itself establish the commercial relationship that has to be documented elsewhere.
12. Impressum / Legal notice. SafeDriver Group GmbH, 2026-09.Primary company documentAppraisal: The company's own legal notice, naming Thomas Mohnke as managing director of SafeDriver Group GmbH, registered at HRB 188652 B, Amtsgericht Berlin Charlottenburg, together with four further group companies at the same Berlin address. It establishes who runs the company and its register details, and nothing more: the notice does not mention Uber anywhere, so the contractual relationship has to be established from another source.
13. Safedriver könnte der große Profiteur des Uber-Streits werden. Kapalschinski, C., Handelsblatt, 2019-12.PressAppraisal: General business press, not trade press, reporting that the Ennoo subsidiary took on the role of nationwide general contractor between Uber and other operators, that SafeDriver formally sets the prices using Uber's technology, and that operators who previously drove for Uber became subcontractors of SafeDriver Ennoo. It dates from December 2019, so it supports how the structure was set up and identifies the founder, but it cannot show that the arrangement still stands today.
14. FAZ-Interview mit Thomas Mohnke: Wie der deutsche Flickenteppich den Mietwagenmarkt lähmt. Frankfurter Allgemeine Zeitung, interview with Thomas Mohnke (wirfahren), 2025-03.PressAppraisal: The industry's own account of enforcement: the head of the operators' association states that road traffic offices checking licences do verify compliance with the return obligation and that every operator keeps an order book recording in real time where vehicles are. It is a claim by an interested party, consulted in the reproduction hosted on that party's own website because the newspaper original is paywalled, and the 30% it cites is higher than the 28% in the study it commissioned.
15. Mit Uber Mietwagenunternehmer in Deutschland werden. Uber Germany, 2026-09.Vendor materialAppraisal: Cited for an absence: the platform's own page addressed to the operators who carry the statutory record-keeping duty describes signup, vehicle requirements and the supplier portal, and says nothing about the return obligation, the order book, what data the platform hands over, or any app feature supporting compliance. Being commercial copy, its silence cannot show that no such feature exists or that no data is provided, since the Ansbach file is built on platform data; it shows only that the platform does not document that role where it speaks to those who bear the duty.

All field notes