Platforms that have carried for years

Starting point

For a vehicle manufacturer, telematics has long ceased to be an accessory. It is the foundation of the business: fleet operators expect to see their vehicles in real time, to document cold chains without gaps and to prevent breakdowns before they happen. Our customer’s existing on-premise architecture had been built for a smaller fleet and was now visibly reaching its limits. Every newly connected vehicle increased the load, scaling meant hardware procurement instead of the push of a button, and the stability risks grew with every quarter. At the same time it was clear: the future of the business lies in data-driven services, and that requires a foundation that not only carries today’s operations but enables what is coming.

The requirement was therefore more demanding than a classic migration: the platform was to be rebuilt cloud-native, take over the existing fleet without service interruption, and be designed from the start as the digital core for new business models.

The solution

We rebuilt the platform completely as a distributed, horizontally scalable cloud architecture and migrated the productive fleet in just 14 months. Instead of copying the old architecture into the cloud ("lift and shift"), we rethought it: event streaming as the backbone for message processing, container orchestration for elastic scaling, specialized stores for analyzing large volumes of sensor data. The entire infrastructure is described as code, so that every environment is reproducible and every change traceable.

Today the platform processes the telemetry of more than 150,000 actively connected vehicles, across manufacturers: it integrates not only our customer’s vehicles but also third-party makes, regardless of the telematics generation installed. Each vehicle delivers more than 600 different sensor values, from position and temperature to tire pressure, door status and braking data. In normal operation the system processes 250 to 350 messages per second and reliably absorbs load peaks of more than 1,000 messages per second.

On top of this, a portal emerged for fleet operators that goes far beyond position display: real-time visibility of the entire fleet regardless of hardware, freely configurable alarms and notifications, complete histories of events, positions and temperatures. At order level, tours can be monitored in real time, including arrival forecasts and automatically generated temperature reports as required by the cold chain. A monitoring module combines proactive surveillance with predictive maintenance, with particular attention to the cooling units: anomalies are detected before they become failures, maintenance is planned ahead instead of improvised in the event of a breakdown.

Communication runs in both directions. The platform not only receives data but commands functions such as the cooling unit or door control remotely and rolls out over-the-air updates through the same channel. A dedicated update management system plans and documents these campaigns in line with the regulatory requirements for software updates in vehicles: every update is traceable, every vehicle remains compliant, without anyone having to intervene physically.

And the platform has become the business foundation it was meant to be: for its users we brought an AI-powered assistant into production that answers questions about vehicles and their data in natural language, coupled to the actual data context.

The result

Since the start in the cloud, the number of connected vehicles has grown from around 30,000 to more than 150,000 — without an architecture rebuild, because the design carries up to more than 600,000 vehicles. Today the platform carries the telematics operation of one of Europe’s largest connected trailer fleets. And it is not a finished project: as part of our platform stewardship we have been securing, operating and continuously evolving the system since go-live: with the same team, monitored around the clock, with automated load and integration tests in every release.

Starting point

It sounds like a detail but is a problem with many dimensions: for every vehicle produced, the right combination of tires, wheels and valves must be determined, down to the specific complete wheel. What seems trivial is in truth a high-dimensional optimization problem, because the selection follows competing criteria: costs argue for one combination, the efficiency values relevant to CO₂ regulation for another, contractually guaranteed supply shares and supplier capacity limits for a third. Whoever decides by one criterion alone inevitably violates another.

Until now this assignment was made largely manually: a laborious step that tied up expert capacity, took hours, and where nobody could prove that the result was really the best across all objectives. The customer was looking for a solution that performs this assignment fully automatically and verifiably optimized and fits seamlessly into the existing configuration process instead of rebuilding it.

The solution

Classic rule-based automation fails at such problems: the number of possible combinations is too large to try out, and rigid rules cannot balance competing objectives. We therefore implemented an evolutionary optimization method, an approach borrowed from nature: the system evaluates an assignment via a score function, modifies it in a rule-based way and keeps only the changes that improve the overall result. Over many generations it thus approaches a solution that is very good across all objectives: not perfect in one criterion at the expense of all others, but optimal as a whole.

The selection is multi-stage: first the tires, then wheels and valves, finally the specific complete wheel. It considers commercial, ecological, technical and contractual criteria simultaneously. Several weighted objectives can be specified for each component; the weighting is configurable, so changing priorities can be reflected without redevelopment. In addition, the system distributes quantities across time blocks as evenly as possible among suppliers to avoid clustering effects in procurement.

Honesty of the system was important to us: unresolvable assignments or exceeded capacity limits are not silently solved "somehow" but reported back in a structured way via an alert function. The decision on how to handle them stays with humans. Simulation runs also allow alternative objectives and time-period splits to be played through before real scheduling: what does it cost to weight the efficiency targets higher? What happens if a supplier fails? A BI dashboard makes the effects of every optimization immediately visible: metrics such as average efficiency values and costs before and after each run.

The result

A manual process that took hours and tied up expert capacity became an automated, traceable run of under five minutes, mathematically optimized across 15,000 orders and more than 300 materials, with guaranteed compliance with contractual supply shares and capacity limits. The effect of the optimization is not a claim but a measurement: the dashboard reports it for every run.

Starting point

A globally active sales organization has a structural problem: the knowledge of the best sales experts lives in heads and scattered file shares: which arguments work, which documents convince, which options fit the customer. Every region presents differently, documents become outdated unnoticed, and regional specifics such as winter equipment in northern or sun protection in southern markets have to be gathered anew every time. On top of that comes a very practical hurdle: sales conversations happen where the customer is: in depots, in factory halls, in regions without reliable network connection.

The customer was looking for a globally uniform tool with which sales staff can compile and present individual showcases for every customer: multilingual, regionally appropriate, centrally maintained and yet ready for use anywhere, including offline.

The solution

We developed a guided-selling application that makes the pooled knowledge of the best sales experts available in a structured way. At its core is a flexible presenter that prepares the right content in categories, from general product information through flyers, brochures and value argumentation to photos and videos. Products, campaigns and the associated documents are managed centrally and can be configured per customer situation: the salesperson assembles the individual presentation from verified, up-to-date building blocks instead of reinventing it every time.

The foundation is a central media asset management: marketing images, graphics, documents and videos are versioned, kept multilingual and can be found precisely via individual keywords. An administration interface covers product management, the assignment of data and texts, multilingualism as well as roles and permissions. Maintenance stays central, use stays global.

The real technical challenge was offline operation. We designed the application consistently local-first: required content is checked out in advance and provisioned locally on the device; a local database makes the application fully functional without any connection. A powerful synchronization mechanism distributes subscribed content to the clients and keeps it current as soon as a connection exists again. Users notice nothing; their documents are simply there and up to date. A responsive design makes this work from notebook to tablet, in the meeting room as in the depot.

The result

Worldwide sales received a uniform tool to address customers individually, multilingually and regionally appropriately, online and offline, on any device. High-quality, always up-to-date sales documents are available anywhere in the world; the knowledge of the best sales experts arrives in every conversation, not just those of the most experienced.

Starting point

Anyone who wants to control who may access what in an industrial company with thousands of employees fights on many fronts at once: joiners, leavers and internal moves must reliably translate into permissions; devices and software need to be managed, ordered and assigned; and in the end, IT costs should be allocated to the divisions according to cause. At our customer, these processes ran through scattered individual channels: emails, shouts, forms. This created effort and cost traceability. Particularly critical: fully ensuring that all access is revoked when employees leave is practically impossible via individual channels, a risk every auditor finds immediately.

What was needed was a central solution that controls identities, permissions and the associated IT operating processes across the entire company, from an employee’s first working day to the last.

The solution

The system we developed controls the on- and offboarding processes end to end: on joining, access, rights and equipment are set up reliably and traceably; on leaving, they are completely revoked again. What sounds like a matter of course is rarely solved cleanly in grown organizations and is one of the most important compliance building blocks of all: no orphaned access remains, every permission has a documented reason and a documented end.

A central building block is the self-service functions that give all employees structured access to the common IT procedures: ordering software and devices, requesting permissions for systems, folders and mailboxes, each via configurable approval paths that adapt to the organization instead of the other way around. Requests run in a structured and traceable way, the IT organization is relieved of routine procedures and can concentrate on what requires judgment.

In addition, the system manages the company’s assets, both devices and software, and maps the company-internal recharging of IT services fully automatically, including reporting: every division sees which IT services it causes, and cost allocation runs in the background without anyone maintaining lists.

The result

The customer has central control over identities, permissions and IT operating processes: less manual effort, higher traceability, cause-based costs and a permission system that withstands audits. After implementation we handed the system over to the customer’s internal development in an orderly manner, including documentation and knowledge transfer. They have been responsible for operation and further development ever since. That, too, is part of our understanding of partnership: we do not tie customers to us, we enable them.

Starting point

At the end of the nineties, "IoT" was not yet a term, but the problem already existed: an energy supplier operates thousands of industrial plant components distributed across a large network, and the operation of this infrastructure ran on grown isolated solutions and decentralized shadow IT. Faults were discovered when they took effect; operational processes such as maintenance, incident management and asset administration ran side by side instead of into one another. For an infrastructure on which the heat supply of entire regions depends, this was not a sustainable state in the long run.

Standard software failed at the extreme requirements: at the quantity and variety of the plants, at the depth of integration, at the governance requirements of critical infrastructure. So the platform was created as a custom development: from the ground up, together with the customer’s business units.

The solution

The platform manages thousands of heat and energy components with millions of sensors across the entire network. It processes telemetry and sensor data from generation and distribution in real time and makes the condition of the plants visible at all times, from the individual unit to the overall network. On this data basis, the system detects anomalies, supports predictive maintenance and, in case of irregularities, triggers automated incident management: a measured value becomes a fault report, the fault report becomes a dispatched deployment, without media discontinuity.

Because the platform is far more than pure IoT: it handles the operational business processes end to end: from project planning through maintenance and incident management, plant and network administration and meter management to the administration of hazardous substances. Only this combination of machine data and business processes turns measurements into decisions and decisions into actions.

A substantial part of the task lay in integration: via hundreds of interfaces, the platform connects the existing systems and plants, including the secure integration of control systems via established standards of the automation world such as OPC UA. For governance reasons and with a view to the requirements for critical infrastructure (KRITIS), the solution is deliberately hybrid: cloud services combined with operation in dedicated environments where data sovereignty and regulation demand it. A high-availability architecture with cross-site redundancy and comprehensive backup strategies ensures consistently reliable operation. For critical infrastructure that is not a nice-to-have but a duty.

And then comes the dimension that makes this reference special: time. For 24 years we operated and continuously evolved the platform as platform stewards, through several technology generations, changing market conditions and steadily growing regulatory requirements. What began in 1999 was never "finished" and never became legacy: it grew along.

The result

An infrastructure that contributes to the heat and energy supply for millions of people: real-time monitoring, predictive maintenance, operational process handling and end-to-end integration in one solution, under the strict governance requirements of critical infrastructure. In 2024 we handed the platform over to the customer’s internal IT in an orderly manner, with complete documentation and structured knowledge transfer. A quarter century of partnership that remained professional even in parting: perhaps the strongest proof of what we understand by platform stewardship.

Starting point

In the international trade in marine fuels, risk assessment is not an administrative task but a question of existence: every bunkering is a credit decision, every portfolio changes by the minute, and the counterparties sit distributed across the globe. Our customer’s existing order and risk management system had grown over years and was no longer up to this business: risk assessments were produced in nightly batch runs: whoever looked at the figures in the morning saw yesterday’s state. In a time-critical, globally distributed business, that is not just inconvenient but dangerous.

Added to this was the classic dilemma of the irreplaceable: the system was business-critical, deeply woven into the workflows, and yet could not stay as it was. Switching off impossible, continued operation risky. Exactly the situation we are built for.

The solution

We redeveloped the platform completely and reversed the architectural principle. Instead of collecting data and settling it at night, the new system is consistently event-driven: every new piece of information, whether an order, a payment or a limit change, immediately triggers exactly the calculations and process steps affected by it. At the core of the platform is a risk calculation engine that processes incoming financial data in real time and immediately recalculates the risk assessment. Changes in the portfolio thus show up in the risk figures within seconds instead of the next morning. For an annual risk volume of more than one billion euros, this means the assessment is reliable at any time, not just once a day.

Beyond risk assessment, the platform maps the entire order management, from capture to settlement. The associated decision processes are configurable: approval paths and rules can be adapted to the requirements of different markets and business situations without code changes. The domain knowledge about processes thus lives where it belongs: with the business units, not in the source code.

The most delicate phase was the replacement itself. A system through which live trading is handled cannot be stopped for a changeover. We therefore executed the migration as a controlled cutover during live operation, prepared by parallel operation and comparison tests until the new system had proven itself. Trading continued without interruption; a high-availability architecture with hot standby has secured uninterrupted operation ever since.

The result

More than 1,000 users in over 100 countries work on the platform today, which processes more than 100,000 bunkering transactions annually. Nightly batch runs became risk figures in seconds; a legacy system nobody wanted to touch anymore became a modern, event-driven platform that can be evolved. And the proof is delivered that the irreplaceable can be replaced — during live operation, without stopping the business.