The CFO asks: “What is Müller GmbH worth to us in the long term?” Three minutes later comes the answer, based on gut feeling and the most recent order history. Yet all the data is there: in CRM, ERP, service system, Siemens Insights Hub. The installed base is documented, every maintenance logged, every spare parts order recorded. Nobody has connected them into a lifetime view. In mechanical engineering, Customer Lifetime Value has a special characteristic: A machine runs for 15 to 50 years. During this time, the customer buys spare parts, service, upgrades, training and follow-up machines. This lifetime perspective requires linking the installed base with customer master data, history and external market data, plus a model that predicts which customer generates which value.
Why classic CLV models fall short in mechanical engineering
B2C CLV follows simple logic: order frequency times average order value times average customer lifespan. In mechanical engineering, these models fall short because they don’t capture the complexity of capital goods.
A machine tool isn’t replaced after three years, but modernized after 20 years. A customer doesn’t buy twice a year, but every five to ten years, but for six-figure amounts. Between these new machine cycles lie 15 to 50 years of machine life, during which aftermarket revenue is continuously generated. According to analyses by IMPULS Consulting, mechanical engineers generate 20 to 25 percent of their total revenue with after-sales service, with profit margins of 15 to 25 percent, significantly above the 2 to 5 percent net margin in new machine business.
Lifetime value in mechanical engineering is not a linear value, but a sum of several lifecycle phases: new machine cycle, aftermarket (spare parts, maintenance), upgrades and modernization, and training and software subscriptions. The challenge is not that the data is missing. The challenge is that it sits in silos, in CRM, ERP, service system, Insights Hub, and nobody connects it into a predictable lifetime view.
Four value sources per customer in the equipment lifecycle
A customer who buys a CNC machine today for €250,000 generates over 20 years of machine life not just this purchase price, but a multiple through four parallel value streams:
1. New machine cycle: The probability that the customer will buy a follow-up machine in five to ten years is 30 to 60 percent, depending on industry and utilization. The installed base is the best predictor: those who have three machines in operation statistically buy more frequently than someone with one.
2. Aftermarket (spare parts and maintenance): About 40 percent of mechanical engineering companies report spare parts revenue exceeding 60 percent of total revenue. Annual after-sales revenue per machine is approximately 3 to 6 percent of the acquisition value, for a €250,000 machine that’s €7,500 to €15,000 per year, extrapolated over 20 years that’s €150,000 to €300,000.
3. Upgrades and modernization: After ten to 15 years, modernization is often due: new controls, expansion modules, retrofit to Industry 4.0 capability. These projects often reach five-figure sums and extend machine life by another ten years.
4. Training, consulting, software subscriptions: Service contracts, operator training, software licenses for predictive maintenance or remote monitoring. These value streams are smaller but high-margin and scale with the installed base.
A customer with one machine generates a CLV of €400,000 to €600,000 over 20 years, two to three times the new machine price. Those who don’t model these four value sources underestimate lifetime value by a factor of two to three.
Data foundation for predictive CLV: From installed base to forecast
Predictive CLV in mechanical engineering stands and falls with the quality and linking of data sources. The central foundation is the installed base: Which machines are in operation at which customer, since when, with what utilization, which service contracts?
Three additional data sources are added:
Order history (CRM + ERP): Which machines has the customer bought in the last ten years? Which spare parts ordered? Which upgrades performed? The order history provides the pattern: customers who regularly book maintenance have a higher CLV than those who only call reactively during breakdowns.
Service history (service system, Insights Hub): How often was the machine serviced? Which spare parts were installed? Which malfunctions occurred? Service history is the best indicator of installed base condition, and thus of the probability of a follow-up order or upgrade.
External market data (Dun & Bradstreet, Destatis, industry associations): How is the customer’s industry developing? Which investment cycles are typical? Which regulatory changes are pending (e.g., compliance pressure, energy efficiency requirements)? External market data provides the context to model CLV per customer and segment.
An AI model predicts the value per lifecycle phase per customer based on this data. The model learns from history: Which customer characteristics (industry, company size, utilization, service frequency) correlate with high aftermarket revenue? Which installed base constellations lead to follow-up purchases? Predictive CLV models in B2B combine account characteristics, product usage, service history and external market data into a forecast that is significantly more precise than historical average values.
The difference between historical and predictive CLV: Historical CLV simply sums all previous contribution margins of a customer. Predictive CLV models what’s to come, and considers that not every customer is equal. A customer with three machines, regular maintenance and a growing industry has a different CLV than a customer with one machine, sporadic service contacts and a shrinking industry.
CLV as a management tool: From metric to strategy
Predictive CLV is not a metric for reporting, but a management tool for operational decisions. Three use cases show how CLV changes sales strategy:
KAM prioritization by lifetime value instead of current pipeline: The classic key account manager prioritizes by current opportunity size. The CLV-based KAM prioritizes by expected lifetime value. A customer with a small current pipeline but three machines in the installed base and high aftermarket frequency gets more attention than a customer with a large pipeline but only one machine and low service frequency.
Investment decisions per account: How much discount is justified to win a customer? How much effort makes sense to retain an existing customer? CLV provides the answer: A customer with a projected lifetime value of €500,000 justifies €20,000 acquisition costs. A customer with €100,000 CLV doesn’t. In the B2B sector, CLVs can reach millions per account, putting investment decisions on a new foundation.
Service level differentiation by expected customer value: Not every customer gets the same service. Customers with high CLV get 4-hour response times, priority support, proactive maintenance offers. Customers with low CLV get standard service. This sounds harsh but is economically rational: According to the Pareto principle, 20 percent of customers generate 80 percent of profit, resources should be distributed accordingly.
Another use case: Churn prevention. If a customer with high CLV suddenly stops ordering spare parts or doesn’t renew service contracts, that’s a warning signal. A Customer Health Score based on CLV forecasting enables proactive intervention before the customer switches to competitors or decommissions the machine.
Where the MARINI platform comes in: From silos to lifetime view
MARINI, the platform for Customer Intelligence with Data Integration, Data Cloud and Agentic, solves the three central challenges in implementing predictive CLV in mechanical engineering:
1. Data integration across the entire system landscape: The SAP S/4HANA integration brings order history and installed base data into the MARINI DataEngine. The Salesforce integration delivers CRM data and opportunity history. The Siemens Insights Hub connection provides service history and machine condition data. Dun & Bradstreet delivers external market data and credit information. Everything lands in the DataEngine as a single source of truth for the lifetime view.
2. Data quality and Golden Records: The Data Cloud deduplicates customer master data, creates Golden Records across system boundaries and links accounts with their installed base. Enterprise account hierarchies are resolved: Which machines belong to which parent account? Which subsidiaries buy autonomously, which centrally? This data foundation is the prerequisite for valid CLV forecasts.
3. AI Agents for predictive modeling: The Agentic phase brings AI models directly to customer data: An AI agent classifies customers by projected CLV. Another agent identifies churn risks. A third agent suggests cross-sell and upsell opportunities based on the installed base. AI Agents in sales use the Model Context Protocol (MCP) to access customer data in natural language: “Show me all customers with CLV over €500,000 and expiring service contracts in Q3.”
MARINI Professional Services accompany the implementation: CIEF-based roadmap development, AI-supported data cleansing during migration, building customer-specific AI workflows for CLV forecasts. The result is a platform that not only calculates CLV but makes it operationally usable, as a foundation for KAM prioritization, account planning and service strategy.
Mechanical engineering needs a lifetime model, not a transaction model
A machine runs for 50 years. The customer buys spare parts, service, upgrades, follow-up machines. Classic CLV models from B2C fall short, mechanical engineering needs a lifetime model across four value sources: new machine cycle, aftermarket, upgrades, training.
The data foundation already exists: installed base, order history, service history, external market data. The challenge is not data availability, but connecting it into a predictable lifetime view. Companies that establish this connection have a measurable advantage: they prioritize the right accounts, invest in the right customer relationships and systematically exploit their aftermarket potential.
Predictive CLV is not a nice-to-have, but the foundation for data-driven account planning in B2B sales. Semiconductor companies like ASML expect annual growth of 12 percent in installed base revenue, not through more new machines, but through systematic monetization of the existing machine base. Those who still prioritize by gut feeling today are giving away contribution margin.
