Industrial Service Portfolio Design: What Customers Buy

Somewhere in a customer’s plant there is a machine you built. You were paid once, on the day it shipped. Somebody is being paid every month to keep it running — and there is a fair chance that somebody now earns more from your machine than you did.

Otis publishes the arithmetic, because it reports its two businesses separately. In 2025, new equipment produced 9% of segment profit and service produced 91%. On revenue the split is 35% and 65%. The profit split is far more extreme than the revenue split, which is the part worth pausing on: the service business is not simply bigger, it is structurally better.

Lifts, filling lines, compressors, turbines, robots, crushers — the pattern is the same wherever a machine runs for twenty years inside somebody else’s operation. The equipment is the entry ticket. The relationship afterwards is the business.

Most industrial companies have already bought the sensors, the connectivity and the platform. Very few have redesigned what they actually sell. That gap is not a technology problem, and it will not be closed by another capital programme.

1. What those numbers mean for the customer

Behind the split sits a maintenance portfolio of roughly 2.5 million units and about 37,000 service mechanics, with 1.1 million units connected at the end of 2025. That is not a software story. It is a very large field organisation with data underneath it.

Two figures from the fourth quarter show where this goes. Modernisation orders rose 43% at constant currency with backlog up 30%, and management stated that modernisation margins now exceed new equipment margins. Upgrading a machine already in service is more profitable than selling a new one.

Loyalty appears in the same accounts, expressed the way industrial customers actually behave. The maintenance portfolio grew 4%. Nobody in this market completes a satisfaction survey. A customer who is unhappy signs with an independent maintainer at renewal, quietly, and you find out from the revenue line. Portfolio retention is the experience metric.

2. Four decisions that have nothing to do with technology

In December 2025 Samsung Heavy Industries and its customer Evergreen opened a joint monitoring centre — inside Evergreen’s own headquarters in Taipei. Most manufacturers would have built a portal and issued a login. Samsung moved people and responsibility into the customer’s building.

Look at how the offer around it is assembled. The SVESSEL platform sells fleet management as a cloud service, a separate warranty-period service, troubleshooting alongside partners, and a public API so others can build on top of it.

Four decisions: where the service physically sits, what happens during warranty as distinct from after it, who else may build on your platform, and who certifies your claim.

I have sat through a great many service reviews on three continents. I have never once been in a room where anyone asked where the service should sit. The questions are always which platform, which vendor, which integration. The location of responsibility belongs to no function, and therefore to nobody.

Yet that decision is the customer experience. The shipowner in Taipei does not experience an algorithm. They experience an engineer down the corridor. A plant manager does not experience predictive maintenance; they experience a technician who arrives already knowing what is wrong and carrying the right part. Portfolio design is where industrial customer experience is actually decided — long before anyone writes a service-level agreement.

3. Price it in the customer’s unit, not yours

Rolls-Royce charges TotalCare per flying hour and says plainly that it is only rewarded for engines that perform, reporting overhaul intervals extended by around 25%. Atlas Copco sells compressed air the same way: under AIRPlan the customer pays for the air used while Atlas Copco owns, maintains and upgrades everything down to the piping — 600 installations worldwide by 2024.

Flying hours and cubic metres are not machine metrics. They are the units the customer’s own business is measured in, which is why their finance director can approve the contract without translating anything, and why the deal survives a change of plant manager.

Here is the test: if your services are priced per visit, per part and per machine, you are asking the customer to do that translation themselves. Most will not bother. The ones who do will use the result to negotiate you down.

4. Seven places to find the services worth building

The customer, asked properly. Customers describe problems in the vocabulary of the contract they already hold, so asking what they want returns a request for cheaper parts. Ask instead about their three worst production days of the year. The service is in the story, not the answer.

Your field engineers. They hold the most valuable unstructured knowledge in the company and are almost never debriefed. Every service organisation has engineers quietly doing unpaid work because it is easier than arguing about scope. That unpaid work is a catalogue of your next products.

Your platform data, read as behaviour. It records what people did rather than what they said. An alert nobody acts on is a dead service you are still paying to run.

Sales, for objections rather than the roadmap. Sales will say you lost on price. Treat that as a value gap — and accept that a sales force paid on equipment will not hand you ideas that reduce equipment sales.

Partners and dealers. They see your machines beside your competitors’ in mixed fleets, and they know which of your services are undeliverable in their market.

Adjacent industries. Charging by hours of use crossed from aerospace into tyres and then into compressed air. Borrowing a proven model beats inventing one.

Market research, for sizing rather than invention. It tells you how large a market is. It cannot tell you what to build, because it samples what already exists.

Then decide what to refuse

Not everything a customer asks for should be built. Some services cannot be delivered consistently in every country you serve, and one that works in Germany and fails in Brazil damages you more than not offering it. Some are not good for the customer at all: keeping an ageing machine alive with a cheap subscription can cost more across five years than replacing it, and selling it anyway trades a decade of trust for a small annual fee.

The conversation where you explain what you will not sell does more for a relationship than the three services you did.

5. Five tests before anything reaches a price list

Then sequence what passes. Start with what is easy to sell and quick to earn, because it funds the rest — FANUC reports 1,300 hours of unexpected interruption avoided across 16,000 robots, valued above $40 million. Move next to the hard, transformative offers: ANDRITZ sells Metris OPP as a process-performance contract with guaranteed results, which means selling a performance level and carrying the risk of missing it.

And keep the unglamorous services. At ENGEL, a parts finder that identifies a component from a photograph means a technician spends 50% less time identifying spares. Bühler’s thirty-year account of a Mauritius flour mill lists repeated training of the customer’s staff beside the automation upgrades, not after them. Customers remember those at renewal.

6. The plus and the minus

This is a trade, not a free move. You gain revenue that recurs when capital spending freezes, better margins, a relationship two levels higher in the customer’s organisation, and data that compounds against every competitor.

You also absorb risk that used to be the customer’s — Rolls-Royce’s model is described as transferring cost uncertainty from the airline to the manufacturer. Assets may move onto your balance sheet, since under AIRPlan the installation stays the manufacturer’s property. Revenue recognition slows: a machine books at once, a subscription arrives over years. And you need commercial people who can model risk and engineers comfortable being judged on someone else’s output.

The decision that cannot be delegated

In 2000 Michelin began selling kilometres instead of tyres. Three years later profitability was poor, and one documented reason deserves reading twice: the sales team believed the service undermined their target of selling new tyres. The idea was never wrong. Rebuilt from 2013, it now runs across hundreds of thousands of vehicles, and the XPO relationship alone grew to 13,000 assets.

Every chief executive reading this has the same private question: why has my services director not done this already? I have asked service directors that question directly. The answer is almost never a lack of ideas. It is a version of one sentence — I would be penalised for it before I was rewarded for it. Service leaders are measured on service margin and response time; an outcome offer puts margin at risk for growth that lands elsewhere. That is a compensation decision, and it is where organisations move slower than the technology they bought.

A service portfolio is not a list of what your company can do. It is a statement of what you are prepared to be accountable for, in a customer’s plant, on a Tuesday, when something stops. Otis makes that statement across 2.5 million units and is paid nine tenths of its profit for it. The machines in your installed base are already running. The only open question is who gets paid for the next twenty years of them.