Your installed base is not an asset register. It is a demand forecast.
Your installed base is not an asset register. It is a demand forecast.
The most useful number in one of our recent filtration mandates was not about our client's equipment at all. It was the number of a competitor's filter units in service across alumina refining worldwide.
That gap was not a market-share statistic. It was a service revenue forecast, and it belonged to somebody else.
We reached that number the same way you would reach any forecast: by building the installed base from the bottom up. Every site, every process line, every filter type, every unit, every unit's capacity. Then dividing by equipment lifetime to get replacement demand, and multiplying by maintenance intensity to get parts and service demand.
From asset register to demand forecast
The conversion chain in a bottom-up installed base build.
Across sixteen end applications, that build produced roughly €720 million of annual capital and operating spend, and it showed value was concentrated in silica, nickel, copper and calcium carbonate mostly because of large installed bases, not because of growth. By 2030, replacement accounted for 85% of serviceable value, and 60–70% of annual demand in the largest application came from replacement rather than new capacity.
Sizing the base you don't own
The competing filter technology against our client's, in the same application.
Nothing about that analysis required a new market. It required knowing what was already in the field, including the part of the field that belongs to a competitor.
Two findings fell out of that build that no equipment forecast would have produced. The first is how much of the prize is recurring rather than one-off.
In the largest markets, most of the money is recurring
Split of each end-application's annual opportunity between the equipment sale and the parts it will consume. Eight of sixteen applications shown; labels withheld.
The second finding is the one that reframes the whole exercise. We plotted every application's annual opportunity against its expected growth rate, and the two largest markets in the study were growing at zero per cent.
The biggest markets are the ones that stopped growing
Annual opportunity (indexed, largest = 100) against expected CAGR to 2030. Absolute values withheld.
The register and the forecast are different documents
Most OEMs hold an installed base record of some kind. Serial numbers, ship-to addresses, commissioning dates, warranty status. It is maintained because finance and legal need it, and it is structured the way an asset register is structured: one row per machine, accurate at the moment of dispatch, decaying quietly ever after.
A demand forecast is a different object built from the same rows. It asks what each unit will consume (which wear parts, at what interval, under what duty cycle, at what point in its life) and it aggregates that into a number a commercial team can be held to.
The distance between the two is where the aftermarket leaks. You can have a complete asset register and still have no idea which two hundred customers are due for a rebuild next quarter.
Scale does not produce visibility. Scale destroys it.
The register decays faster than most OEMs assume, and the pattern is counterintuitive: the larger the OEM, the worse the visibility.
Visibility decays as the base grows
Typical data accuracy and annual decay, by OEM size band.
| OEM revenue | Machines in base | Systems involved | Data accuracy | Annual decay |
|---|---|---|---|---|
| < €50m | < 5,000 | 2–3 | 65–75% | 8–12% |
| €50–250m | 5,000–25,000 | 4–6 | 35–50% | 18–25% |
| > €250m | > 25,000 | 7+ | < 25% | 28–40% |
Around 90% of OEMs report problems with their installed base data, and for most of them the critical customer and asset records sit across three to four separate enterprise systems. Scale does not produce visibility. It destroys it, unless someone is explicitly accountable for preventing that.
You may not even be able to see your own channel
There is a harder version of this problem, and we ran into it while studying a distributor return policy for a global minerals and aggregates manufacturer.
The client treated every distributor purchase order as end-customer demand. It is a reasonable default. It is also wrong, and the cost of being wrong compounds in both directions. When a distributor over-orders, the OEM reads it as demand and builds more. When the distributor then asks to return the surplus, the OEM refuses, because its own warehouse is already overstocked on that part, for exactly the reason the distributor is.
When you say “sorry, I have three extra to return,” we are actually piled up with three extra too. This vicious circle is because we do not have visibility. Head of Global Inventory Network, global minerals & aggregates manufacturer
That is not a returns problem. Returns is where the problem becomes visible. The problem is that a channel-level demand signal was being read as an end-customer demand signal, in a business where roughly 95 distributors sit between the factory and the machine.
The commercial consequence was measurable. At one large distributor, only about 17–20% of requested return items were approved, even on a return request worth more than $1 million. Returnability was not governed by a policy the distributor could plan against; it was governed by whatever the OEM's own stock position happened to be that week. Every rejected return teaches a distributor to order more cautiously next time, which degrades the demand signal further.
We modelled the way out, and the arithmetic was unexpectedly forgiving. With a 10% returnability cap and a 10% restocking fee, the manufacturer only needed roughly 0.4% of returned value to be replenished through the channel to break even. The barrier to a predictable, published return policy was never the economics. It was that nobody could see enough to be confident in them.
We also timed the process, function by function, across a year of returns. What the clock showed was that the deadlock is not adjudication, it is administration.
Where the time actually goes
Observed hours per return by function, against the modelled position once the process is automated. Based on 106 returns in the base year.
The removed effort was worth roughly €200,000 a year against a first-year build cost of about €160,000, a nine-month payback on what everyone involved had been describing as a policy dispute.
What actually changes when you build the forecast
Three things, in our experience.
- The commercial target list stops being a guess. Once the installed base is expressed as consumption rather than as machines, you can rank customers by what they are due to spend rather than by what they spent last year. The two lists are not the same list, and the second one is the reason field sales calls the wrong accounts.
- You can size what you are losing, not just what you are earning. The 3,827-versus-76 number above is only visible if you build the competitor's installed base alongside your own. Most OEMs never do, so the leakage stays invisible, and an invisible loss never gets a budget line.
- Pricing gets a denominator. It is very difficult to argue for a price premium on a wear part without knowing how many of that part are consumed annually across the field, by whom, and against which alternative. Cost-plus pricing survives in roughly three-quarters of the industry partly because it is the only method that does not require this data.
The uncomfortable part: nobody owns it
Building this is not a systems project, and treating it as one is how it stalls. In our filtration work, the installed base model was assembled from public capacity data, technical parameters and structured interviews, not from a data warehouse. It took weeks, not a transformation programme.
The reason it usually does not get built is more mundane. Sales owns customers, service owns work orders, supply chain owns parts, and the installed base is the one asset that sits between all three and belongs to none of them. So it stays an asset register: accurate at dispatch, decaying ever after, and quietly setting a ceiling on how much of your own aftermarket you are able to ask for.
There is a second, sharper version of that ceiling, the revenue you are already delivering and never invoicing, and the service agreements your customers say they were simply never offered. That is the subject of part two.
SprintlyWorks runs eight-to-ten week sprints for industrial OEMs on installed base visibility, aftermarket capture and service channel design. The installed base build described above took weeks, and ended in a model the client's commercial team owns, not a recommendation deck.
This is part one of a three-part series on aftermarket and installed base.
