Ask a service director at a machine builder how parts revenue is doing and you get a number. Ask how much parts revenue they are missing, and the room goes quiet.
Which makes sense. You cannot measure what never came in. In capital equipment, figures of 30 to 50% get quoted for the share of parts revenue that goes to third parties: local suppliers, dealers in generic components, the supplier around the corner who can deliver tomorrow. Whether that holds for you is something you only know once you measure it. But that it happens is obvious to anyone who has stood in a machine hall.
How the leak starts
It nearly always starts the same way. A machine has stopped. The technician standing next to it knows the serial number, the year it was built and the noise it made before it failed. What he does not know is the part number of the component that needs replacing.
So a photo goes to service@. Then comes the exchange: is this the one, which configuration, what is the serial number, could you send a photo from the other side. Someone at the service desk digs through the ERP, produces a quote, waits for approval and types the order over by hand. Two days later the part arrives.
Meanwhile a local supplier could have delivered that same bearing set tomorrow. For a bearing set, an O-ring or a sensor, that trade-off takes about two seconds to make. And once made it stays made: next time the technician does not call you at all.
Why your reports do not show it
Your ERP records what you sold, not what you could have sold. Parts revenue that holds steady year over year looks healthy — even when it sits structurally thirty percent below its potential. No report turns red over revenue that was never requested.
The only place the leak becomes visible is your installed base. You know how many machines are out in the field and roughly what each one consumes per year in wear parts. That consumption happens regardless. The only open question is who gets paid for it.
A rough estimate in five minutes
You need three numbers:
- The number of machines still in active use. Not everything you have ever shipped — what is still running.
- The average annual parts spend per machine. If you do not know it: look at your customers on a full service contract. There you do see the complete consumption, because all of it runs through you.
- Your actual annual parts revenue. Parts and consumables only — leave out new machines, service hours and contracts.
Machines times spend per machine is what your installed base consumes. Subtract your actual revenue and you have an upper bound on the gap. Not a measurement, but an order of magnitude — and that is usually enough to decide whether the subject deserves attention.
Two things to stay honest about. Machines on a full service contract already consume through you, so they do not belong in your machine count. And a part sourced locally in an emergency is not by definition revenue you could have captured: downtime beats loyalty, every time.
Three numbers that confirm whether the gap is real
The estimate tells you whether this is worth looking into. These three tell you whether something is genuinely going wrong in your process — and they sit in systems you already run.
1. The share of requests that took more than two rounds of email
Take two full weeks from the service@ mailbox, spread across the year so you do not catch a seasonal peak. Per thread, count how many messages went back and forth before an order existed. Anything above two means there was uncertainty about which part it had to be. This doubles as your cost-to-serve: every round is work for your service desk and waiting time for your customer.
2. Returns booked as "wrong item"
Pull all return and credit lines on parts over twelve months from your ERP and set them against the total number of parts order lines. Filter on "wrong item" and "does not fit", not on "damaged". Expect the return reason code to be poorly filled in — reading a sample by hand is part of the job. And know that you are measuring a lower bound here: many wrong parts never come back, because a customer with a stopped machine simply orders the right one alongside it and bins the wrong ring.
3. Second shipments within ten working days
Count how often a second parts shipment went to the same customer, for the same machine serial number, within ten working days. That is a fairly clean error indicator, and the express shipping costs attached to it give you a figure in euros. A figure convinces an internal audience better than a percentage does.
A fourth number is worth having if you can get at it: the share of requests arriving outside office hours or from another time zone. If part of your installed base sits in Australia or North America, that single number proves there is demand at moments when nobody at your end is picking up the phone.
Why a webshop does not fix this
The reflex is a webshop. Catalogue online, search bar on top, order button underneath. But a webshop assumes a visitor who knows what he is looking for, and that is precisely what your customer does not know. A search bar that expects a part number the technician does not have only moves the problem around.
What does work is starting from the one thing your customer knows for certain: his own machine. That takes four things an off-the-shelf package generally does not deliver.
- An installed base per customer: which machines they own, with serial number, year and options.
- Part selection through exploded views or machine configuration, so people click on a drawing instead of searching for a number.
- Customer-specific pricing live from your ERP. If the customer sees a different price than the invoice shows, he will call you anyway.
- Punchout through OCI or cXML for customers who buy via SAP Ariba, Coupa or a system of their own.
There is more on how such a portal fits together on our page about spare parts portals.
Start by measuring, not by building
The biggest risk in a project like this is not the technology. It is that a year from now you cannot demonstrate what it delivered, because nobody ever took a baseline. Write the definition down — including the exact query and the period — so you can repeat the same measurement later.
Watch one trap while you do. After go-live you measure neatly and completely through portal data, while your baseline came from a rough mailbox analysis. Compare those two and you are measuring your method, not your improvement. Use the mailbox analysis to size the problem, and the return and credit measurement from your ERP to prove the improvement. That one you can repeat identically.
And if the outcome is that the gap is smaller than you thought: also fine. You will have avoided an expensive rebuild with a few days of analysis.
