Spare parts and MRO
One missing part can stop a production line, so you stock everything or trust averages. A model sets each buffer from your past demand pattern.
- Free 30-minute call
- No commitment
- You own the code
Typical problems in spare parts and MRO.
A bearing nobody needed for three years fails on Friday night. The parts store has fifty bearings for a machine you scrapped and none of this one.
One missing part stops a production line
A seal worth a few francs fails, and the production line stops until a replacement arrives. Every hour of waiting costs production.
Lumpy demand
A part goes unused for months, then three fail in a week. The average says a little every month, and that never happens.
“Stock everything” fails
The store holds parts for machines you scrapped years ago. Money stays tied up in those parts while critical parts run out.
Repair or replace?
A worn pump could be repaired or replaced. You decide from habit. What it costs when that machine stops plays no part.
In short: A spare parts inventory optimisation sets each part’s stock from its past demand, which often shows long gaps and sudden spikes. Methods such as Croston forecast two things separately: how many pieces are needed each time, and how much time passes between two demands. The more it costs when a part is missing, the bigger its buffer.
Stock everything, or trust the average
Friday, 10 pm, and a gearbox bearing fails on production line two. The store holds dozens of parts for machines you scrapped years ago, but not this one. The production line stops until a courier brings the part on Monday. In plants with many machines, one missing part can shut down a whole production line.
So you’re tempted to stock everything. Then a lot of your money is tied up in parts that stay on the shelf for years. Or you trust average demand. But spare parts demand is lumpy: nothing for months, then several at once, and an average hides that risk. You end up holding too many parts you rarely need and none of the one that just failed.
Spare parts inventory optimisation for asset-heavy operations
Today, your stock of parts for maintenance, repair and operations (MRO) runs on habit and memory. With our model, each part gets a buffer set by its own demand pattern and how critical it is. Parts that can stop a production line get the largest buffer. Parts you rarely need take up less shelf space and less capital.
You see, part by part, what each buffer costs and how rarely the part would still run out. When a component wears out, you know whether repairing or replacing it is the better call.
Croston, ADIDA and intermittent demand forecasting
Your forecast for these parts needs methods built for long gaps and sudden spikes. The Croston method forecasts two things separately: how big each demand is, and how long the gap lasts. ADIDA, the aggregate-disaggregate intermittent demand approach, adds demand up into longer periods, forecasts those and splits the result back down. Our model takes that forecast as its input. It sets each part’s stock by how critical the part is, across several levels such as a central warehouse and site stores (multi-echelon inventory). Where demand is very uncertain, stochastic optimisation plans for the whole range of possible demand.
Which method fits which of your parts?
Guides explain Croston on its own. You can sort your own parts with two numbers. ADI, the average demand interval, is how many periods pass per demand: a part used in 6 of 24 months has an ADI of 4. CV², the squared coefficient of variation, measures how much the order sizes vary. Syntetos, Boylan and Croston (2005) set the cut-offs at 1.32 and 0.49. Parts above both are lumpy, and parts with a high ADI but steady sizes are intermittent.
The choice of method follows. Croston tends to forecast a little too high. The Syntetos-Boylan approximation (SBA) corrects this by making the Croston forecast slightly smaller. For intermittent and lumpy parts, SBA is a sensible default. Where a part may go obsolete, the TSB variant (named after Teunter, Syntetos and Babai) adapts faster. Run that sort on your history and you’ll see which parts the average has been getting wrong.
How a project runs.
Each of the three phases ends with something you can check before we go any further.
-
Understand
We start with your operation. What do you plan, where does it go wrong, and what would a better week look like? Then we tell you straight whether optimisation can help, and sometimes the answer is no.
- A written assessment
- A clear go or no-go
-
Model
We write your decisions, goals and rules into a mathematical model, which is just a precise description of how your operation works. Then we run it on your own data. You see side by side what each plan costs and what it delivers, and pick the one that fits your priorities.
- A documented model
- Results on your own data
-
Implement
We test with your data and go through the results with your team. Once they trust it, we connect the model to your systems, so it runs every day without anyone stepping in by hand.
- Source code and documentation
- Training for your team
Common questions.
Something else? Ask us directly.
Ask what happens if it’s missing, and how fast you can get it. Say a production line stops without the part, and delivery takes weeks. Then keep the part in stock, even if you need it only once every few years. A part you can buy the same day may not. How important a part is depends on how fast you can get it as well as on the machine.
When a part has long gaps between demands. Sort your parts by ADI, the average number of periods per demand, and by how much order sizes vary. Above an ADI of 1.32, classic methods struggle. For those parts, use Croston’s method or its corrected version, SBA, which removes Croston’s tendency to forecast high.
Set them from three things. The first is how often and in what quantities the part is needed. The second is how long delivery takes. The third is how often the part must be in stock when it is needed (the service level). At 95 percent, the part is there in 95 of 100 cases. Many spare parts are needed only rarely, and then sometimes one piece and sometimes ten. The usual formulas assume steady demand and are wrong for such parts. A model uses each part’s demand pattern and the criticality of the machine. Review the levels whenever lead times or the installed machines change.
It depends on the problem, so we agree the timeline with you up front. Every project starts with a diagnostic, a short look at how you plan today. It shows which data, how much time and what budget a model for daily use needs. Only then do you decide.
No. We connect to your ERP, your warehouse management system or your planning spreadsheet. We build the model only for the planning that gives you trouble today. Your systems keep running as before.
Almost nobody’s data is clean. The diagnostic shows what the model needs and what’s missing. Then we build the model to cope with messy data.
Off-the-shelf tools are built for many companies at once, so they don’t know your special rules. We build a model that follows exactly your rules. Then we hand it over, and your team owns it.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
- ADIAverage demand interval
- CV²Squared coefficient of variation
- Croston’s method
- SBASyntetos–Boylan approximation
- TSBTeunter–Syntetos–Babai method
- ADIDAAggregate–disaggregate intermittent demand approach
- Cycle service level and fill rate
- MROMaintenance, repair and operations
- Stochastic optimisation
Tell us what you’re planning.
In a free 30-minute call with a specialist, you find out whether optimisation can help.
- Free, no commitment
- You talk to a specialist directly
- A clear go or no-go