What is fleet attrition, and why it changes your installed-base number

An old piece of farm machinery left to rust, abandoned in an overcast field.

Every installed-base model has a leak in it, and that leak is attrition. You can count units shipped going back fifteen years, sum them up, and call it an installed base, but that number assumes every unit ever sold is still sitting on a job site somewhere. It isn't. Machines get scrapped, cannibalized for parts, written off after a bad accident, or just left to rust in a yard once the repower cost exceeds the trade-in value. Fleet attrition is the rate at which that happens, and if you don't model it, your denominator is too big, which means your market-share numerator (your own unit sales divided by that denominator) comes out too small.

Most OEM strategy teams know this in the abstract. Fewer have a clean way to put a number on it, because attrition doesn't show up in any single dataset. Dealers report new unit sales. Nobody reports retirements.

Retirement rate isn't flat across a fleet

The mistake we see most often is applying one write-off rate across an entire regional fleet regardless of age. A five-year-old wheel loader and a twenty-two-year-old wheel loader do not retire at the same pace, not even close. Early in a product's life, attrition is close to zero. Machines are under warranty, dealers are still chasing parts margin on them, and owners haven't hit the point where a rebuild costs more than the machine is worth. Then somewhere past the midlife point, usually tied to a major component interval, the curve bends and retirements start stacking up fast. By the time a unit is past its second or third major rebuild, it's either been kept alive by an owner with no better option or it's gone.

A scaling factor for fleet age behaves like a curve: flat for years, then bending sharply once a fleet clears its midlife rebuild point. That curve carries more weight for older regional fleets, where a larger share of units sit past the inflection point and small errors in the bend compound across more machines. If your region had a sales boom eight years ago, you're sitting right near that inflection point, which is exactly where the model is most sensitive to getting the curve shape right. Get it wrong there and the whole regional total swings by a meaningful percentage, not a rounding error.

Why this matters for a market-share denominator specifically

A market-share figure is a ratio, and the denominator (total installed base) is exactly the number attrition corrupts. If you're comparing your position across three regions and one of them has an older average fleet age, applying the same flat retirement assumption everywhere will understate the installed base in the younger regions and overstate it in the older one, or the other way around depending on which direction you got the slope wrong. Either way, your cross-region comparison is measuring attrition-curve error, not actual market position.

This is also where dealer-survey-based estimates tend to drift. A survey asks dealers what they think is still running in their territory, and dealers answer based on what they see come through for service, which skews toward units still active enough to need parts. Machines that went fully dark, sold to a scrap buyer, cannibalized, exported out of the territory, don't get a vote. The survey is structurally blind to exactly the units it's supposed to be subtracting out, because those units stopped generating any signal a dealer would see.

The alternative is to work from what's actually sitting on sites right now and work backward, rather than starting from cumulative shipments and trying to guess the write-off curve. That's the logic behind a site-level count scaled up with coverage correction rather than a shipment-minus-attrition model: you're counting what's still there and scaling for the fraction of the region you didn't image, instead of estimating how many units should have retired by now. Attrition still matters for sanity-checking the result against known shipment history, but it stops being the thing your whole denominator rests on.

None of this makes attrition modeling pointless. It's still useful for forecasting parts revenue, for timing rebuild campaigns, for flagging territories where an aging fleet signals a replacement-cycle opportunity. It's just a poor foundation for a number you're using to defend market share to your own leadership.

If your regional installed-base figure is still built on a retirement curve you inherited from someone else's spreadsheet, it might be worth comparing it against a count that doesn't depend on guessing when machines die.

Request early access

← Back to the blog