A container replacement budget is usually built from an assumed service life – the fleet is this many containers, they last this many years, so the annual requirement is the first divided by the second. That calculation is simple and consistently wrong, because it ignores the shape of the failure distribution and it underestimates loss. A forecast from measured data takes a year to build and is accurate thereafter.
Why the simple calculation fails
Containers do not fail at a uniform age. A fraction fail early through damage, a fraction are lost, and a long tail survives well past the average. Dividing the fleet by the average life produces a number that is too low in the early years – because the early failures and the losses are happening – and too high later. In a fleet that is still growing, the error compounds.
What to record
Every container retired, with its date of entry into service, its age, and the reason. Three fields. Where containers are individually identified, the age is known; where they are not, a batch-level date is enough to start. A year of that record produces a survival curve, and the survival curve produces the forecast directly.
Reading the survival curve
The slope at any point is the retirement rate at that age. The area that matters for next year’s budget is the number of containers currently in each age band multiplied by that band’s failure rate. Summed across bands, that is next year’s requirement – and it is far more accurate than any average-based estimate, because it accounts for the actual age profile of the fleet.
Loss is the term that is always low
Containers that leave the loop and never come back are retirements, and they are frequently not recorded at all – a container that is missing is not visibly retired, it is just absent. Where a fleet has never reconciled its count, loss is almost certainly larger than assumed. A periodic physical count is the only way to establish it. The mechanics are covered in where containers actually go missing.
Growth, and how it changes the shape
A growing fleet has a young age profile and therefore a low current retirement rate, which rises as the fleet ages. A static fleet has a stable profile and a stable rate. Forecasting from the current rate alone therefore over-predicts for a young fleet and is right for a mature one, and knowing which applies is part of the calculation.
What accuracy is achievable
With a year of records and a stable process, a forecast within about ten per cent is realistic. Without records, the error is usually an underestimate of a third or more. That difference is the cost of the recording – and the recording is a few minutes a week once the retirement criterion exists.
Using the forecast for more than buying
The same data shows which failure modes dominate, which feeds the next container specification, and how the fleet’s condition is trending, which feeds the maintenance budget. A replacement forecast built on real data is therefore not only more accurate – it produces two other useful outputs at no extra cost.
What to do in year one
Start recording before the data is needed. A fleet that begins recording retirements when it first suspects a problem has to wait a year for the answer; one that started recording earlier already has it. The record is cheap and it is the basis for everything else in fleet management.