Death by a Thousand Cuts: Telling a Trend From an Event

The most common way a car ends is not a dramatic failure. It is a sequence: an alternator, then a wheel bearing, then a leaking rack, then something with the exhaust, each individually defensible, each arriving before you had recovered from the last. At no point does anyone make the decision to stop owning the car. The decision makes itself, badly, on a day when you happen to be annoyed.

The fix is not stoicism. It is a ledger.

Event versus trend

An event is a single failure in one system, with a known cause, on a vehicle that has otherwise been unremarkable. A water pump at 90,000 miles is an event. A collision-damage repair is an event. Events are priced with the repair-versus-replacement arithmetic and usually the answer is: fix it, because you are buying remaining years at a reasonable rate.

A trend is unrelated systems failing in sequence, with the interval between failures shortening. Cooling, then electrical, then suspension, then fuelling. Nothing is causing the others; what they share is age, heat cycles and mileage. A trend is a forecast, and it does not care what the last repair cost.

The reason people get this wrong is that they evaluate each repair the moment it arrives, in isolation, when the correct comparison is against the pattern the previous repairs established.

Build the ledger

One line per visit. Date, mileage, system, cost, and one word: wear or failure.

Wear means you always knew you would buy it — tyres, pads, wipers, a battery, scheduled fluids. Failure means something stopped working before it was supposed to. The distinction matters because wear items cluster naturally around intervals and can masquerade as a trend when they are only a coincidence of odometer readings.

Then, twice a year, compute two numbers:

• Failure cost per 1,000 miles over the last 12 months.
• Number of distinct systems with a failure entry in the last 12 months.

That is the entire method. It works on the back of an envelope and it is more reliable than any feeling you have about the car.

The three patterns

Flat and low. Failure cost per 1,000 miles roughly stable, one or two systems. This is a healthy older vehicle. Keep it, keep spending on the maintenance that actually extends life, and stop worrying.

Spiky. One large year surrounded by quiet ones. Almost always an event, or a cluster of wear items landing together. Do not extrapolate from a spike; that is how people sell perfectly good cars.

Rising and broadening. Cost per 1,000 miles climbing year over year and the count of affected systems increasing. This is the trend, and it is the one that justifies changing your plan. Note that either signal alone is weak — rising cost in a single system is often one unlucky component, and many systems at low cost is often just an old car with cheap problems.

A worked example

Assumptions stated, and these are arbitrary illustrative figures rather than any measured average.

Suppose over the last 12 months you drove 12,000 miles and recorded failures in three systems totalling F. Your failure rate is F/12 per 1,000 miles. The year before, over 11,000 miles, you recorded two systems totalling roughly half F: a rate of about F/22.

The rate has roughly doubled and the breadth has grown by one system. That is a trend. It does not tell you to replace the car tomorrow. It tells you the next twelve months should be budgeted at something like the current rate rather than the historical one — and if that budgeted figure, plus your insurance and fuel, exceeds what the replacement you would genuinely buy would cost you annually, you have your answer without any drama.

Notice what the ledger does not ask: what the car is worth. Resale value is irrelevant to whether keeping it is cheaper than the alternative. That error is the single most common one in this whole subject.

What to do with the answer

A confirmed trend does not mean sell immediately. It usually means three things.

Stop discretionary spending on the vehicle — the theatre tier of maintenance, the cosmetic refresh, the tyre upgrade. Set a replacement date twelve to eighteen months out, which converts panic into planning and lets you buy on your schedule rather than on a tow truck's. And apply a simple filter to each new repair: does this cost less than the months of use it buys at your budgeted rate? If yes, do it. If no, that repair is your exit.

Handled this way, the thousand cuts stop being fatal. They become a data series, and data series are much easier to argue with than a bad morning and a large estimate.