↬ Car odometers as a flawed metric

2012-04-16

(Editor’s note : Taking a break from your newly setup blog usually means it wasn’t a good idea to start it in the first place. It shows a clear lack of enthusiasm for the subject, a poorly thought out project and a less than optimal discipline. But please believe me, it was due to professional constraints that are now out of the way. Now back to our schedule.)

Metrics are important : it is easier to make decisions when you have a small number of clear indicators to check. Gamers know this : for example, in my all-time favourite, The Legend of Zelda : A Link to the Past, the protagonist’s health is depicted as a simple fraction of hearts out of a potential of at most twenty. This is quite easy to manage for the player : if you have 12 hearts out of 12, everything’s ok. If you only have 3 hearts remaining, the next action is clear (find hearts, and quick!).

This gets tricky when it comes to real health. For this, BMI, body fat ratio and blood pressure are indicators that synthetize a few facts about the tricky machinery that is the human body to give us a clear message. But no doctor would only look at one number and declare you healthy or not. A high BMI can be explained by muscles and high blood pressure can be genetic or environmental.

So let’s take an even simpler machine : a car.

A few weeks ago, La Facture (a consumer protection investigative show aired on the CBC’s french network) presented a report on car odometer fraud. I’ll leave to others the thrill on commenting on this matter’s legislative ramifications, but it struck me that when it comes to used cars, it is not unusual to rely on only one metric for assessing a vehicule’s health : the odometer’s reading (beyond looking for rust or very useful ‘check engine’ lights).

It is true that it is unequivocally unethical to change this reading in order to trick some buyer into paying more than what the car is worth. However, this is a trust issue more than a technical or legal one : even with ample police and technical work against fraud, it has to remain legal to be able to modify a car odometer (you wouldn’t want an old car with a new odometer that reads 0, so the mechanic who replaces an odometer has to punch in the right numbers). We are likely to always have to live with the fact that some people trick odometers. Therefore, our trust in this metric should not be 100%.

Let’s not go ballistic and simply put the fault on the buyer for relying on this number when he should know that it can be tricked. But knowledge is powe (why no final ‘R’? read this) and we have to remember not to rely blindly on artifical or synthetic metrics.

When we realize what metrics represent, we are better equipped to make decisions. What numbers (or qualitative data) do you rely on daily? Can you replace these with better metrics, and can you explain what sources of bias they have?

(At least, I know that in The Legend of Zelda : A Link to the Past, you had to check for what was in Link’s jars to know the complete situation.)