After a visit to a bicycle shop I realised that I need to increase my budget. It makes sense to buy the more expensive bike, as it’s more fun, nicer to ride, and I totally need the fancy features, right?

With this premise I was happy to find Olof Hoverfält‘s post about data-supported decision making. In the genius piece, Olof uses his wardrobe as a case-example of the effect of value-vs-cost. Through meticulous data collection over three years (and counting) he is able to make informed statements about clothing categories, quality, pricing, value, cost of preference, and actual frequency of use. The significance of the post is that it explains important concepts of informed decision making in familiar terms and a relatable context.

Let’s take a look at some highlights.

Real cost. Expensive can be cheap and vice-versa. You can’t know the real cost of an item unless you know it all: purchase price, depreciation rate (or lifetime + value at divestment), actual frequency of use, and quality. A pair of shoes may cost a lot, but if they’re used daily during the looong winter and they can take it (durability), they turn out very cost-effective.

Category differences. There may be subtle differences between seemingly similar categories. Every item in knitwear category is available for wearing every day (unrestricted category). Underwear may spend days in wash cycle after use, becoming available after a significant delay (resricted category). Value of investments can’t be compared directly across categories as the competitive attributes are different.

Cost of quality. The definition of quality is not obvious. Durability is a factor, so it makes sense to buy cheap and durable items. But it makes no sense to buy cheap and durable items that are used very rarely. An expensive shirt may not be cost effective but has other attributes: nicer style and cut, better details and materials, etc. There is a cost related to perceived quality, and the cost can be quantified. In Olof’s post, the cost of “fancy shirts” is 500 euros per year.

Value of long term data. You’d think that after a year of daily tracking you’d have a pretty good data set for making informed decisions about something as simple as clothes. Not so: Olof’s analysis after a year is very different from after three years, and the results keep changing as more data flows in. The actual frequency of use may be very different from the estimate. In terms of data the saying holds: The best time to plant a tree is 20 years ago, the second best time is today.

The data collection template is available at https://hoverfalt.github.io/.