Read time 24 min
Have you ever wondered whether expensive clothes are worth their price? Or had that subtle feeling of guilt when buying something pricey, and then justifying it because you will wear it so many times, even if you have no clue if it’s actually true? If you thought yes, then this is for you.
I took a deep-dive into my closet to find out what my clothes truly cost, and learnt a ton about performance, sustainability, and myself in the process. In this blog post, we go on a journey of data-supported decision making. Even with small data, you can get big insights – and some nice shoes along the way.
I have kept a continuous daily log of the use of all my clothes for the past three years.
What started out as a simple question of whether it makes sense to buy expensive clothes has now turned into a rather deep discovery into the cost performance of clothes. You can explore my continuously updated wardrobe performance here.
My principles for more or less any consumption have for a long time been to buy what I need, use what I buy, and take good care of what I have. With clothes, however, I had no idea of how I was really doing.
The insights from this project have helped me improve the choices I make. I also ended up building an analytics platform for the performance of not only clothes, but any kind of durable good.
Three years into this journey, it’s time to reflect on everything I’ve learned and hopefully help others understand the logic, practices, and benefits of data-supported decision making. And if you’re only here for some wardrobe-perfecting tips, head over to the end of this post for a practical guide to clean out your closet and consumption habits.
Easy daily wear data
Let’s start with a brief look at the data, as it is the core enabler of everything to come.
My wear data is a continuous daily log of my use of each piece of garment since January 1st, 2018. To date, that is a total of 426 items, 1106 days, and over 300,000 data points of use and non-use. All items are divided into a simple MECE structure of 12 categories: jackets and hoodies, blazers, knits, shirts, T-shirts, pants, shorts, belts, socks, shoes, underwear shirts, and boxers. Sportswear is an additional category that is outside the portfolio of “daily” clothes.
“Data entry takes less than a minute per day. I spend four times as much brushing my teeth.”
My first version was a simple Excel that contained the wear data and a few plots. At some point, Excel started choking on my growing data set. I needed more powerful computing and data visualization capabilities, so I started building my own platform using R, a programming language for statistical computing. All computing and plotting is now automated in R. It took some 2300 lines of code. I also moved the master data from Excel to Google Sheets in order to make data entry as easy as possible and to keep all data in the cloud.
At the moment, the data entry UI in Google Sheets looks like this:
These are my jackets and hoodies that are active, aka items that are still in use now, as opposed to those that have been sold, donated or recycled, and are no longer available for use.
I count “wears” by day. Every evening, I open the browser tab and check the clothes I used that day in the list of active items. It takes less than a minute per day, which is roughly six hours per year. I spend four times as much brushing my teeth. Easy data entry is indeed critical to any voluntary, continuous, non-automated data collection. And yes, I have also played with ideas of RFID tags and image recognition to automate even that one last minute.
Once the platform and processes are set up, the data set itself is “cheap”, at least in comparison to the world of insights it unlocks.
Unveiling the real cost of clothes
Clothes are a form of durable goods. Therefore, a garment’s purchase price alone tells close to nothing about its actual cost performance.
To account for the costs of assets that deteriorate with use, the standard approach is to depreciate such assets over their expected economic lifetime. It works fine when the underlying estimated lifetime is acceptably accurate, as it is in most cases of continuous use of identical assets. As we shall see shortly, most of a wardrobe is nothing like that.
“I had no idea of what the lifetime of my various garments might be.”
When I started looking into my clothes, I had no idea of what the lifetime of my garments might be. I of course tried to estimate my use for various items but later learned that nearly all of my initial guesstimates back in 2017 were way too high.
This is where daily use data corrects the distorted picture and lets you explore what the use really looks like, cleared of biases and wishful thinking.
Actual versus imagined use
Let’s start with the simplest thing: use.
The figure below shows the cumulative number of wears for all my shoes. The items marked green are divested and provide a benchmark to compare the active items against. The shaded green areas show one and two standard deviations from the mean for the divested items.
In the case of shoes, there is quite a large spread in times worn. This is understandable as the shoes are all unique. My black winter shoes (the ones at the far right) really stand out with over 300 wears and still going.
Items that are identical or highly similar, such as underwear T-shirts, tend to have a much smaller spread. The animation below shows the historic time progression of wears for all of my underwear shirts. We can see that this use is rather stable and predictable.
Based on the divested (green) sample of 41, we now have a picture of the distribution of the number of wears for items in this category. Underwear shirts typically last between 20 and 25 times. That means an active shirt with about 28 wears is likely to be nearing the end of its lifecycle, and that a shirt that starts falling apart after 10 uses is no good.
Now that we know the number of wears for each item, and the typical use performance range for each category, we are finally ready to start exploring cost performance.
Cost per wear standardizes cost performance
Tracking the use of a garment, or any durable good, enables studying its real cost performance over its lifecycle. In the context of clothes, I use Cost Per Wear (CPW), a “wear” being one day.
The impact of purchase price on CPW is linear and easy: an n% decrease in price means an n% decrease in CPW.
The impact of use is non-linear as CPW follows a rational function curve a/x, which means the amount of additional wear needed to push down an item’s CPW keeps increasing. This is mathematically simple and intuitive, yet a visualisation may help make this dynamic and its implications stand out.
Below is a plot of CPW by use for all my shoes. The gray dots trace the use of each item. The blue dots are guide trajectories for items with various purchase prices. Again, items marked in green are divested and no longer move.
The goal is to push each item as low down on the chart as possible through continued use. As mentioned, this becomes slower and slower with increased use. Note that the CPW axis scale is logarithmic.
CPW is the real cost of each of the times the garment has been worn to date. While that is useful in and of itself, perhaps the biggest benefit of CPW is that it makes garments comparable across price tiers. It helps answer the question of whether quality correlates with price.
“In some cases, buying cheap is provenly more expensive.”
The table below shows the performance of all my white sneakers, active and divested. Note that the maximum limit for the light red CPW column is that of the sample of divested items, as that is the meaningful benchmark.
So what does CPW tell us? Here are a few remarks.
The 90 euro Converse sneakers and the 30 euro Mywears have a similar CPW of 0.87 euros and 0.70 €, respectively. Their effective cost is roughly the same, which means that walking around in the cheap Mywears is roughly as expensive as walking around in Converses. In this case, money buys quality, at least when measured by durability. It takes about two pairs of Mywears to match one pair of Converses. From a sustainability perspective, I would imagine one pair to be better than two, even though the items may have differing individual footprints.
The more expensive sneakers, the 150 euro Diesels, perform remarkably worse than the two aforementioned. Their actual CPW of 1.88 € is more than twice as high. In this case, money does not seem to buy quality, at least not durability. The Diesels did last twice as long as the Mywears, but they were five times more expensive, leaving them at a far lower level of cost performance.
The clear leader at the moment are the Hugo Boss sneakers that are still active and may improve further. The Hugo Boss CPW of 0.57 € is 0.30 € lower than that of the Converses at 0.87 €. This might seem like a rather small difference, but it is 34 percent lower. From a cost performance point of view, that is quite a lot.
Then there are other still active Jim Rickeys and Helly Hansens in their own league. This is understandable as they are still (hopefully) in the early phases of their lifetime.
So the conclusion seems to be that, disregarding potential brand and style preferences, the Boss sneakers are the best. We have also seen that in some cases, buying cheap is provenly more expensive. This would support the old saying “I cannot afford to buy cheap.”
Now this is, of course, not always the case. Let’s look at such an example next.
Price might buy quality, but it doesn’t always turn into a comparably low CPW. In my wardrobe, this is the case with shirts, which are plotted below. Again, note the logarithmic CPW axis scale.
My shirts can be divided roughly into two categories: relatively expensive ones (100 euros and up) and less expensive ones (40 euros and below). These tiers show in the graph as the upper group on a gentle slope, and the lower four on steeper slopes.
The expensive divested shirts have a mean CPW of 3.44 €. The less expensive shirts (all of which are still active) are at an average CPW of 1.29 € and decreasing. While the less expensive shirts are still active and may still improve, we can confidently say that they perform significantly better from a cost efficiency perspective.
“I now know the monetary cost of my preference for quality, and I am happy to pay for it.”
This brings up the obvious point that quality is not limited to, and cannot be measured solely by, durability. In my personal view, the expensive shirts are of higher quality in many regards. This means each wear provides a more valuable subjective experience (better materials, style and cut, details, brand, etc.), which may justify the difference in CPW.
Through past purchase and use data we now have a way to quantify the difference in the perceived value, i.e. how much more I am willing to pay for a premium experience. Based on my behavior, I can only conclude that I am indeed willing to pay two to three times more for it or roughly 2 euros per wear. I now know the monetary cost of my preference for quality, and I am happy to pay it.
The example with shirts goes to show that for myself, wardrobe performance is not about driving down absolute cost, but about finding an optimal balance between value and cost.
Frequency of use is the underlying driver of performance
Since use is a fundamental driver of performance, it makes sense to look at what affects it and seek to understand what might limit it.
The fastest increase in use for an item is for it to be used every day. Anything less than this means a slower accumulation of use, and hence means a slower progress toward a lower CPW. An item’s “use effectiveness” can thus be measured as frequency of use. In this context, I have defined it as average times worn per month, which is intuitive at the single garment level.
To continue with the shoes as an example, below is a plot showing cost per wear and frequency of use, as well as the weighted average for each dimension among divested items (green). The weighted averages show the performance at the category level and provide a benchmark against which to compare the active items. This for me is the most useful plot to actively follow.
With CPW, lower is naturally better. An active item that is still above the dotted CPW average line is underperforming compared to divested items. It would weaken the category performance if divested. An active item below the same line is already performing better than the divested items. It will improve the category performance when eventually divested.
The dimension of average times worn per month shows the frequency of use for each item relative to the others. Higher is better. Items to the right of the dotted average line have been used more frequently than the category average, while items to the left have been used less frequently.
The dimension in essence shows how “popular” each item is. This is powerful as it shows real preference versus imagined. This plot is where you will identify those really-like-but-never-wear garments in the upper-left corner. It is pretty brutal. You can probably spot the shoes that I might explain to myself as the ones I “have yet not started using”, but which I perhaps never should have purchased in the first place.
Items may have good reasons to be relatively low on frequency. They may be specialised for special occasions. In my shoe plot, the Helly Hansen white sneakers are at my summer house, which limits their potential use significantly. More generally, you can really only use winter clothes in the winter season and summer clothes in the summer season. In order to keep things simple and comparable, I have decided not to account for this “availability of potential use” factor.
Let’s quickly look at the frequency of use over time. This animated visualisation may help see the dynamics we have discussed in action all at once. Below is the same plot, but as a time progression.
The animated plot shows rather brutally which items are rarely used. As time passes, they keep slipping further and further to the left. Here we really see how time keeps eating away at their use frequency, as if they are constantly pushed back by the wind in their fight forward. In a weird way, I feel like I want to cheer them all in their race against time, and each other, to reach the lower-right corner.
Speaking of competition, the point of “competitive dynamics” is also particularly important. The items in a category are generally substitutes (with important aforementioned exceptions). They compete for use in a game where usually only one at a time can take a step forward.
Since the total potential use of any garment is ultimately limited by the number of days available (there are only so many days in a month), available days become the scarce resource within the category. This is why frequency of use is so helpful in seeking to broaden the understanding of performance.
“A wardrobe with nothing but favorite clothes sounds nice. It may also be the best in terms of cost performance.”
We need to make an important distinction between two types of categories: those whose availability for use is heavily affected by the wash cycle, and those for which it isn’t. Shirts, T-shirts, socks, underwear shirts, and boxers I generally wash after each use. That makes them unavailable for use until clean and back in the wardrobe, which might be six days later, or even longer. Let’s call these “restricted”. The “unrestricted” garments then fall into the other categories: jackets, blazers, knits, pants, shorts, belts, and shoes.
This obviously has a significant impact on what frequency of use an item can achieve, and what should be considered good performance. But there is a more profound difference in competitive dynamics.
In unrestricted categories, substitute items compete with each other for use. In these categories, use frequency essentially shows my preference hierarchy between the items, in a way their relative “popularity”. This tends to lead to big differences in performance. And the more substitutes there are, the poorer they all perform.
In restricted categories, substitute items don’t compete with each other, provided that the amount of items only slightly exceeds the maximum amount of items idling in the washing process. As a simplified example you might have seven shirts, one worn every day, and all of which are washed every Sunday. In this case they don’t compete at all. But if you have seven shoes, one pair worn every day, they compete quite fiercely.
A good example of a restricted category with high substitutability is underwear shirts. I only have two kinds, t-shirts and tanks, and two colors, black and white. So four subcategories within which all items are perfect substitutes. The only way I can tell them apart is by the number I have added to their label. The plot below shows their performance is very similar and easy to predict.
This can be contrasted with belts, which is an unrestricted category with medium substitutability (I consider whites, blacks, and browns to be substitutes within their color but not across colors). Here we see widely varying frequency of use and the most “popular” belt clearly stands out at the expense of the others.
So competitive dynamics matter. They influence use and therefore cost performance in a rather complex and indirect way. For myself, the key takeaway is to recognize these dynamics and choose an optimization approach accordingly. In short, unrestricted categories should have only a small number of items that you actually use: the “popular” ones. Restricted categories can be more varied, but should only have the amount of items that take you through your wash cycle.
Someone once said their goal is to have a wardrobe with nothing but favorite clothes. That makes sense not only from a value perspective, but in light of my data, that may also be the best alternative in terms of cost performance.
Category performance shows how well the whole thing works
The category level is about how all items in a category perform as a whole. In addition to average CPW, category cost performance can also be described as category daily cost over time. This has the benefit of showing progress over time.
The plot below shows the daily cost of underwear shirts over time. Every dot is one day in time. Green dots are days in which all items are divested at the time of plotting. They no longer change. Red dots are days for which at least one item used that day is still active at the time of plotting, which means that that day’s daily cost may still decrease. Most days, only one item was used. The blue line shows the 30-day rolling average. The yellow shaded areas show summer holidays, which tend to have quite an impact on what I wear – we Finns take our summer holidays very seriously.
We can see that the daily cost varies quite a bit depending on the item(s) used. We can also see that many items are in use for quite some time. Some items from 2018 are still in use today (red). This means it takes a rather long time for the average cost trend to show and to settle.
Category level daily cost has one very significant benefit. It describes the cost of the “utility” of the particular kind of clothes, regardless of how many or which items were used on how many days. This plot essentially says: “This is how much the use of underwear shirts has been costing you daily”, which is a little over 1 euro per day. This is where improvements in item level performance and the effect of pruning the wardrobe eventually show, although with a significant lag.
Portfolio performance is the ultimate measure of progress
We have finally come to the point where we can look at the entire wardrobe. I call this the portfolio level, because it contains my entire portfolio of goods, or assets. It is here that all improvements, big and small, ultimately show.
Let’s start by taking a quick look at how the different categories compare. In order for the comparison to be more intuitive, I decided to change the category frequency of use measure from number of wears per month to share of days.
Now that we have developed a way to describe category performance in a normalized way, we can compare the categories. The plot below shows the average CPW and category frequency of use for all categories.
At last, we have a complete and exhaustive overview. Even a quick glance reveals which categories are expensive and which are less so, and also which categories are heavily used and which ones only seldom. There seems to be a negative correlation between CPW and frequency or use. It would seem natural – and justified – to accept a higher CPW for clothes that are worn seldom, provided that they last for a long time of course.
Blazers seem very expensive, and they are, if we look at CPW. They cost more than six times more than pants per wear. On the other hand, they are used much more seldom, which means their annual cost might still not be that bad. In order to find out, let’s take the final step and convert CPW into a yearly cost.
The plot below shows the yearly cost for each category.
Finally, we see category performance in an intuitive and familiar form. In all simplicity, it shows the real cost per year of each category. At this point, the expensive but rarely used categories don’t look that bad anymore. But my shirts stand out in a league of their own. I spend 751 euros on shirts yearly. Do you recall my costly affection to high quality shirts? Well, here it shows. I now know I could save some 500 € per year by downgrading. But I won’t. I like my fancy shirts and I’m fine with the cost now that I know it. This is a conscious choice. Gone is also the subtle guilt of not knowing whether this tradeoff is sensible. It is.
The final measure of real cost and its possible trend is daily cost plotted over time. Below is a plot of the total daily cost of all my clothes over time. As with the category view, every dot is one day. And again, green dots are days in which all items used that day have been divested. The red dots are days in which at least one item is still active today. Red dots may still move, green ones don’t.
We can see that the rolling average daily cost hovers around roughly 10 euros per day. The yearly dip during the summer holidays (the yellow shaded areas) shows that my summer chillout outfits tend to be cheaper than my work and city wardrobe.
You may have noticed there aren’t many green dots. That means most days during the past three years, I wore one or more items that I am still using today, which in turn means the daily price back is still decreasing. Not by much, but it is. This goes to show just how long time ranges are needed to see the effects of changes in purchasing and use.
The animation below neatly shows how this lag works.
We can see how past days become cheaper with future use. The impact goes far back. If we were to look at the daily cost for 2018 at the end of 2019, it would be too high. In the same manner we cannot say much about 2020 at this point.
I can tell by the shape of the curve that 2020 will eventually differ from previous years because of the pandemic.. As a result, the daily clothing cost for 2020 is likely to be significantly lower. But I will not know for sure until much later when the new covid-era garments have been divested. Only time will tell.
The perfect wardrobe: Tips for cleaning and restocking your closet
In my context and with my general principles and preferences, here are some things that have helped me prune my wardrobe. I’m by no means claiming to be any kind of expert nor authority on the subject, but if any of this is of use or inspiration to anyone, I’m happy. If not, I’m still happy.
Find what you need AND love, then only buy that. Sounds easy. Yet building this discipline is hard. There is a difference between need and “need”, as there is a difference between love in the store, and love two weeks later.
Focus on use, not price. Use is about value. As with any value-cost tradeoff, it makes sense to look for value first, then work out what constraints the cost might bring. Remember that pricier clothes may actually be less expensive.
Avoid “second tier” clothes. Those are the ones that are kind of nice, but never seem to get picked into an outfit as you prefer very similar but better ones. It really does make sense to only have favorite clothes.
Avoid clothes for narrow contexts. Seek to have everything as compatible as possible with as broad a set of contexts as possible.
Maximize compatibility across categories. One way is to limit your wardrobe color palette. Non-compatible clusters increase inventory, decrease use, and lower performance. This also lessens the cognitive load of finding suitable combinations.
Shop for the long-term. Some high quality garments may need to be worn for years in order to perform as they should. In order to give them a fair change, you may need to prepare to wear them for a very long time.
Take good care of your garments. Continuous care and the occasional small fix might double a garment’s lifetime. It might also make sense to send high quality clothes to a tailor to be repaired. Zippers, oh those frequent garment-killers, can be replaced.
Know when to divest. Even though we seek long use, there is a limit at which it is no longer justified to continue pushing. Enjoy the lack of guilt when divesting an item that you know has served well.
Make your garments look good. Picking clothes in your orderly and well-lit wardrobe should be as nice as browsing through a store. Well, at least the feeling can be replicated. We tend to value what looks valuable.
Accept your special weakness. I guess we all have our weaknesses. I certainly do. One of mine is blazers. Despite all this performance optimization, I still buy too many. At the current rate, I have a 10-year supply. Sensible, no. Lovely, yes. And that’s fine.
Up next: The sustainability factor
This is where my journey into wardrobe performance has taken me so far. I feel like it might make sense to broaden rather than deepen. While I will definitely keep spending that minute every day entering my wear data, and keep seeking to improve my purchase and use choices on a continuous basis, I have two areas of special interest to perhaps pursue next.
One of these is sustainability. Monetary cost is of course not the only kind of cost related to clothes. I would very much like to somehow include the sustainability perspective to be its own dimension of performance. For most items, I have collected their weight and material composition. Let’s see what could be done with that.
The other is to expand comparability to all spending. Since the use data is exhaustive both in terms of items and time, each category and the portfolio describe the “utility of clothes”. This makes clothes comparable to not only other types of physical assets that deteriorate with use, but to service models providing any kind of continuous utility, be it physical or intangible, metered or flat rate. This in turn enables evaluating consumption alternatives and choices across types of spending and across business models.
It would be fun to see what a journey into utility performance might reveal.