The Pareto Estimator
Pushing the Pareto approach to its logical extreme to turn it into a universal estimation tool, usable even when information is scarce.
Long car rides have this ability to turn the mind into an ideas laboratory. On the road between Lugano and Turin, a thought obsessed me for nearly two and a half hours. It’s fairly simple in principle, but its implications struck me as deep enough to deserve being put into writing.
The idea: push the Pareto approach to its logical extreme, and turn it into an estimation tool usable even when information is scarce.
The starting point: cyclical nature
We all know the Pareto principle: 20% of inputs produce 80% of outputs. We try to take it into account, more or less consciously. But what struck me is the cyclical side of this observation. If I put in 20% of the effort and get 80% of the result, what happens if I reapply the same logic to the remaining 20%?
In two cycles, I can reach 80% + 80% of the remaining 20%, or 96% of the result. In only two iterations of targeted 20% effort. And as a general rule, whenever I lack information, I’ll assume the Pareto principle applies.
= 96% of the result with ~36% of total effort
Becoming excellent at Go
Let’s take a concrete example. I want to develop an excellent level at Go — the strategy board game, not the programming language. My level is relative to a population. So I define my goal: be better than 80% of people (i.e., in the top 20%).
To get there, I look for the 20% of knowledge and practice that will produce most of my progress. Let’s say that after a time T, I’ve reached this first plateau.
Now I want to keep going: be better than 80% of the top 20%, meaning I’d be in the top 4%. I’ll need to find the new relevant 20% at this level — concepts that didn’t even exist in my initial field of vision. And something interesting happens: the time required isn’t the same.
In fact, everything I’ve invested so far only represents 20% of the effort needed to unlock this second cycle. In other words, I need 4T more. Total: 5T to be in the top 4%.
If I push further — top 0.8% — the same reasoning applies. The 5T invested only represent 20% of the next cycle. I therefore need 20T more, or 25T total.
If the first cycle takes 1 month, the second will be reached after 5 months, the third in 25 months — just over 2 years. This illustrates why becoming an excellent Go player is difficult: the depth of knowledge increases geometrically. It’s not just “more of the same” — it’s a constant rediscovery of what matters at each level.
The Pareto estimator: formalization
This is where the idea of an estimator takes shape. I define the Pareto estimator as a set of measures: the sequence u_n, the cumulative effort I_n, and the output O_n.
| n | u_n | I_n (cumulative effort) | O_n (top %) |
|---|---|---|---|
| 0 | 1 | 1 | 20% |
| 1 | 4 | 5 | 4% |
| 2 | 20 | 25 | 0.8% |
| 3 | 100 | 125 | 0.16% |
| 4 | 500 | 625 | 0.032% |
What makes this estimator useful is that it works even when you lack information. Imagine someone gives me a task. I think: “Roughly, if I focus well, in two hours I should be able to produce 80% of what’s asked.”
I get started. After two hours, I take stock. Did I really reach 80%? If I feel like I only got halfway, that gives me two indications:
- Either I didn’t focus enough on the 20% of actions that matter most
- Or I misjudged the reference time scale
In both cases, I can readjust. The estimator becomes a constant feedback tool.
Applying this to a behavior: coffee
Where it gets really interesting is when you apply it to behavioral goals. I want to have a reasonable coffee consumption.
What does “reasonable” even mean? I don’t even need to answer precisely. I can directly apply Pareto.
- Worst case: coffee every day — 30 times per month
- Best case: 0 coffee
My goal becomes: not consume coffee in 80% of cases, which gives me a failure rate of 20%, or 6 days per month.
In other words, if I limit myself to 6 coffees per month, I’m on the right track. And most importantly, if I fail on certain days, it’s not the end of the world — it’s accounted for in the model. The O_n metric becomes an acceptable failure rate, allowing me to be less harsh with myself while staying on a trajectory of progress.
The fractal property
And this is where a fascinating property emerges: self-similarity.
Let’s go back to coffee. I have my goal (6 days max). Now, what is the main action that will ensure I succeed in 80% of cases?
The answer: don’t drink coffee at home.
If I impose this rule on myself, most of the time I simply won’t have access to coffee. But I can go one level deeper. What will make this rule work 80% of the time?
- Option 1: don’t have any coffee at home at all — radical but effective
- Option 2 (what I did): have an alternative — decaf, chicory
The same structure appears at every level: one or two actions produce most of the result. It’s fractal — the structure repeats as you zoom in, which explains why the approach remains usable even with little information.
Universality — and its instructive limits
What gives this approach its power is that it applies everywhere. The domain changes, the metrics change, but the structure remains.
Let’s take an example: evaluating your relative position in Bitcoin.
21 million BTC / 8 billion humans = roughly 250,000 sats per person.
| Cycle | Threshold | Position |
|---|---|---|
| 1 | 250,000 sats | Top 20% |
| 2 | 1 million sats | Top 4% |
| 3 | 5 million sats | Top 0.8% |
| 4 | 25 million sats | Top 0.16% |
But wait — if we apply 4 Pareto cycles, we should be in the top 0.16%. Yet in reality, someone with 25 million sats (~0.25 BTC) is probably not in such an exclusive elite. The model doesn’t fit exactly.
And that is precisely where the estimator becomes a diagnostic tool.
If reality deviates from the model, it means there are disruptive factors. And those factors themselves probably follow a Pareto distribution. So I list them:
- The actual wealth distribution is more unequal than Pareto — a handful of wallets hold a disproportionate share
- 3-4 million BTC are lost forever
- The global population isn’t 100% exposed to Bitcoin
- Time horizon matters — investment window, ability to secure holdings
Now, which of these factors explains 80% of the gap with my model?
Probably the first: the extreme inequality of the distribution. If I understand this factor, I already have 80% of the explanation. And I could apply the same reasoning to identify what, within this inequality, weighs the most — for example, early adopters and exchanges.
The estimator doesn’t predict exact reality. It provides a baseline from which I can identify what disrupts the model — and prioritize my understanding of those disruptions.
A tool for radical prioritization
At its core, the Pareto estimator is a clarification tool. When you have fog in your head, you list everything that comes to mind. You rank it. And you keep the top 5 — which represents 80% of 80% of what truly matters.
The rest, you can set aside. Not permanently, but for now.
And this operation can be repeated at every level of zoom. Whether you’re planning a project over a year or deciding what to do in the next two hours.
One last thing
The Pareto estimator also lets you consciously decide whether you want to move on to the next cycle. Knowing that the next level demands 4 times more effort changes the perspective. Sometimes, staying at the current cycle is the right choice.
And incidentally, what you’re reading here is itself a first Pareto cycle. It took me a certain amount of time to produce. If I wanted to improve it significantly — better structure, more visualizations, polishing every sentence — it would probably take me four times as long.
For now, two Pareto cycles are enough for me.