The AI Pareto Analyst skill for Claude

You download the zip at the top of this page, put it in your Claude environment, and from that moment Claude analyses every list according to this logic. What a skill actually is and how that file format works, you can read in what Claude skills are NL.

You will probably recognise the problem this skill solves straight away. You have too many customers, projects or tasks to do them all justice, and you have long suspected that a handful of them deliver most of the value. But you do not have it in black and white, so you keep doing everything by half. The frontmatter of the file is written for exactly that feeling: the skill does not just trigger on the word Pareto or 80/20 analysis, but also on phrases like where should I focus, where am I losing time, I am short on time, too many customers not enough margin, I want to cut back, and what is my 20 per cent. Anyone who says something like that to Claude while the skill is loaded gets the analysis automatically.

It is written for business owners and teams who feel overwhelmed by too many projects, customers, tasks or products and want to know what genuinely has impact. It is part of the library we make freely available from the AI and automation service, with no account and no sales email afterwards.

One thing you should know up front: the skill calculates with whatever you supply. Measure only revenue, and you get a revenue analysis, and the file itself warns about that: a customer can deliver a lot of revenue and still be loss making because it costs even more time. That is why every analysis closes with the questions that still need data, rather than with a conclusion that sounds more certain than the figures allow.

The core the skill takes from that is stated literally in the file: it is not a law, it is a pattern. The ratio can just as easily be 70/30, 90/10 or 95/5. The core idea stays that most of the impact comes from a minority of the input, but which minority and how much impact differs per list. That is why phase two of the procedure is a mandatory calculation step: determine the real ratio, and note it in the output instead of the default 80/20.

That sounds like a detail, but it is the difference between an analysis and a slogan on a tile. In diff form:

The source list in the file counts five titles: Pareto (Cours d'économie politique, 1896), Juran (Quality Control Handbook, 1951), Koch (The 80/20 Principle, 1997), Hennessey (Pareto's Principle in Business Management, 1972) and Kahneman (Thinking Fast and Slow, 2011). The last one is not there by accident: Kahneman is the source for the survivorship bias warning that the skill carries into every analysis. Want to learn to set up a well founded instruction file like this yourself? The approach sits in writing a SKILL.md NL.

Before anything gets calculated, the skill checks whether it has three things: the list itself, the value or metric per item, and what result you consider it to be. That last point is subtle but important: revenue is a different result from profit, and time freed up is a different result from customers won. If one of the three is missing, it asks for it in at most two short sentences before carrying on. So you do not get a form, at most a targeted question.

Next comes the core: a fixed procedure of five phases, and the file states explicitly that no phase gets skipped. Phase one: sort the list from high to low and calculate the total, the percentage per item and the cumulative point at which eighty per cent of the value is reached. Phase two: determine the real ratio, because Pareto is not a law, and note the actual ratio in the output. Phase three: split the list into three groups rather than two, the vital top, the middle band and the tail. Phase four: give a concrete action per group, literally what you do on Monday. Phase five: check the three pitfalls, survivorship bias, too short a measurement period and hidden costs.

The output is also fixed. Every analysis comes back in the same structure: what I analysed, the real ratio, the three groups each with an action for Monday, three sharp observations the user probably did not see, the pitfalls at play here, and to close, the heading what I could not see, with two to four questions that still need data. Those last two headings make the difference between an answer and an analysis: the skill does not just tell you what it sees, but also what it cannot see.

The file closes with four style rules that shape the character of the output. Never write that it is just 80/20, but calculate the real ratio. Concrete figures over ranges whenever there is something specific to say. Never give an action that depends on more information, but ask for that information first. And the word interesting is banned: every observation must be something a business owner can use without further explanation. How those rules play out in the conversation, you can see below in FIG.03.

After that, using it is simple: paste your list and say what you are measuring. Whatever is missing, it asks for itself. If you get stuck anywhere, the full step by step plan per environment sits in installing Claude skills NL, and the broader explanation of working with AI you will find in the knowledge base.

There are also situations where you are better off leaving it be. A list with no value per item is one: the skill then asks for the metric first, because without a figure there is nothing to sort. A question about the cause of a problem is a second: Pareto tells you where the impact sits, not why. For that question the 5 Whys Analyst was built, which keeps asking until the root of the problem lies exposed. And the most important limit the file names itself in its pitfalls: the analysis is only as good as the measurement period and the metric. One month of revenue says little, and revenue without hours says less than you think.

One more honest limit: the skill makes the choice visible, it does not make it for you. That customer L is almost certainly loss making stands in black and white after the analysis. Whether you part ways, raise the rate or keep them for strategic reasons stays your decision, and the file acknowledges that by asking at the end whether there are items with a value that does not sit in the metric, such as a customer with PR value or a customer who brings in referrals.

A Pareto analysis shows which customers carry your revenue. The next question is where the new enquiries need to come from that grow your vital top. That is where visibility begins: being found by the people looking for you. The free SEO scan shows in a few seconds where your site stands. And if you want to talk further about what else AI can do for your business, from separate skills to full automation, we simply do that in a conversation.