Method 02 · Prioritise
Pareto analysis
A Pareto analysis puts your categories in order of size and shows which small group causes the bulk of the problem, so that you go and work on the right one.
The name comes from Vilfredo Pareto, who noticed around 1900 that a small share of the population owned most of the land. Joseph Juran turned it into a working method for quality management and called it the vital few against the trivial many.
The 80/20 ratio is an observation, not a law. Sometimes it is 70/30, sometimes 95/5, and sometimes the load is spread evenly and there is no clear Pareto distribution. That last case is a result too: it means you cannot prioritise this way and will have to do something else.
When to use it, and when not
| A good fit when | A poor fit when |
|---|---|
| You have many separate incidents and too little time to investigate them all. | You are dealing with a single incident. There is nothing to count. |
| The incidents fall into categories that do not overlap. | Every incident is unique. You end up with fifty categories of one. |
| You can measure the impact, not just the count. | You only have a feeling about which category is worst. |
| You need to justify a choice to other people. A Pareto chart is persuasive. | The choice is already made and you are shopping for a picture. |
What you need
- Data over a period. Long enough to average out chance, short enough to still be current. A quarter is often a good start.
- Categories that exclude one another. Every incident belongs in exactly one category, or the same problem gets counted twice.
- A measure that reflects the damage. Counts are the easiest and the least useful; see below.
How to run it
-
Choose what you measure
Counts, money or lost time. This is the most important decision in the whole analysis, and it is usually made by accident, by counting whatever is easiest to count.
-
Settle your categories
Five to fifteen. Fewer and there is nothing to choose between; more and the tail grows so long that the chart stops saying anything.
-
Count
Over a fixed period, with a fixed definition. Write down which period it was, because in six months somebody will want to know whether things got better or worse.
-
Sort descending and add up cumulatively
The cumulative line shows where the eighty per cent sits. The categories up to and including that mark are your vital few.
-
Check that it is a Pareto at all
If the load is spread evenly there is no small group to attack. See whether the categories can be cut differently; if they cannot, prioritising this way is not the tool for this problem.
-
Narrow your problem statement and start again
The result is not a cause but a problem area. Restate your problem down to that one category and run your root cause analysis on that.
A worked example
A service desk counts a quarter’s tickets and wants to know where improvement should start.
| Category | Tickets | Average handling time | Total hours |
|---|---|---|---|
| Forgotten password | 420 | 0.2 h | 84 |
| Printer not working | 180 | 0.5 h | 90 |
| VPN drops | 95 | 1.5 h | 143 |
| Slow laptop | 60 | 4.0 h | 240 |
| Other | 145 | 0.6 h | 87 |
On counts the winner is obvious: passwords, with nearly half of all tickets. On hours lost, passwords come last and slow laptops are by far the biggest item, at nearly a third of the time.
The same data, two different answers, and both defensible. If you want to lower the load on the service desk, tackle passwords. If you want to lower the time the organisation loses, tackle laptops. What you must not do is draw the first chart and then answer the second question.
Note the “other” category as well. At 145 tickets it comes third, which is a sign that the breakdown is not finished: something in there deserves a category of its own.
Common mistakes
- Counting whatever is easy to count. Counts are in every system; impact has to be worked out. That is exactly why people prioritise on counts and end up working on the wrong thing.
- An “other” category that finishes high. That is not a finding but a sign that your breakdown is not done. Split it.
- Overlapping categories. If an incident fits in two buckets, the same problem is either counted twice or half lost.
- Drawing the chart and narrowing nothing. The result is supposed to change your problem statement. If it does not, you made a picture.
- Mistaking the result for a cause. “Slow laptops” is not a root cause; it is the place where you are going to go and look for one.
Combining it with the other methods
- Is / Is-not and Ishikawa come after this, on the chosen category. The Pareto tells you where to look, not what you will find.
- 5 Whys comes last, once you have a concrete cause to keep asking about.
- Repeat the Pareto six months later with the same categories. The difference between the two charts shows whether your countermeasure worked.
Frequently asked questions
Does it really have to be 80/20?
No. Eighty per cent is a customary place to draw the line, nothing more. You can move it on the worksheet.
What if the distribution is flat?
Then there is no small group causing most of it and prioritising this way will not help. Check your categories first: a flat distribution often comes from a breakdown that is too fine or too coarse.
How many categories should I have?
Five to fifteen. Below four there is nothing to choose between; above twenty you get a long tail of singletons that makes the chart unreadable.
Why can I not send the result straight into 5 Whys?
Because a Pareto category is a problem area, not a cause. Putting it in as a first why would start the analysis from something that is not an answer to a why question at all. The right next step is to narrow your problem statement to that category, and begin again.
Can I compare two periods?
Run them as two separate analyses with the same categories and put the charts side by side. Cramming two periods into one chart makes it harder to read than it needs to be.