Recipe for a Pareto Analysis

[hic]

“Which products contribute to the first 80% of our turnover?”

[hic]

This type of question is common in all types of business intelligence. I say “type of question” since it appears in many different forms: Sometimes it concerns products, but it can just as well concern customers, suppliers or sales people. It can really be any dimension. Further, here the question was about turnover, but it can just as well be number of support cases, or number of defect deliveries, etc. It can in principle be any additive measure.

It is called Pareto analysis. Sometimes also known as 80/20 analysis or ABC analysis.

The logic is that you first sort the products according to size, then accumulate the numbers, and finally calculate the accumulated measure as a percentage of the total. The products contributing to the first 80% are your best products; your “A” products. The next 10% are your “B” products, and the last 10% are your “C” products.

And here’s how you do it in QlikView:

1. Create a pivot table and choose your dimension and your basic measure. In my example, I use Product and Sum(Sales).
   
   
2. Sort the chart descending by using the measure Sum(Sales) as sort expression. It is not enough just to check “Sort by Y-value”.
   
   
3. Add a second expression to calculate the accumulated sales value:
        RangeSum(Above(Sum(Sales), 0, RowNo()))
   Call this expression Accumulated Sales. The Above() function will return an array of values – all above values in the chart – and the RangeSum() function will sum these numbers.
   
   
4. Create a third expression from the previous one; one that calculates the accumulated sales in percent:
        RangeSum(Above(Sum(Sales), 0, RowNo())) / Sum(total Sales)
   Format it as a percentage and call it Inclusive Percentage.
   
   
5. Create a fourth expression from the previous one; one that calculates the accumulated sales in percent, but this time excluding the current row:
         RangeSum(Above(Sum(Sales), 1, RowNo())) / Sum(total Sales)
   Format it as a percentage and call it Exclusive Percentage.
   
   
6. Create a fifth expression for the ABC classification:
         If([Exclusive Percentage] <= 0.8, 'A', If([Exclusive Percentage] <= 0.9, 'B', 'C'))
   Call this expression Pareto Class. The reason why the Exclusive Percentage is used, is that the classification should be determined by the lower bound of a product’s segment, not the upper.
   
   
7. Create a conditional background color, e.g.
         If([Pareto Class] = 'C', LightRed(), If([Pareto Class] = 'B', Yellow()))

You should now have a table similar to the following. In it you can clearly see the classification of different products.

In this table, there are five different expressions that you can use for Pareto analysis. The graph in the beginning of this post uses Sales and Inclusive Percentage for the bars and the line, respectively; and Pareto Class for the coloring of the bars.

Further, you may want to combine the Pareto Class and the Exclusive Percentage into one expression:

     Pareto Class =
         If(RangeSum(Above(Sum(Sales),1,RowNo())) / Sum(total Sales) <= 0.8, 'A',
         If(RangeSum(Above(Sum(Sales),1,RowNo())) / Sum(total Sales) <= 0.9, 'B', 'C'))

Good luck in creating your Pareto chart.

HIC

Further reading related to this topic:

Recipe for an ABC Analysis

Recipe for a Pareto Analysis – Revisited

[#3]

Great post as always. I like the use of a background colour for each Pareto class.

[#4]

is there any way to count number of elements in A-class and number of elements in b-class ? and if i want to put this formula in a textbox ?  thanks

[thanstad]

Could this work If ([Exclusive Percentage] <= 100,count(Row)) ?

I made a count in the script instead, that seems to be more correct to use as sum of xxx as a expression.

/Tormod

[#6]

Very useful and I have been using this already. However how does this work in a straight table or a pivot with more dimensions? The percentages will go wrong when I try this. Basically I want more dimensions for more detailed overview of the products (example including brands)

[ebrambilla]

Thank you, very useful explanation!

[#8]

Has anyone a idea for a pivot chart with more than one dimension? I try to represent an ABC analysis with ABC as the first dimension and for example the goods as the second dimension.

[#9]

In the Running Total for Accumulated Sales, the total is adding the absolute value of the sales. I have some credits where there would be line items that are negative. Is there a way to exclude the negative values or have them properly calculated in the running total so they are not treated as an absolute value?

[#10]

What`s about using the fabs expression?

[#11]

This is just what I needed. Thank you!

[#12]

Please advise me how to get and display the totals separately for each Pareto class?