# Method for determining the root cause behind a drop in business performance

What methods are there for determining the root cause behind a drop in business performance? My question is best explained with a very simple example...

There are two Widget Factories (A and B).

Each factory has an output ($$y_X$$) given by the product of number_of_workers ($$N_X$$) and avg_production_rate_per_worker ($$r_X$$):

$$y_X = N_X \cdot r_X$$

where $$X$$ refers to the factory, and the total output is:

$$y = y_\mathrm{A} + y_\mathrm{B}$$

From week 1 to week 2, the total output decreases from 1,000 to 500.

Is there a way of determining the root metrics/dimensions behind this drop? i.e., has $$y$$ gone down due to $$N_\mathrm{A}$$, or $$N_\mathrm{B}$$, or $$r_\mathrm{A}$$, etc.?

I have data in the following form:

                 Number of  Avg. Production   Total
Week   Factory   Workers    Rate per Worker   Output
----------------------------------------------------
1      A         10         10/week           100
2      A         15         10/week           150
...    ...       ...        ...               ...
1      B         10         10/week           100
2      B          5         15/week            75

• For example, a drop in $y$ may be due to a: 1) local change - a single metric in one factory decreasing, e.g.: a) a bout of flu meaning many staff are off sick (reduces number_of_workers), b) automated production line breaks meaning production has to be done manually (reduces avg_production_rate) 2) global change - a single metric over both factories decreasing, e.g.: a) a national holiday meaning more staff off due to annual leave (reduces number_of_workers) or b) a new company-wide QA process meaning more checks are added (reduces avg_production_rate)
– Ben
Jul 26 at 11:22
• Any ideas? Can Sensitivity Analysis help? Can I change data to logs and use Structured Equation Modelling?
– Ben
Jul 26 at 12:56
• Graph models perhaps?
– Ben
Jul 26 at 13:28