Let us start with a few background of the objective.
There are 3 actions that a user can perform on the website namely - View Example Cards, Design Custom Card and View Pricing Page. The objective is to identify which action among these 3 most affects the Revenue (since revenue is the most important metric identified by the analytics team).
The following would be the methodology to deduce the action which most impacts revenue :-
1. Feed data on Excel and then plot the 3 vs. revenue to see the generic trend
2. Find mathematical correlation between each of the actions and the revenue; Correlation shows the degree in which the two variables are linearly related/dependent [it is a value from -1(most negatively correlated) to +1(most positively correlated)]. Find the action that is most 'correlated' to revenue by the magnitude of the correlation value (can be +ve or negative).
The steps mentioned below in detail:
Step 1
A. Copy data in excel with all columns in the above table as separate columns in cells B3:G16
The table looks like below -
Month | Sessions | View Example Cards | Design Custom Card | View Pricing Page | Revenue |
March | 1695292 | 17,758 | 15,052 | 25,964 | $ 1,14,43,801 |
April | 1670136 | 20,146 | 15,008 | 19,231 | $ 1,14,73,438 |
May | 1868864 | 22,857 | 18,408 | 9,704 | $ 1,40,37,828 |
Jun | 1727159 | 17,394 | 17,236 | 13,877 | $ 1,32,35,195 |
July | 1781272 | 17,926 | 13,523 | 7,136 | $ 1,04,65,325 |
August | 1956408 | 24,360 | 14,135 | 14,721 | $ 1,06,41,739 |
September | 1930809 | 20,046 | 13,016 | 17,226 | $ 1,00,93,305 |
October | 2367562 | 26,298 | 14,223 | 19,816 | $ 1,09,33,293 |
November | 2568950 | 30,039 | 15,517 | 19,671 | $ 1,18,72,900 |
December | 2741061 | 28,246 | 14,284 | 7,613 | $ 1,07,13,635 |
January | 2485498 | 27,426 | 15,846 | 12,973 | $ 1,17,19,280 |
February | 2202975 | 24,113 | 13,149 | 9,586 | $ 1,00,22,618 |
March | 2593533 | 27,960 | 13,950 | 14,277 | $ 1,03,55,119 |
B. Then we start plotting the actions with revenue by following - select the column say view example cards (D) and then press control and then select the column Revenue (G). Then go to Insert and under the chart select X-Y scatter to plot the below-
Repeat for other two actions.1. Design Custom card
2. View Pricing Page
Inference: It seems that view example cards and view pricing page are not related / weakly related to revenue and Design Custom Card is strongly,positively and linearly related to Revenue (as the Design Custom Card value increases the revenue increases accordingly).
Step 2 : Now we calculate the correlation of the 3 actions with the revenue by using the formulas below and typing them in cells D19, E19 and F19.
View Example Cards | Design Custom Card | View Pricing Page | ||
Correlation with Revenue | =CORREL(D4:D16,$G$4:$G$16) | =CORREL(E4:E16,$G$4:$G$16) | =CORREL(F4:F16,$G$4:$G$16) |
View Example Cards | Design Custom Card | View Pricing Page | ||
Correlation with Revenue | -0.158795 | 0.990788 | 0.021484 |
On formatting, The results look like below where green shows highest correlation. (values are formatted as %)
View Example Cards | Design Custom Card | View Pricing Page | ||
Correlation with Revenue | -15.87948% | 99.07878% | 2.14841% |
The above correlation table shows that Design Custom card influences the Revenue to the maximum extent in a positive manner.
Thus it can be concluded that the focus of redesign should be The Design Custom Card page which would most positively impact the revenue.
5) Consider the following fictional scenario: A mid-sized firm has hired Viget to redesign its website....