Pandas
Numpy
Feature Engineering
Pivot Tables
Functions
Filtering
EDA
Matplotlib
Hypothesis Testing
T-Test
The purpose of this project remains to analyze the data provided by the telecom operator Megaline. With an offering of two plans, Surf and Ultimate, the goal of this project is to determine optimal capital allocation. We will determine which plan brings in more revenue. This will result in an adjustment of the advertising budget, as a means to further increase revenue. The dataset provided is a sample of the population of Megaline customers, across different cities in 2018. We will conduct further analysis on the client behavior, as well as look at other important insights found in the data.
Initial thoughts suggest the Surf plan would bring in more revenue, as the overage charges, combined with the limited plan allotment, would lead to many customers paying fees. The Ultimate plan is more than double the price of the Surf plan, and the company lacks and middle tier plan. As a result, we hypothesize that the Surf plan would be far more popular than the Ultimate plan, further contributing to the differences in revenue. Yet another factor could be the overages charges on the Ultimate plan, as they are far lower than those of the Surf plan. We expect to see differences in plan preference based on age, as well as revenue when looking across age groups.
The data shows statistical differences in mean revenue among the two plans, as the Ultimate plan brings in more revenue. Our significance level was set to 5%, and our p value was much higher. In simpler words, we reject our null hypothesis that the mean revenues were similar.
The data shows us that capital allocation to marketing the Ultimate plan would likely yield a better cash on cash return, not based on popularity, but on revenue. As the Surf plan is more popular, new customers should be lead to the Ultimate plan instead. We saw many Surf customers would experience overages on their plan. These would be the prime customer base to push towards the ultimate plan.
We see that the mean revenue of customers in New York appears to be similar to that of all the other cities combined. Yet, Honolulu, Albany, and Colorado Springs are the cities with the highest average revenue. A marketing push may also be a good idea in those areas, to further increase revenue, while also considering market saturation. We did not see a preference of plans of customers of different age groups, as most preferred the Surf plan.
Overall, the Ultimate plan is not very popular. As such, maybe it would be beneficial to test a middle tier plan, in order to capture customers who may be dismayed by the gap in plan prices. Another method that would lead to increased revenue would be to slightly increase the overage fees on the Surf plan. Yet, a smart revenue strategy remains in rounding up minutes, and more substantially, rounding up data used to the nearest gigabyte. Data usage appears to be the largest contributor to revenue.
Finally, Hypothesis testing suggests the mean of the call durations and number of messages were not different. On the other hand, internet traffic is different, when conducting statistical tests on the means.