Data validation

• The dataset had 15000 rows and 8 columns
• customer_id has 15000 unique customer ids without missing values thus matching the description
• nb_sold is a numeric variable with no missing values and matches the data description
• revenue has 1074 missing values that were replaced by the mean
• sales method variable had 5 non unique values with no missing values in the column but had duplicate categories with non unique names so I lumped the categories em + call and Email + Call into one category called Email + Call , also lumbed the category email and Email to one category called Email. at the end i ended up with 3 categories as required by the description
• there are 39 years since the company has been formed so for the variable years_as_customer i removed all rows or observations with years greater than 39 as these where outliers
• week is made up of 6 unique categories that is 1-6 and it matches the description so no data cleaning was needed
• nb_site_visits is a numeric variable with none missing values and does not need cleaning
• state has 50 unique categories and no cleaning was required
• after validating , the dataset remained with 14998 rows and 8 columns

How many customers were there for each approach?

library(scales)
prod_data |> 
  group_by(sales_method) |> 
  summarise(count=n()) |> 
  ggplot(aes(x=fct_reorder(sales_method,count),y=count))+
  geom_col()+
  geom_text(aes(label = comma(count)), position=position_dodge(width=1),
            hjust = 2, vjust=.5, colour = "black", size=3.5)+
  labs(x="sales method")

comments

• the most sales were made through the email (7465 people) followed by call(4961) and lastly email+call(2572 people)

What does the spread of the revenue look like overall? And for each method?

prod_data |> 
  ggplot(aes(x=revenue))+
  geom_histogram(bins=30)

comments

• from the histogram we note that, most sales fell between 40-60 and 80-115 with more sales concentrated on 60 and 90 specifically

prod_data |> 
  ggplot(aes(x=revenue,y=1))+
  geom_boxplot(outlier.colour = "green")

Comments

• the plot suggests that the revenue per each individual is quite spread out ranging from approximately 50 to 200

correlations

## select only numeric values
cor_data<-prod_data %>%
  select(-customer_id) |> 
  keep(is.numeric)
## create a correlation matrix
corl<-cor(cor_data)
corrplot::corrplot(corl,method="color",addCoef.col = "black")

Comments

• number of new products sold seem to increase as the weeks progress , we see that in the high correlation between the two variables
• revenue also significantly increases with number of new products sold since there is a high correlation
• with an increase in number of site visits , number of new products sold also increase and so does revenue to some extent

prod_data |> 
  ggplot(aes(x=sales_method,y= revenue,fill=sales_method))+
  geom_boxplot(outlier.colour = "green")+
  facet_wrap(~sales_method,scales="free")

comments

• the plots suggest that calls were the least productive by looking and the range of sales revenue for the method
• this also concludes our initial hypothesis of why calls had a smaller percentage of total revenue

proportion of revenue

prod_data |> 
  group_by(sales_method) |> 
  summarise(tot=sum(revenue)) |> 
  mutate(perc=tot/sum(tot)) |> 
  ggplot(aes(x=fct_reorder(sales_method,perc),y=perc))+
  geom_col(fill=ggthemes::tableau_color_pal()(1))+
  geom_text(aes(label = scales::percent(perc)), position=position_dodge(width=1),
            hjust = 2, vjust=.5, colour = "black", size=3.5)+
  labs(x="sales method",y="Proportion")

comments

• the plot suggest that calls were the lowest revenue generating activity ,making up only around 17% of the total revenue produced this period
• Email + call accounted for around 31% which was better than calls alone
• Emails was responsible for the greatest proportion of revenue

Was there any difference in revenue over time for each of the methods?

prod_data |> 
  mutate(week=as.numeric(week)) |> 
  group_by(week,sales_method) |> 
  summarise(tot_revenue=sum(revenue,na.rm=T)) |> 
  ggplot(aes(x=week,y=tot_revenue,color=sales_method))+
  geom_line()+
  scale_x_continuous(breaks=1:6)+
  labs(y="revenue",color="method")

comments

• we can note that email produced the highest amount only in the first week which accounted to a value of 247519.25. revenue gradually decreased over the next weeks
• Email + call method started much lower but ended the period much higher than other methods claiming the top spot
• the calls method started out lower and ended up lower than other other methods,there was litle to no change in the revenue generated from this method

prod_data |> 
  mutate(week=as.numeric(week)) |> 
  group_by(week,sales_method) |> 
  summarise(average_revenue=mean(revenue,na.rm=T)) |> 
  ggplot(aes(x=week,y=average_revenue,color=sales_method))+
  geom_line()+
  scale_x_continuous(breaks=1:6)+
  labs(y="average revenue",color="method")

comments

• the graphs desplay average revenue for each sales method over time.
• Email + call have the highest or up above in terms of the trend in average revenue over time
• Email Alone comes in second then calls alone comes last

prod_data |> 
  mutate(week=as.numeric(week)) |> 
  group_by(week,sales_method) |> 
  summarise(average_revenue=mean(revenue,na.rm=T)) |> 
  group_by(week,sales_method) |> 
  summarise(`maximum average`=max(average_revenue)) |> 
  arrange(desc(`maximum average`)) |> 
  head(n=7) %>% 
  gt() %>%
  tab_header(title = md("** rows showing maximum values**"))