Analyzing three different sales strategies for a new line of products

A new line of office stationery products focused on items to enhace brainstorming was launched and three different sales strategies were used for this over the last six weeks: targeted email and phone calls, as well as combining the two. Email: Customers in this group received an email when the product line was launched, and a further email three weeks later. This required very little work for the team. Call: Customers in this group were called by a member of the sales team. On average members of the team were on the phone for around thirty minutes per customer. Email and call: Customers in this group were first sent the product information email, then called a week later by the sales team to talk about their needs and how this new product may support their work. The email required little work from the team, the call was around ten minutes per customer.

The goal is to provide compelling information to the sale department by answering this questions:

• How many customers were there for each approach?
• What does the spread of the revenue look like overall? And for each method?
• Was there any difference in revenue over time for each of the methods?
• Based on the data, which method would you recommend we continue to use? Some of these methods take more time from the team so they may not be the best for us to use if the results are similar.

The data set used contains 15000 records, an initial validation is described below :

Data Validation The dataset contained 15000 rows and 8 columns before validation. Each variable in the dataset was verified against the criteria in the dataset information:

• week: 6 weeks, numbered as weeks since the product launch from 1 to 6 with no missing values. No cleaning was needed.
• sales_method: 5 methods were found before cleaning, standarization of category’s name was requiered, resulting in the 3 methods. In lower case ‘email’, ‘calls’, ‘email + call’.
• customer_id: 15000 unique values with no missing values. No cleaning was needed.
• nb_sold: Numeric values with no missing values as described. No cleaning was needed.
• revenue: Numeric values with 1074 missing completely at random values, since it represents a 7% of this column data, all of the records with NAs in revenue had an associated number of items bought, and we want to avoid loss of data, replacement of NA’s with the median of the total distribution of revenue was peformed. Median imputation resulted in 15000 numeric values, all of them rounded to 2 decimals and no missing values.
• years_as_customer: Up to date in 2023, a client couldn’t be a customer for longer than 39 years since foundation of the company in 1984.Two anomalous values were found, and after checking distribution of values, median imputation was performed to replace them. It resulted in 15000 numeric values ranging from 0–39 and no missing values.
• nb_site_visits: 15000 numeric values with no missing values as described. No cleaning was needed.
• state: 50 states with standarized characters and no missing data. No cleaning was needed.

After cleaning and validation,the dataset contains 15000 rows and 8 columns with no missing or anomalous values.

How many clients were there for each approach?

After six weeks, it’s evident that the email approach has attracted nearly half of the total registered customers, with the call approach and the combination of email and call following behind in terms of customer acquisition.

Aditionally, we can see that the distirbution of total number of customers across the country, is pretty similar for each method. Regardless of the approach utilized, California(CAL) holds the highest total, followed by Texas(TX) and New York(NY); this is something you might be expecting already, since those are probably the most crowded states in the nation.

The analysis shows that regardelss of the state, e-mail approaching was in general, the best method at getting clients, with almost 7500, which is pretty much the sum of clients in calls and email+calls.

Now, it is clear that the e-mails and calls appear to be the best at getting large amount of clients, however, more clients does not necesarily translate into bigger revenues, and also, the type of method that worked for some clients, could provide information on how expensive the purchase is going to be; hence, we need to review it the revenues separately.

How much revenue did we manage to earn using each method?

Due to the volume of customers acquired through each method, we had a clear expectation that the email approach would yield the highest total revenue after six weeks. However, despite calls alone attracting a greater total customer count compared to the email+call method, the latter generated higher revenue. This suggests that customers apporached with e-mail+call expent more money.

What does the spread of the revenue look like overall?

The revenue in the last six weeks apporaches a right skewed bi-modal distribution indicating two peaks, one of them is around USD 50 and the other one relies over USD 87, which is basically the median revenue selected as a central tendency value and imputed during cleaning (USD 87.5). The standard deviation of revenue is 45.7, and the density plot is wide, suggesting a wide spread of revenue between each sale. Although the majority of sales (~75%) lies below USD 106, there are a few clients who purchased more than USD 185.62 and can be conisdered outliers in the distribution (red points).

In general, the revenue spread among clients shows a high variability, with a median revenue of USD 87.5 and some outliers indicating extraordinary revenues over USD 185.62 per client.

What about each method?

After six weeks, when comparing the three methods, we notice that the email+call approach, although attracting fewer total clients than the other methods, manages to secure more substantial purchases from them.

To illustrate, the highest revenue generated from sales closed through phone calls is now comparable to the lowest revenue generated by the email+call approach. Meanwhile, the email-only method maintains a middle position, with sales ranging from USD 78 to 149 per transaction.

At this point, we know that the e-mail approach attracted the largest number of customers and also generated the highest total revenue (USD 721,006). However, a composed method e-mail + call, yields the highest median revenue per sale (~USD 182), which indicates that customers in this group tend to expend more. This relationship suggest that the first method is pretty effective at reaching a larger colume of customers, resulting in higher overall revenue. Meanwhile, the composed method, is more effective at maximazing revenue from each customer it acquires, which might be due to the personal touch and the opportunity of directly explain the benefits of the product during a call.

Was there any difference between the revenue over time for each method?

Despite initially generating higher total revenues compared to the other methods, the email approach exhibited a decline in performance over time. It started strongly in week 1 but ended up at the bottom of the rankings after week 6, suggesting a significant drop in its effectiveness. In contrast, the calls method displayed remarkable consistency, maintaining sales at approximately USD 27,000 in the first week, reaching a peak value around USD 56,000 by the fifth week, and concluding at USD 30,000 in week 6.

Lastly, the combined approach of email+calls showed a substantial increase in revenues over time. It began at a lower point with approximately USD 18,000 and steadily grew, achieving a weekly total of around USD 120,000 after the sixth week.

Which method should the team continue to use?

In a nutshell, despite the e-mail apporach generated the largest revenue after six weeks, it seems not to be really effective in the long run; the calls are rather steady over time, and the combination of both is showing great potential, specially because it takes shorter than calls alone, generate more revenues and unlike e-mail, the numbers are held or even improving over time; thus, if I were to recommend one method, that would be the email+calls approach.

Bussiness metrics

When evaluating the various methods for launching our new product line, I suggest tracking the percentage of total revenue over time that each method contributes. This approach will help us understand the consistency of each method’s performance.

For instance, let’s take the email+calls method as an example. In week 1, it represented approximately 10% of the total revenue. However, this percentage increased significantly to around 70% of the total after six weeks. This shift indicates that following up an informative email with a phone call resulted in attracting more clients and generating higher revenues.

Recomendations

In the coming weeks, I suggest the following actions:

• Concentrate on the email+call method and use the suggested metric to track how revenue evolves over time.
• Cease using the calls-only method as it consumes considerable time and generates relatively low, consistent revenue.
• Despite the digital age, it appears that traditional methods still resonate with people for brainstorming. Prioritize products that enhance this activity.
• Given the striking similarity in average revenue per sale across all states, direct your focus toward the most populous areas, where you’re likely to attract a larger client base.
• Pay close attention to data collection, particularly regarding missing revenue data. This is critical for making informed decisions, especially because revenue is a vital variable in our dataset. While it’s likely due to a data entry error, comprehending this pattern will assist us in devising strategies to address it in the future.