I’m excited to share my journey with Data Career Jumpstart as I dive into the second module, focused on mastering data analysis with Microsoft Excel. This hands-on experience is a fantastic opportunity to develop my analytical skills through exploration and reporting.
Case Study & Problem Overview
For this case study, I am taking on the role of a Data Analyst for a food delivery company. My primary goals are to strengthen the company strategy by analyzing customer behavior and examining the distinctions between those who engage with campaigns and those who do not. By uncovering these insights, I provide actionable, data-driven perspectives that inform marketing and business decisions.
- Understand the key data elements.
- Identify valuable business insights and opportunities.
- Propose data-driven strategies that will improve the effectiveness of marketing campaigns and create real value.
Strategic Recommendations & Action Plan
This section provides strategic recommendations based on insights from the case study. However, we need to exercise caution in making these recommendations because we only have access to correlations in the available data. There is no information indicating whether previous campaigns effectively targeted their audiences. For instance, we lack a clear understanding of whether the trend in purchase amounts is influenced by specific campaigns. There is no trend data available regarding when each of the six campaigns was conducted or the spending patterns of customers following each campaign. Based on the information we do have, here are the best recommendations I can provide.
- Suppose the previous campaign did not specifically target a particular income level within the customer audience. In that case, the next campaign should focus on higher-income customers, for example, those earning over $60,000. However, if the previous campaigns already targeted high-income customers, the current data may reflect that influence. In this case, it might be beneficial to shift the focus to lower-income customers to see if we can achieve a better acceptance rate, similar to the reasonable acceptance rate we experienced with higher-income customers.
- If the previous campaign did not specifically target households with children, the next campaign should focus on customers without children at home. However, suppose the previous campaigns targeted customers without children. In that case, it might be more effective to shift the focus to customers who do have children, especially those with at least one teenage child, as they make up the majority of customers with children.
- Based on the performance of the shopping channels, if the previous campaign did not specifically target audiences using the catalog, the next campaign should focus on this channel. However, if the earlier campaigns did target catalog audiences and successfully increased purchase frequency in this area, the recommendation is to shift the focus to web audiences. The physical store already has the highest absolute purchase frequency among customers.
- Customers who accepted campaigns tend to make more purchases with deals. Therefore, the recommendation is to run campaigns that include deals.
To further enhance the analysis, we recommend several action plans steps. These include:
- Comparing customer behavior across previous campaigns
- Collecting more detailed data on number of product purchase
- Obtaining campaign dates to help measure changes in purchase volume and spending following each campaign.
- Access to each campaign’s audience criteria to enable deeper, more targeted insights.
Together, these recommendations and action items provide a clear roadmap for informed decision-making and continuous improvement. The following section outlines the analytical process that guided these recommendations, illustrating how data exploration and insight extraction shaped the overall strategic direction.
Analytical Process & Methodology
This section outlines the analytical process used to develop the recommendations presented above. All analysis was conducted in Excel, covering data cleaning, exploration, and visualizations. We will first understand the data, then analyze the differences in behavior between customers who accepted the campaigns and those who did not
Getting to Know the Data
The dataset used in this project reflects food-delivery operations. It has been adapted and further modified for instructional purposes. The provided data is an Excel file containing 2,205 entries of customer information, featuring 36 attributes, including the following:
- Customer’s yearly income
- Total amount spent at the store by each customer
- Number of young kids and teenagers in the home
- Number of days since the last purchase
- Amount spent on various types of products, such as wine, fruit, meat, fish, sweets, gold products, and regular products
- Number of purchases with deals
- Number of purchases made through different shopping channels, including the website, catalogues, and physical stores
- Number of visits to the website in the last month
- Whether the customer accepted offers in the first to sixth campaigns
- Customer complaints in the last two years
- Age of the customer
- Number of days since the customer first joined and the date joined
- Marital status
- Education level
Of the 2,205 records we received, we used Excel’s built-in duplicate-checking feature to identify 184 duplicates, leaving us with 2,021 clean records. Upon reviewing the data in each column, we discovered that 3 rows had purchase amounts for regular products that were negative. We considered these figures anomalies, since customers in these 3 rows had not participated in any campaigns or deals that could have reduced their purchase amounts. Consequently, we decided to remove these rows, reducing our total to 2,018 records. We then conducted an initial analysis of customer backgrounds and join dates, which is illustrated in the visualizations below.
Campaign Response Analysis
We are comparing two categories to understand their behaviors. Customers who accepted any campaigns accounted for 28% of the records, while those who did not accounted for 72%. Although the two groups differ in size, our focus is on the patterns within each group. The difference in record counts does not impact the insights we can gain. The distribution of customer behavior by background and join date is shown in the visualizations below.
We will then compare the two groups to identify differences in their shopping behaviors.
Compare Shopping Habits
Key Insights from Campaign Response Analysis
- Two key factors that differentiate customers who accepted the campaign from those who did not are income and the presence of children:
- Income: Customers who participated in the campaign tend to have significantly higher incomes. Specifically, 57% of those who accepted the campaign earn $60,000 or more. Also, income has a stronger effect on total purchases among customers participating in the accepted campaign.
- Presence of children: Generally, households with children are much more prevalent in the group that did not accept the campaign. This trend is consistent across all age groups of children, including young kid, teens, and mixed-age groups. The group that accepted the campaign has fewer households with children in each category, particularly among those with young kids.
- The product with the most significant purchase difference between customers who accepted the campaign and those who did not is: wine (by item) and regular-tier products (by tiered product line).
However, we only have data on the total amount spent on these products and do not have information on the quantity sold or their respective prices. Therefore, despite observing a notable amount spent on wine and regular-tier products among customers who accepted the campaign, we will not include these product categories in our recommendations. - The number of catalog purchases is noticeably higher among customers in the accepted campaign.
- Customers who accepted the campaigns made more frequent purchases with the deals and filed fewer complaints.
