Investigation Overview¶
Working as a data analyst for an online food ordering and food delivery platform, I've been told by the Chief Marketing Officer that recent marketing campaigns have not been as effective as they were expected to be. I will analyze the data set to understand this problem and propose data-driven insights and recommendations.
Dataset Overview¶
The dataset consists of 2,240 customers of an online food ordering and food delivery platform. The dataset contains data on:
- Customer profiles
- Product preferences
- Campaign successes/failures
- Channel performance
1. Campaign Success¶
Generally, campaigns have been unsuccessful. Only the most recent campaign broke above a 10% success rate at 14.8%. Others have ranged from "underwhelming" at around 7% success to "complete failure" with campaign 2's 1.3% win rate. I'd recommend analyzing what differentiated the recent successful campaign from unsuccessful campaign 2, then shaping future campaigns with this insight.
Further segmenting by country, we can see how poorly campaign 2 did in Australia and America. That is worth investigating. As with campaigns 1, 3, and 4, Australia had been a country of little success. This has changed more favorably in campaigns 5 and 6, however. What changed?
Special attention should be paid to India as campaigns 5 and 6 underperformed. Why were Indian customers not receptive to the most recent campaigns? India was the most receptive in campaign 3. I'd recommend looking at the recent successes in Spain and South Africa then adapting those strategies to the Indian market.
2. Sales Channels¶
The best performing channel was in-store and the worst performing were deals and catalog purchases. There were many customers that only made 1 or even 0 deals and catalog purchases. In-store purchases accounted for nearly 40% of purchases and online purchases—although higher than that of deals and catalogs—accounted for only 27%. Despite being an online food delivery platform, the average customer only made 4.1 purchases over the two-year sample of data.
I'd recommend focusing on web and mobile advertising (social media, email newsletters, smartphone apps, influencers, etc) to better cultivate the online sales channel. I'd advise the CMO to shift away from legacy paper modes like catalogs unless they were targeting an older demographic. It should be noted that this data sample, collected in 2015, was taken pre-Covid-19 so any advice may not be as relevant.
3. Purchasing Trends¶
Wines and meat products are the top 2 best performing products in terms of sales. Customers on average spent \$303.50 on wine, amounting to over 50% of the total share of purchases. Meanwhile, staples like fish, fruits, and sweets all fell well below the average of $100.80 per customer.
I'd recommend using regular sales and promotions to increase the share and average amount spent on these products. Such campaigns may find more success if they were targeted towards families with more than 1 dependent, as I'll explore further next.
4. Number of Dependents¶
The sample of customer IDs averaged 1.0 dependents with most customers having at least 1 dependent. 0 dependents are the next largest bin with about 600 IDs. When examining the correlation between the number of dependents and key numerical metrics like total amount spent and the total number of purchases, they are negatively correlated. Meaning, the more dependents a customer has the fewer and smaller purchases they'll make. This may be due to the trend that customers with more children tend to have lower incomes, at least within this sample.
Although it may stand to reason that those with higher income would spend more, as seen in the bottom scatterplot, left unaddressed this is a wasted opportunity given that the majority of our customers have 1 or more children. This group should be targeted while maintaining sales of the higher-earning, higher-spending "0 dependent group". I would suggest splitting up campaign efforts into two directions: one direction targeting the "0 dependent group" and another targeting families. For the latter, focusing on fish, fruit, and sweets and other lower-cost goods may be more productive. Wine, gold, meat, and the like can be marketed more to the "0 dependents group".
