We’re firm believers in using examples to help explain new topics to people. As we continue to build our customer base, we will continue to put out examples like this to provide potential customers examples of how analytics can help increase sales and decrease costs by enabling better decision making. This examples is using real data from a grocery store in Brazil that was made available via Kaggle.
The Business Questions
The grocery store has two questions they’re looking to have answered.
- How have our previous marketing campaigns performed?
- How can we simplify our marketing efforts to be more targeted by providing relevant campaigns and deals our customers will be more likely to act upon?
Their objective in answering these questions is to increase the impact of their marketing budget without needing to increase the budget allocated to marketing.
This grocery store collects a lot of data via their customer loyalty program. While more data is usually better, by no means do you need to be collecting all of this information to get value from this type of analysis. This means they have some basic demographic information about their customers.
- Marital Status
- Number of kids or teens
The loyalty program also enables the grocery store to track purchasing activities across their website, catalog, and in store. The information they have tracked includes:
- Purchase amounts for various categories (Wine, Fruit, Fish, Meat, and Jewelry)
- Where they’re purchasing the products (Website, In-Store, or Catalog)
- Website visits
Lastly, they’ve run their marketing campaigns through their customer loyalty program, meaning they’re able to identify which customers use which coupons from the campaign.
To answer both of these questions, a customer segmentation analysis will be done to provide a view point from the perspective of customer profiles.
Executing an analysis to segment your customers into specific groups is actually quite simple. However, there are a number of easy to forget/miss steps that can give you an inaccurate result. Customer segmentation is usually done with a “clustering” algorithm. There are a number of specific clustering algorithms, however they generally work by measuring the distance between data points. This means the scale of your data needs to be consistent between data elements. For example, if you’re comparing the annual income of a customer to the amount of fruit they purchase each year, their income will be much higher than the amount of fruit they purchase. Without adjusting our data to ensure having a high income and purchasing a lot of fruit, our customer segment results will not be accurate.
Where the major work comes in with customers segments is digesting what each of the segments mean. It isn’t helpful to call a customer segment “Customer Segment 1”, its common practice to evaluate the group and create a “profile” for what the typical person in that profile is like.
Customer Segmentation Results
The output from the algorithm identified 4 customer segment groups. The existing data is then split into the 4 segments and analyzed for trends or unique traits about those customers.
By putting profiles and stories to each of these groups, the grocery store can then be more targeted in their marketing efforts to each of the segments. This part of the process can be time consuming, but also fun to flexible some creative muscles as well!
Group 1: Channel Agnostic Frequent Shoppers
This group spends the most amount of money on average across all the product categories compared to the other segments. They’ll shop in store, online, or via the catalog. They enjoy options for purchasing from the grocery store. This makes them a highly desirable customer from a profit standpoint. They’re also the group that capitalizes on promotional sales the most. They’re overall a highly educated group.
Group 2: Large Family
This group is primarily distinguished by the size of their family. They’re the most likely to have kids and teenagers. Given the size of their family, they generally don’t purchase much from this grocery store, suggesting they’re looking around at potentially other discount grocery stores. We conclude this by their frequent visits to the stores website, but lack of spending money.
Group 3: Highly Educated Couples
This group is the most likely to have an advanced degree (Master’s or PhD) and most likely to be married. They’re the least likely to have teenagers, suggesting if they have children, they’re younger in age. They frequently browse the company website, but don’t spend a lot of money with the grocery store.
Group 4: In Store Frequent Shoppers
Similar to first group, these customers spend a lot of money with the grocery store and also leverages promotional sales as well. However, this group specifically shops in the store vs. online or via the catalog. While profitable, this indicates digital or catalog advertisements won’t make an impact on this group.
Previous Marketing Campaign Results
The other question from the grocery store is how their previous marketing campaigns have performed. We can use the newly created customer segments to understand the performance of the marketing campaigns.
The grocery store has run 6 marketing campaigns in the last 2 years. Overall, they see about a wide range of customer engagement with the campaigns.
Campaigns 1, 3, 4, and 5 were generally successful with customers in all segments engaging with relative consistency between 5-7%. The campaign that performed the worst was 2. This saw engagement rates of around 1%. The most successful campaign was the 6, which saw engagement rates between 7-24%. Within campaign 6, the Highly Educated Couples segment was most likely to engage with the campaign.
The grocery store can learn from these results by not spending time and money on campaigns like 2 in the future. They can also learn that campaign 6 was by far the most successful, particularly with the Highly Educated Couple segment.
By doing an analysis like this, the grocery store can have better insights into their customers and which marketing messages will be more effective with certain groups rather than a “mass marketing” approach by giving all customers the same marketing touchpoints. While some of the insights may not be novel, some are likely to be new. At the very least, businesses can confirm their suspicions and thoughts, increasing their confidence when making decisions.
Is this type of analysis something you think would be interesting for your business? If it is, reach out to us to get a free consultation scheduled!