A Taste of Offers at Starbucks: Analyzing which offers work for different customer demographics
A deep dive data analysis into one of Starbucks’ products
Everybody loves offers and businesses all over the world know that. Offers such as discounts, bogos and rewards play a huge part in increasing customer loyalty. Starbucks — the world’s largest coffeehouse chain, just like any other business frequently gives offers to its customers in form of discounts, informational ads and BOGOs(Buy One Get One free).
As part of the capstone project for my Udacity Data Scientist Nanodegree program, I decided to take on the challenge of analyzing simulated data from Starbucks that mimics customer behavior for one of its products. The data consists of offer metadata, customer demographics(age, gender, income) and transaction data from customers. The main objective of the data project is to obtain business insights on how customer demographics influence the kind of offers that they complete. ‘Completing’ an offer here means that the customer actually got the offer after spending the minimum required amount to activate the offer. To accurately get customers that completed an offer because of an ad, the analysis takes into account to only study customers who both viewed & completed an offer. This is because it is possible for a customer to complete an offer without ever viewing it. Moreover the analysis goes further to get some insight on key drivers of the amount spent on transactions for this product.
So what kind of offer characteristics would a middle-aged man with an income of about 70,000 likely complete? We could try to answer that below.
How do demographics influence the type of offer given?
There are only 2 types of offers that a customer can complete for this product; bogo and discount.
Type of Offer Vs Age of customer
We see that generally there is no major preference for the type of offer across all the age brackets. The only small noticeable difference is for age bracket (90–100) where about 47% of the offers were discounts while for other brackets it was~50% –54 % discount type and above. Generally discounts are more popular than bogos for majority of the age brackets.
Type of Offer Vs Income of customer
With income we observe an interesting pattern: the highest income earners prefer bogos over discounts! Bogos contributed to ~76% of the offers for customers in the income bracket (120,000 -130,000). This was in contrast to the other income brackets which were more evenly distributed over the type of offers.
Type of Offer Vs Gender of customer
Generally no major preferences for type of offer for both genders although males seem to slightly prefer discounts over bogos while for females the distribution is really uniform over the two types.
How do demographics influence the channels that a customer uses to complete an offer?
The channels that an offer is advertised on are email, mobile, social or web. Notably, an offer can have multiple types of channel where it is advertised.
Channels Vs Age of customer
Generally offers that are given in all the 4 channels available i.e (web,email,mobile,social) are most popular across all age groups taking more than 60%
Channels Vs Income of customer
Again when looking at the channels vs the income brackets of the customers, offers given in all the channels (web,email,mobile,social) are most popular. We can also observe that offers given in the channel combination of (email, mobile, social) seems to be a bit popular with the highest income earners (120,000 –130,000) with 30% compared to other channels which have less than 15%
Channels Vs Gender of customer
It is a similar story when looking at gender vs channels with the offers given through all channel combinations (web,email,mobile,social) taking up more than 60% of the offers in both genders.
How do demographics influence the customers completing offers of various duration
An offer usually has a duration for which it is valid for customers to enjoy. For this product the various duration of offers is 5 days, 7 days & 10 days
Offer duration Vs Age of customer
It is apparent for most of the age brackets, offers with duration of 7 days (a week) resonate well with them. The only small exception is on the oldest age bracket (100–102) where the is a considerable drop in the percentage of offers with a one-week duration — 38% while the rest are between 46%–51%
Offer duration Vs Income of customer
For Income brackets vs offer duration again we observe that one-week duration is the most popular across all the income brackets taking ~50% of the offers given in each group.
Offer duration Vs Gender of customer
Again gender here doesn’t seem to influence a lot the duration of the offers chosen. Generally both male and female customers seem to prefer the same duration of offers — one-week offers.
How do demographics influence the customers in completing offers with various rewards?
Offers come with various rewards of monetary value. For this product we have 4 different reward amounts: 2, 3, 5 & 10 dollars
Offer reward Vs Age of customer
When we look at the distribution of age bracket of customers vs reward amount on an offer, we see that there is no clear preference across all the age brackets except one. Age bracket (100–102) seems to enjoy offers with a reward of 2 dollars much more compared to the other groups — about double the percentage enjoyed by each of the other groups for this reward amount. It’s also worth noting that reward of 2 dollars is only associated with offers of type discount.
Offer reward Vs Income of customer
For offer reward vs income brackets we see that the highest income earners (120,000–130,000) have a significant higher percentage (~58%) of offer rewards worth 10 dollars. This is expected because only bogos have these reward amounts and we saw earlier that the highest income earners prefer bogos a lot over discounts. Reward offers with 3 dollars seem to be least popular across all the income brackets.
Offer reward Vs Gender of customer
There seems to be no significant difference in gender preferences for the different offer rewards.
How do demographics interact with difficulty (minimum required spend to receive an offer) of offers?
In order to earn an offer you have to spend a minimum amount on the product. For this product the difficulty(min amount required spend to receive an offer) can have these values: 5, 7, 10 & 20 dollars.
Offer difficulty Vs Age of customer
Comparing offer difficulty vs age brackets we see a pattern throughout all the age brackets: Customers mostly complete offers with a minimum spent of 10 dollars. Also a minimum spent of 20 dollars is very unpopular across all the age brackets(all having below 7% of the offers). The popularity of 10 dollars min spent could also be explained from the fact that 4 offers (40% of the offer list) have this as the minimum spent amount. Hence customers had a variety of offers(both discount and bogos) to chose from.
Offer difficulty Vs Income of customer
With offer difficulty vs income, we observe almost similar results as above, with minimum spent amount of 10 dollars being the most common. We do however see a bigger share of the 10 dollars minimum spent for the highest income earners — ~64% of the offers in this income bracket; in contrast to the other income brackets where the percentage share is between (47%–55%)
Offer difficulty Vs Gender of customer
Again as seen previously in other comparisons the gender of customers does not seem to have a huge influence on the minimum amount spent that customers chose to get an offer.
Which factors influence highest the amount spent by a customer ?
We also wanted to understand how different variables influence the amount that a customer spends on this particular product. We answer this using a key driver analysis approach with the amount spent being our variable of interest.
Running a key driver analysis using variables about a customer we see how much the income of a customer has an impact on the amount spent. The income variable contributes ~82% in predicting the outcome of the amount spent. Also age of a customer has a considerable impact with 7%
Key Takeaways
From this analysis we get quite a number of business insights that are useful in in understanding how customers interact & consume this product:
- The conversion rate (rate of customers who viewed and completed an offer) was 77%
- Generally the gender of a customer doesn’t seem to influence a lot the characteristic of the offer they complete except for offer type where we saw males slightly preferring discounts over bogos while for females the distribution was uniform
- Generally discounts are more popular across all age brackets. Also highest income earners (120,000–130,000) love bogos! More than three-quarters of offers from this income bracket were bogos.
- Consequently the highest income bracket completed mainly offers with a reward of 10 dollars(attributed to bogos) — 58% of offers within this income bracket
- We also see age bracket (100–102) having a big share of offers with reward of 2 dollars within the bracket compared to other brackets
- Most of the offers completed were those that utilized all the 4 combinations of channels (web,email,mobile,social). We also found that for the channel combination of (email, mobile, social), it was a bit more popular for the highest income earners (120,000 –130,000) compared to other brackets
- One-week(7 days) duration of an offer is most popular across all age brackets and income brackets. However we also noticed for age group (100–102) the offer duration is more distributed across the 3 offer duration as compared to the other age groups
- Offers with a minimum spent of 10 dollars were most popular across all age brackets and income brackets. The highest income earners (120,000 –130,000) have a bigger share of the 10 dollars minimum spent (within the bracket) compared to the income brackets
- The income of a customer has a very huge significance(compared to the other factors) in predicting the amount that a customer spends on the product.
Conclusion
From the analysis done we can see that the age and income amounts of a customer seem to significantly influence the characteristics of an offer that they complete(e.g the type, difficulty,channel, reward levels). The gender of a customer doesn’t seem to have an influence on characteristics of an offer that is completed.
So revisiting our question of what kind of offer that a middle-aged man earning 70,000 is likely to complete — it seems he would mostly complete an offer that is a discount, has a duration of 7 days and where the minimum spent required is 10 dollars!
Link to my GitHub repo containing the analysis code used can be found here
