Question Type: Product

Duration: 40 Minutes

Difficulty: Medium

Domains: Product

Problem

Suppose you have 10 million transactions of Apple customers. How would you find meaningful segmentation of customers?

Solution

Suppose you have 10 million transactions of Apple customers. How would you find meaningful segmentation of customers?

[Candidate] Thank you for the question. I want to first discuss the business problem with you. You mentioned “transactions” and “customers.” I understand that Apple isn’t just a company that sells iPhones but, various other services and products such as subscriptions (i.e. Apple Music, Books, News), payments (i.e. Apple Pay) and app stores. Which slice of transactions would you want me to take a look at?

[Interviewer] That’s a great question. How about just focusing on devices for now?

[Candidate] Okay understood, so devices could be iPhones, MacBooks, iPads and such. Is this correct?

[Interviewer] Yes.

[Candidate] So, I’m going to assume that if these are the transactions people focus on then, the type of customers we care about are store visitors whether brick & mortar or online.

[Interviewer] Sounds good. Anything else?

[Candidate] Just another business-related question. As you said, segmentations should have meanings but it kind of depends on the type of problem. Could I ask a bit more context on where this is plugged-into?

[Interviewer] Let’s suppose that the segmentation is used for a marketing purpose. How would you approach this problem then?

[Candidate] Understood. So, perhaps we are investigating trends and demography of users who would purchase certain device types.

[Interviewer] Sure that could work.

[Candidate] Okay great. I think I have enough information to start formulating an analysis approach.

[Interviewer] Great. Let’s hear it.

[Candidate] There’s a framework that can be readily applied when it comes to segmentation analysis, something called RFM, recency-frequency-monetary value. I know that it’s usually employed in customer segmentation to understand who are top-tiered customers and who aren’t. This framework can help categorize users.

[Interviewer] Interesting approach. What kind of data would you need for such analysis?

[Candidate] I will need a user’s purchase history. Can I assume that Apple can track user’s device and accessory purchases based on user email address?

[Interviewer] Yes.

[Candidate] Great. Then, here’s the data source I’m thinking. The data could consists of the following variables – email_address, transaction_date and sales. For instance, I recently bought the MacBook Air 2020, so my email address XYZ@gmail.com, 2020-02-14 and $1,100, will appear in the transaction data.

[Interviewer] So, how would you use this data to create a segmentation?

[Candidate] I’d roll-up the data at the user level to generate the following – (1) Recency – days since the last transaction, (2) Frequency – count of transactions, (3) Monetary – total sales. I can use this information to apply segmentation on users.

[Interviewer] Okay, how would you segment the user on the variables?

[Candidate] I’m thinking I could apply quantiles. So, the top 33th percentiles on recency, frequency and monetary could represent a bucket of customers who are the most loyal customers of Apple. When a new product launches, they are the first ones to purchase them regardless of the price.

[Interviewer] Interesting approach. Can you think of another customer group?

[Candidate] Each dimension could be binned into three intervals based high (x <= 33th percentile), mid (33th percentile < x <= 66th percentile) and low (< 66th percentile), such that a total of 27 segments are generated. The low levels across all three buckets are users who are the opposite of loyal Apple customers. They are difficult to funnel into sales.

Interviewer Assessment

In the statistics section, a candidate is assessed based on correctness and soundness of statistical methodology, product sense and communication. For each dimension the candidate is rated in the following scale: (5) superior, (4) good, (3) adequate, (2) marginal, (1) not competent.

Assessments	Rating	Comments
Statistical Methodology	5	The methodology from the candidate sounds great. He understood how to apply the RFM framework. When asked about what kind of data source he might use, he didn’t shy away from being able to respond clearly and specifically. He also understood how to find meanings in the segmentation applied which is a huge plus.
Product Sense	5	The candidate understands the business model of Apple. He knew how to turn a generic question on “transactions” and “customers” into specifics (i.e. device users, subscribers). This suggests that he researched things about Apple before the interview which is one of the qualities our team looks for as stakeholder engagements often start with a vague, open-ended business problem.
Communication	5	The candidate structured his response very well. He began with a business discussion, asking questions about the problem and clarifying assumptions about Apple. This shows that the candidate knows how to turn an open-ended question into a specific problem. Most other candidates will often jump into the solution first. But, this candidate stood out given that he made sure there’s no room for vague interpretation. Then, he followed up with a solution laying things out in a step by step manner. He was coherent, clear and confident with his delivery. A+ for his communication performance.

[Apple] Meaningful Customer Segmentation

Question Type: Product

Problem

Solution

Unlock now by joining.