Define the problem
Customer churn occurs when customers or subscribers stop doing business with a company or service.
All the telecom/cable/tv services need to know what happen. What kind of people will leave? Customer segmentation Why the customer leave? Identify churn triggers. When they gonna leave? How they change their mind? Treat cost/effect estimation Explicitly “retention-first” care strategy or marketing strategy to hedge against their risk of leaving.
Everyone new to this problem should take look on Kaggle data first. There are already good solution on predict. This post is more focus on how to utilize NLP data on churn problem.
NLP source
Call log/Email/Online assistance (CRM)history are the three major resource the company can utilize. Do you remember every time you call for customer service you’ll heard For quality purpose, this call might be recorded.
Except that, When customer don’t receive exceptional customer service and experiences, they flock to competitors and share their negative experiences on social media: 59% of 25-34-year-olds share poor customer experiences online.
Ideas of solving problem
- Social media monitoring
- Voice of employee/turnover prevention
- Voice of Customer (VoC) & Customer Experience Management
- Text summery
Pain Point
- Unbalanced data set In big company. There’s enough data set show the customer has churn issue. However it doesn’t mean they’ll all complain on customer service.
- Interpretable model NLP problem and deep learning,
- Integration with other data
-
Long review This problem involve topic modeling/summarization issue.
- Zero-shot discovery of typical user types
-
Transcript doesn’t work well
-
Handling correlated multi-dimensional behavior data
- Model explainability
- Evaluate the solution in production
Solution/Paper review
I Know You’ll Be Back: Interpretable New User Clustering and Churn Prediction on a Mobile Social Application
- Track 1 month snapchat new register account information to predict user behavior
| ID | Name Feat. | Description |
|---|---|---|
| 0 | chat_received | textual messages received by the user |
| 1 | chat_sent | textual messages sent by the user |
| 2 | snap_received | snap messages received by the user |
| 3 | snap_sent | snap messages sent by the user |
| 4 | story_viewed | stories viewed by the user |
| 5 | discover_viewed | discovers viewed by the user |
- Doing cluster on user(cluschurn).
- Build LSTM model predict sequence behavior
Text Classification by Bootstrapping with Keywords, EM and Shrinkage
This is 1999 paper. Pretty old. But it do provide a solution to generate sample set. Dealing with unbalanced issue.
Implement text summarization. Pick the important sentences.
Sample code
Terminology
-
LTV life time value
