GO-JEK Customer
Intelligence
How to leverage data to increase customer’s satisfaction
Today’s discussion
GO-JEK Customers
Customer Intelligence in GO-JEK
Analysis framework
Use case example
Customer?
Products
Consumer App
Products
Driver App
Products
Partners App
Customer Intelligence
in GO-JEK
How customers behave
Determining key metrics Determining ways to measure it
Consumers
lifetime value, GO-PAY usage behaviors,
most frequent service type used
Drivers
productivity hours, bid acceptance rate,
driver ratings
Partners
best-selling dishes, monthly merchant
growth, stores with highest
cancellations
Customer Intelligence in GO-JEK
Platform level
Weekly Cohort Funnel Analysis
Customer Intelligence in GO-JEK
Service Type level
GO-FOOD Live Content Tracker GO-KILAT Online Store Tracker
Analysis
Framework
Analysis Framework
1. Problem 2. Hypothesis 3. Experiment 4. Result &
Statement Evaluation
What problem are you What statement you What approach do you How does the result
trying to solve? want to validate use to prove the satisfy your
through this analysis? hypothesis? hypothesis?
How the analysis
result will impact to What are assumptions What works and what
business decisions? do you use in doing doesn’t
the experiment?
Plans on what to do
next
Example
Order Process
Is there Do the
Yes Yes
Customer place any driver driver
Order completed
an order accepting completed
the bid? the order?
No
No
DriverNotFound Order not completed
Bid Booking
Example : GO-CAR Acceptance Rate Experiment
1. Problem Statement
We need to get better conversion rate in order to achieve better customer experience (faster time to
get a driver after they place an order). Many ways to make it happen, and one of the ways is
acceptance rate.
~20%
Period : April 1 - 15, 2017
Example : GO-CAR Acceptance Rate Experiment
2. Hypothesis
Social pressure will increase acceptance rate.
Example : GO-CAR Acceptance Rate Experiment
3. Experiment
"(Info) Anda termasuk 20% mitra dengan performa paling rendah di GO-CAR.
Ayo tingkatkan performa dan raih bonusnya!"
Control
Experiment
Example : GO-CAR Acceptance Rate Experiment
4. Result & Evaluation
We got an average increase of 2.5% from the action!
Next step : set up an sms blast to be sent regularly to bottom 20% of drivers with few different versions
but same purpose.
Key Takeaways
1. Every analysis should be actionable
We are doing the analysis to give recommendations to a problem(s)/make a business
decision.
2. Be hypothesis-driven
- Always start with a hypothesis, then do an experiment to prove/disprove it
- Focus on 20% possible analyses that will likely generate 80% of the actionable outcomes.
3. Embrace failure
We do the experiment to prove whether our initial hypothesis is true or not. The goal is to
achieve a better understanding on what’s actually going on and how we can make it better.