By analyzing a cohort, you can discern patterns across the customer journey and answer any number of questions — which marketing channels delivered the highest-value customers? At what point do users stop engaging? Where is fraud coming from and how do we shut it down?
In this webinar, Looker and Avant discuss the value of cohort analysis, tips and tricks to get the most out of it and specific examples of cohort analyses done by Avant.
Some topics covered:
-How Avant used cohort analysis to identify and optimize their key acquisition channels
-How Avant identified sophisticated fraud rings by cohorting applications by banking and network data
-Best practices for using cohort analysis to bubble-up insights
-How to identify where cohort analysis could help your business
Find the recorded webinar at:
http://www.looker.com/resources#ufh-i-180584862-cohort-analysis-get-the-most-out-of-your-data
16. LookML
Transform at Query
Looker's Architecture Enables "Near Infinite" Cohorting
FFF
OEM
Embed
App
API
Simple & Repeatable
• Intuitive & reusable modeling language
• Github for collaborative development
• Transformation at query (ELT over ETL)
100% In-database
• Leverage the power of the existing database
• Direct connection to SQL DB for row-level drilling
• Drill into and view the row level details of cohort
members
Adjust, Analyze & Visualize
• Flexible end user exploration in complex datasets
• Consistent definitions of metrics
• Ad-hoc manipulation of cohort definitions
Database
Connection
Connect Describe Explore
- name:
first_purchasers
type:
single_value
base_view: orders
measures:
[orders.customer.all]
The Looker App
Visualization & Exploration
21. What acquisition channels would it be wise for Avant to invest in?
-What acquisition channels does Avant use?
-What are the KPIs of an acquisition channel?
-What are we cohorting said KPIs by?
22. Choosing the Best Acquisition Channel
- What conclusions can we draw from this?
- What are our next steps?
24. Can we identify fraud trends using customer application data?
-What is fraud for Avant?
-What data are we working with?
-What methodology will we use?
25. Conclusions / Next Steps
-Cleanse output and identify false positives
-Automate via scheduling and filtering
-Improve general fraud identification practices
26. Conclusions
• Cohort Analysis groups subsets of users and observes their behavior over
time.
• New database technologies are emerging that allow for easier capture,
movement, and performance.
• To gain more understanding and control over your business, evaluate if you’
re using the right metrics. Test and iterate on your data model - it's a
product too!
• Create KPIs that address the goals of the organization as a whole, and of the
individuals and teams within it.