Bio Clustering led Social E-commerce App to get 23% better acquisition ROI

02/12/2017

Tuple's Bio Clustering model uses advanced machine learning algorithm to create a persona of incoming customers by using all the information attached to them. By finding patterns from the persona, our system can accurately predict which customer is going to convert as soon as the customer registers. Also, our platform then creates bio-clusters based on the persona by understanding which factors lead to more profitability. Companies can use the most profitable set of customers to create a lookalike audience on platforms like Facebook and acquire more customers just like them.


The Client: Our client is prestigious social e-commerce application serving millions of users across SEA. They have a monthly subscription based business model.



The Problem: Our client was making gut based decisions about which customers should be acquired and what budgets should be allotted to each campaign. The team was not sure what and how much data should be considered for creating an effective segmentation.

This was leading to a lot of wastage in acquisition spend because client business relied heavily on the new acquisition. Team was getting no visibility into which clicks & views on the ads are actually generating profits for the company.


The Solution: Using 'Marketing Intelligence Assistant', our client was able to segment users basis on the information they shared at the time of registration.

This helped the team to simply take the most profitable clusters and create a lookalike audience on Facebook to acquire more customers like that.

They also used the insights from segmentation to create targeted acquisition campaigns for their most profitable users. They were also able to track the shifts in acquisition patterns over time and were able to change their strategy accordingly.


Highlights

  • 23% increase in ROI from the acquisition campaigns
  • Data  from ~ 150K users was integrated, cleaned, segmented and monitored over time
  • 89.42% accuracy in identifying potential high converts by using just the profile information