MindWave AI for Consumer Profiling
With a limited amount of data, MindWave AI managed to profile customer preferences for the bank to target their customers.
By collecting and analyzing a vast variety of unstructured public data we identified core characteristics of each customer and profiled them; and made sure that all the data solely resided on the bank’s server.
How we did it
To forecast customer preference through our advanced statistical techniques
Mining relevant multi-lingual unstructured data and translating for consumption
Identifying and segregating customers with similar traits and assign them into clusters for value-added targeting
“We were amazed at how the AI managed to capture relevant data and profile our customers based on just their names, emails and locations. It enabled us to push more personalised offers to our customers and generated greater interest from them on our offerings.”
The main challenge was the lack of information from the bank’s existing customer database. We were provided limited information like name, e-mail address and location of each customer and from this, we had to produce a variety of insights about customer. Also, we had to recognize and translate the data from European languages other than English.
- Identify ideal customers to be targeted for specific offers
- Identify distinguishing characteristics of each customer such as their interests and preferences
- Segment, cluster and classify customers based on multiple criteria to help grow their business
- Make sure that our entire setup scales well to handle a large customer database efficiently
- Create reasonable predictive models with no data to start with
We developed a wide range of innovative ways to construct reasonable predictive models including some that were based on leveraging external data such as economic and social profiles of different regions/neighborhoods, and demographic information among others.
- Core NLP to recognize, translate and use data in non-English European languages
- AI algorithms to search and capture relevant data given only the rudimentary information from the bank
- Predictive modeling to identify customer preferences
- Clustering of search results filter irrelevant data
- Filtering out irrelevant results so that the filtered set of results matches a high likelihood of the target customer
- Created appropriate set of queries to search the suitable external services from the rudimentary information provided by the bank
The result was the design and implementation of new algorithms to produce the best outcome. The solution we built leveraged several powerful and unique features of our platform and was also improved over time based on the feedback and adjustments received from the bank.
- Collect and analyze the vast variety of data from the public web
- Identify core characteristics of each customer by extracting relevant concepts and respective categories
- Filter out irrelevant results
- Core natural language understanding engine
- Ensure all data solely resided on the bank’s server
- Predictive model refinement incorporating customer feedback