The right product, at the right place, among the right accomplices, for the right buyer, at the right time
consumers likely to be a repeat buyer after a personalised shopping experience
higher conversion rates when consumers engage with recommended products
more likely to complete their purchase, for shoppers clicking on recommendations
more unique products viewed per visit by shoppers who clicked on a recommendation
With an objective to streamline sales and target easy-to-convert customers by recommending optimized product bundles, this recommendation engine use case was developed for a global premium auto major. The sales process was being managed by their dealers in exchange for a hefty sum which resulted in the client incurring 10X additional cost. Therefore, they wanted a better cost to value ratio. The aim was to suggest the most appropriate product bundle / product combination that the customer is likely to buy.
28% increase in revenue achieved by recommendation engines for some dealers
Offer similar products that the customer is likely to look for based on customer's purchase history and similar users to the customer
Additionally discounts given for recommended similar products as in pure upsell
Suggest the customer a more appropriate product than the one searched for using relevant factors like history, currently high-selling products etc.
Reminders suggesting relevant items with and without promos in times of reduced usage or ineffective browsing
Offer discounts while giving product reminders
The Finarb Formula
Content and context based recommender systems using Collaborative filtering and Deep Learning approaches provide high boost for sequence based recommendation. The goal of our recommender system is not only to provide the content on corresponding topic but also to rank content elements by quality and expertise level to fit the resulting feed to each user’s level of expertise
We used almost all data available to various degrees to extract information e.g. User interaction data(complete 360), Web analytics data, Reviews data, Competition data, Vendor information, Social media data and Inventory data
The recommender system is deployed using apache beam pipelines using the apache stack elements like griffin etc and controlled by apache airflow to send the data into elasticsearch engine. The data is analysed by python/scala environment. The entire stack is compatible with and has connectivity to most front end and other backend systems
The deployment was done on a pay per call basis in some verticals. In the ones where there was higher use, it was deployed with a periodical licensing fee. The entire stack was on cloud infrastructure in serverless scalable deployment.
The APIs were integrated to web analytics systems, POS systems, inventory management systems whole was also compatible with the RDBMS/NOSQL data including graph DBs. The api s were integrated to web apps and mobile apps.