I designed it so a store only adds a script to its header to show recommended products
A SaaS that recommends products based on store visitors' page views and purchase history. It uses data gathered by VODA, the same company's web analytics service, and a store only adds a script to its header to receive and show recommended products. Eight people took part; as software architect, I designed the data flow, API, database and recommendation algorithm structure.
- 6 areasDesigned by me, from the data flow to the recommendation algorithm structure
- 3Statistics-based recommendation algorithms I developed
- 8 peopleA team of an architect, AI, back-end and front-end engineers, a PD and a PM
An area that recommends the most viewed products to all members. The store's top menu has been cropped out.
Translated reference · View Korean original- Assigned role
- As software architect, I designed data collection and analysis, how recommendations are displayed, the REST API, the relational database, customer segments and the recommendation algorithm structure, and I developed the statistics-based recommendation algorithms.
- Results
- I built a structure that turns VODA's data into per-customer recommendation lists and shows them in the store. A "most viewed products" recommendation ran on a real store's main page.
- 1Data flowDesigned analysis around purchased and frequently viewed products
- 2Showing recommendationsA header script fetches the recommendation list and shows it in a set place
- 3API and databaseDesigned the script API, the VODA integration API and the behavior pattern tables
- 4Recommendation algorithmsDesigned customer segments and the module structure; developed three statistical recommendations
Expected changes from this technology
Integration with little work on the store's side
A store gets a recommendation area just by adding a script to its header, so nothing has to be built separately for each store.
Different recommendations for each customer
By focusing on purchased and frequently viewed products and grouping similar customers, it can show different products to different customers.
A structure that makes adding and removing algorithms easy
Recommendation algorithms are managed like modules inside Django, so adding or removing one touches little else.