Effective Recommendation System With Machine Learning for Ecommerce
It gives users access to trusted recommendations, a platform to showcase their style, and the opportunity to earn money.
Their founder, Manon Roux, makes shoppers’ lives easier by building style tribes around their interests and preferences.
The benefits for users include:
- Shopping together with friends,
- Finding style inspiration,
- Showcasing their unique taste,
- Giving and receiving recommendations,
- Earning money through creating content.
With more than 160 retailers in its network, Countr can truly have an impact on the way users enjoy online shopping by giving them more ways to interact with friends and make recommendations.
Developing an effective recommendation system
Netguru joined Countr’s team in January 2018, when the app’s alpha version required a general reorganisation. Our process involved several steps:
- We performed an initial code review and a scoping session to gather all necessary data, then began designing a product that would fit the client’s vision.
- Using what we learned, we were able to recommend solutions for tracking relevant metrics, building a better app architecture and using effective designs.
- Restructuring the app improved its scalability, while rewriting legacy code according to best practices allowed us to optimise the platform and ensure that it wouldn’t be difficult to maintain.
- We also worked on the machine learning solution for product discovery.
After extensive testing and making sure users would be able to take full advantage of the value it introduces, The new version was ready. It took months and a lot of user testing to polish the personalised feed.
Users who took part in the process left incredibly positive feedback about the app’s look and usability. Soon enough, several celebrities entered talks about extended collaboration with Countr’s founders.
Using technology to innovate shopping
The app isn’t a simple social platform with a beautiful feed. Netguru’s development team took on the challenge of developing complex features and systems, including:
- A machine learning discovery algorithm. Upon sign-in, users select stores and categories they are interested in. This information is used for providing tailored recommendations. The available categories include fashion, beauty and lifestyle, which then divide into narrower areas of interest, such as skincare or workwear.
- The algorithm’s ability to determine the optimal ratio of adding preferred retailers to other retailers to each user’s feed, aiming at improving user satisfaction. Users can easily keep track of brands they trust while also discovering new favourites, without sifting through content that is useless to them.
- Single cart checkout, which makes the process of making multiple purchases at once incredibly smooth and user-friendly.
- Cheer and tipping between users as a response to shared content, which builds on the community and social aspects of the app.
"Netguru took on some legacy code and helped us – they’ve been fantastic on that front. We’ve received a lot of support in terms of thinking about how do we track metrics, how do we design this properly, and how do we build the architecture. We are extremely grateful for making our platform what it is today."
Manon Roux
Founder of Countr
The challenges of bringing an innovative shopping app to the market
Countr’s user journey needed enough features to gain a competitive advantage, yet it needed to remain intuitive and easy to use.
Product discovery had to be advanced enough to truly reflect users’ preferences and tailor their feeds. The platform also needed influencers to spread the world and build traction. The client already had an extensive network of contacts, allowing for larger impact upon release while turning Countr’s launch into a high-stakes event.
Countr also needed to implement a lot of functionalities to quite a demanding app in a short time – our Ruby on Rails team helped the client take advantage of pre-built "Gems" to do this effectively and in the shortest time.
Ruby "Gems" helped us enable third-party integrations with services such as ‘Elasticsearch’, an open source search and analytic system, and ‘searchkick’, to help enhance the client search process and give better results – our team also implemented an integration with Google Vision.
The next step: learning from user behaviour
A future goal for Countr is implementing learning from users’ actions, such as liking or disliking products, and making purchases.
This information could be used to better understand each user’s preferences. It’s not easy for an algorithm to do this, as people like various things for different reasons. They could be attracted by the price, colour, brand or style. A lot of data will be needed to refine the recommendation system to this level.
But it will be worth it. For an online ecommerce platform, keeping up with online shopping trends is mandatory if one does not want to fall behind their competition.
It’s a good time to develop and improve an ecommerce platform.
More and more people shop online, especially on mobile devices as 73% of all online shopping will happen via a mobile device by 2021, according to predictions. Social media is playing a bigger role in the decision-making process, and AI automation is changing the way people interact with retailers.
Personalisation – tailoring the shopping experience to each user – is one of the main trends determining the future of ecommerce, an industry with an exciting future ahead.
