Strategies refer to the content that we personalize & provide for your customers to experience enriching Journeys. Strategies are built on top of our underlying Machine Learning Models that dictate the logic behind various approaches to product recommendations. Powered by several algorithmic decisions, these models analyze, learn, and discover patterns in customer, product, and contextual data to present every customer with a personalized experience.
We provide several out-of-the box strategies, which you can choose from and customize and tweak parameters to suit your needs.
For each model chosen, there are specific levers or parameters exposed, that you can tune in order to suit your use case.
Eg: For Grocery segment, a similarity model will have the following levers. If you want to provide more weightage to any particular attribute, the model will take that as input and start returning results accordingly.
You can also configure to add more fields available in your catalog to this attributes list.
Business Rules are manually configured filtering conditions applied on top of the strategy to target a specific subset of products. Only those products that match the filtering conditions are returned in recommendations. In case of Contextual models, business rules can also be configured to be applied only when the product in context meets specific criteria.
Eg: Business rules can be set to return only products with price less than 100$, when contextual product belongs to category Dresses.
This turns on dynamic personalization for the strategy. To begin with, the configured model parameters will be used as a starting point to deliver product recommendations. As we know more about customer's affinities and preferences, our recommendation engines return more personalized results for the strategy for that customer. This leads to better shopping experience for your customers, thereby increasing engagement and conversion rate.
Broadly our models fall under one of these categories:
|Contextual||Any model that returns recommendations based on a contextual input falls under this category.||Similarity, Collaborative Filtering|
|Popularity||Models that work on top of population behavioral data.||Trending in a category|
|Journey-aware||Models that work on top of individual user's history and past journeys. Models based on user's visual and behavioral preferences fall under this category.||Recommended for you, Recently Engaged, Inspired By customer's Browsing History|
|Bundling||Bundling models can be Cross Product Bundling, that displays cross-product recommendations, based on the product in view and Product Curations Bundling, where collection of products or product groups related to a particular theme are returned.||Cross Product, Product Curations.|
|Search & Listing||Models that work on top of our search engines.||Search|
Recommends products with similar attributes. Matches similarity vectors and picks products from the catalog with attributes similar to the product currently in view. It enables your customers to discover other similar products from the catalog and quickens the purchasing decision.
You can configure this model to create various types of recommendation strategies by assigning varied importance to product attributes such as brand, color, and category. There is also an option to apply 1:1 affinity based personalization on the model.
Establishes patterns between customers (creating a group), and recommends products within the group. Collaborative Filtering is a technique used for recommending products that a user might like based on interactions, data, and preferences collected by the system from other similar users. These algorithms enable you to harness the crowd wisdom to drastically increase cross-sales and overall revenue.
Collaborative filtering can be:
User based: Measures similarity between users and other users.
Item based: Measures similarity between the items that users interact with and other items.
This model recommends products that are most popular on the website based on the traffic data. This takes into account varied customer events such as Add to wishlist, Pageviews, Buy and Add To Cart over a period of time along with attributes to identify popular products.
This model recommends products considering the customer's behavioral preferences as well as attributes similar to the product currently in view.
You can configure this model to create various types of recommendation strategies by assigning varied importance to product attributes such as brand, color, and category, customer events such as add-to-cart, buys, page views, etc.
This model displays products that were last engaged by the customer, with the most recently engaged appearing first. The recommendations are typically based on data from the last X days.
You can configure this model to create various recommendation strategies by adding a single event or combining events.
This model looks at products that your customers have been browsing and recommends very similar products of different shapes, sizes, and brands to help them find a product very similar to the one they have browsed earlier.
This model can be configured to create different strategies based on:
- Customer events such as page view and add to carts
- Assigning varied importance to product attributes such as brand, color, and category.
Bundling models return products bundled together in relation to a product in view or curated for a theme. In Fashion segment, this can be a Style it with model suggesting outfits to wear for different occasions.
Cross Products displays cross product recommendations, for the respective theme, based on the product in view. This model is based on a source product.
This displays a collection of product outfits for a particular theme.
You can configure both of the above models for a variety of themes available for example casual, party, etc.
VueX comes with preset strategies based on your business. We can also build custom strategies based on your requirements. Get in touch with us to know more.