The increased emphasis on customer experience and the rise of digitalisation has drove the need for many sellers to adopt the use of Artificial Intelligence (AI) models and Machine Learning to enhance their services. One such use of AI models is recommendation engine, a tool which suggests products, services or information to users based on analysis of data collected. It is the technology behind customised recommendations of dramas one may like on Netflix, product recommendations from similar buyers on Amazon and many more.

Despite the availability of established recommendation engines, the uptake is still relatively low due to difficult implementation from open source engines as well as high cost from commercial ones. As such, IMDA Digital Services Lab introduced the X-Selling Engine, as containerised packages and royalty-free, to help more SMEs ease into this technology and remain competitive in the digital era.

X-Selling Engine is a machine learning engine that uncovers latent interests and knowledge from different consumer-related datasets, to make timely and customised recommendations for consumers. The various AI models use transactional (from the Customer Relationship Management (CRM) database) and real-time user interaction data as training datasets to generate recommendations of products or services for the customers.

X-Selling Components

The technologies developed can be used to identify two types of recommendations for buyers. Each type of recommendations is generated by aggregating the product rankings from its respective components.

The first type of recommendations suggests one’s favourite products or services (“exploitation”) via the following components:

  • Siloed-Demographics

    This technology segments buyers’ interests (products or services) based on age groups and gender using the transactional data collected. The products or services are then ranked according to the proportion of buyers.

  • Collaborative Filtering

    This technology identifies similar buyers of a target buyer based on their demographics and ranks a list of products or services that similar buyers have bought which might interest the target buyer. It is also based on the assumption that people like items similar to other items they like, and they also like items that are liked by other people with similar tastes.
  • Graphed Algorithm

    This technology is adapted from Hyperlink-Induced Topic Search (HITS), a Link Analysis Algorithm which rates webpages. The graphed algorithm helps identify Authority nodes (popular merchants) and Hub nodes (buyers) in a visual representation. Hub nodes attain a higher score when it has more connections with Authority nodes, and Authority nodes attain a higher score when it is linked to more Hub nodes (iterative process). The Authority nodes are then ranked according its score which shows each store’s popularity.

  • Matrix Factorisation (MF)

    This technology decomposes user-item matrix into the product of user latent factors and item latent factors to capture the latent features which predict the rating of the item that users might give. For example, it decomposes a customer-store matrix to capture the latent features which predict the ratings of stores that customers might give. Thereafter, the technology ranks the items according to its predicted ratings.

    The technology is also very efficient for dataset with high sparsity such as click history and is used by many large tech companies like Netflix and Amazon with great success.

The second type of recommendations suggests similar products or services (“exploration”) via the following components:

  • Merchants Collaborative Filtering

    This technology averages all shopping patterns from its customer population for each merchant. The averaged pattern indicates the correlation between merchants and a list of recommendations is generated consisting of merchants who have the strongest correlations.
  • Custom-Made Merchants Similarity

    Each merchant has its own unique customer population and between any two merchants, there exists a common pool of customers who visited both merchants. The larger this common pool is, the stronger the two merchants are correlated.

    This technology uses the above idea to design a formula which measures merchant similarity and generates list of recommendations from the most similar merchants.

  • Co-Purchasing Probability

    This is a customized graph technology which estimates the likelihood that a customer would purchase another product or service based on previous purchasing history. A list of recommendations is generated based on the probability of each product or service.


X-Selling Engine augments and enhances current technological offerings to improve customer engagement, which can potentially boost sales for sellers. It is also an affordable system which can be adopted by various industries, such as:

  • Retail (e-commerce)
    • Recommend products or services that the customer might be interested in based on their past purchases or similarity with other customers.

  • Media (advertising)
    • Display advertisements that users might be attracted to view or click into based on their past click or search history.

  • Education (e-learning)
    • Recommend courses that learners might be interested in based on previously attended courses and their demographic.


  1. What are the terms of use for X-Selling Engine?

    If you are keen to use X-Selling Engine, please reach us at the email address provided below. The deployment of the technology will require a Technology Licensing Agreement with IMDA. For more information on terms and conditions, kindly refer to the form here (199.69KB).

  2. Is there a fee to pay in order to use X-Selling?

    No. The engine will be provided for free with approved TLA.

  3. What kind of technology licence is offered by Digital Services Lab?

    The technology licence is perpetual, world-wide, non-exclusive and sub-licensable.

  4. What are the obligations of the licensee?

    By signing the technology licensing agreement, the licensee commits to the use, the commercialisation, and the propagation of the technologies.


For further enquiries on X-Selling Engine, please contact DSL_Tech@imda.gov.sg.

Last updated on: 15 Apr 2021