Skip to content

Background intelligence

This page provides the general information about the intelligence running underneath the surface. The following sections describe our AI-based models: segmentation, customer lifetime value, recommender engine and search engine.

Customer activity segmentation

Customer activity segmentation represents a convenient way to better understand the customer base in regards to their spending habits, and thus identify several groups of customers who have similar activity level. That said, customers can be segmented at a basic level by their lifecycle activity.

A lifecycle represents the period in which customer purchases are observed, and in retail it usually means a period of several months to a year.

Customer activity is described through the following dimensions:

  • when the customer was last active in a given cycle (recency)
  • how many purchases the customer had in a given cycle (frequency)
  • how much money the customer spent in a given cycle (monetary value)
  • when the customer made the first purchase in a given cycle (tenure)

Based on these features, best-fit clustering algorithm is used in order to identify segments of similar customers and thus provide a rough picture of the entire customer base and basic understanding of their activity, on the basis of which corrective and preventive actions can be defined.

Customer segments are served within Customer Studio and can be used for analysis within c360 and Customer Insights components, as well as filters for creating target audience within Audience lab component.

Customer Lifetime Value

Customer lifetime value models the future customer monetary value. The general idea is to estimate how much money can be expected from a particular customer, taking into account the previous behavior, activity level and purchasing trends.

Customer activity level is described through the following dimensions:

  • when the customer was last active in a given cycle (recency)
  • how many purchases the customer had in a given cycle (frequency)
  • how much money the customer spent in a given cycle (monetary value)
  • when the customer made the first purchase in a given cycle (tenure)

And is observed in a specific period of time - customer lifecycle.

The model tries to identify which customers are likely to stay active, and provides CLV estimation accordingly. For customers who are likely to go inactive, the model estimates low to zero value CLV, whereas for customers having only one purchase the model cannot estimate CLV, since only one purchase is not enough to estimate the CLV with high confidence.

Customer lifetime value estimations are served within Customer Studio and can be used for analysis within c360 and Customer Insights components, as well as a filter for creating target audience within Audience lab component.

Recommender engine

Recommender models are the essential part of personalized customer offer. As a tool, they can be used to present every customer with a relevant recommendation or to find the right buyers for specific products.

The next generation recommender systems are expected to include the following features:

  • more personalized recommendations – recommender systems would become more capable of digging deep into the customers’ data insights which will help them in presenting them with more-relevant, customer centric recommendations.
  • reach customers through multiple channels – the recommender systems in the future would be more capable of reaching out to the users across various mediums like emails, social media channels, on off-site shopping widgets, mobile apps, etc.
  • real-time recommendations – recommender systems are able to provide real time predictions. They aim to present the right items to a user, at the time that it is most useful to them.

Besides, not only emerging e-commerce, but other industries as well find the recommendations powered by intelligence are the crucial part in reaching the final goal of tailor-made targeting, including: retail, banking, telecommunications, real estate, etc.

AI starter-pack recommender

  • Best-seller recommender - recommending top selling products based on specific criteria (most frequently bought in a specific category and given time period), which makes it convenient for homepage carousel, leaflets, search banner,...

  • Trending-product recommender - recommending products having increasing sales trend based on specific criteria (in a specific category and given time period), which makes it convenient for homepage carousel, leaflets, search banner,...

  • Newest-product recommender - recommending products most recently included in the assortment based on specific criteria (in a specific category and given time period), which makes it convenient for homepage carousel, leaflets, search banner,...

  • Most recommended products - recommending products that are most frequently found as recommendations extracted based on specific recommender algorithm which is convenient when communication towards big audiences is needed or cold-start problem exists, i.e. homepage/home dashboards shown to new visitors and/or new customers on websites, mobile applications, POS, where no prior information about them has been collected...

AI personalized recommender

  • Community-based recommender - recommending products based on similarities between customers and products they have interacted with, which makes it convenient for cross-sell marketing strategies.

  • Next-gen ecommender - recommending product based on long-term patterns and preferences combined given the real-time occasion/intention the user is having at the touchpoint, which makes it convenient for real-time recommendations and e-commerce platforms.

  • Description-based recommender - recommending products based on similarities extracted from their features (description, category, name,...), which makes it convenient for add-to-cart” banner, search banner, POS recommendations,...

This engine is efficiently incorporated and served through:

  • Touchpoint studio - where recommendations are served in order to ensure that every customer is equipped with relevant recommendations he comes across every digital platform
  • Customer Studio - where recommendations can be used for analysis and filtering within the c360 and Audience lab components
  • Campaign Studio - where recommendations are used for creating dynamic and personalized offers on a customer level

Smart search engine

Smart search represents a bundle of AI modules used for searching various data types and sources. Currently, there are 3 main modules:

  • Smart search - used for searching the textual information stored in formats like csv, excel, json, xml,…
  • Document search - used for searching the textual information stored within unstructured documents stored in pdf format
  • Image search - used for searching a database of images based on provided image

Smart Search is the search engine that relies on intelligence and has the following characteristics:

  • multilingual search - understands any language you speak
  • letter agnostic search - supporting alphabet and cyrillic
  • typos tolerant search - tolerates typing and spelling mistakes and shows relevant results
  • recognizing synonyms - understands if use synonyms, and gives options to define more synonyms specific for the banking industry
  • recognizing context - understands what you mean even if you use verbs and adjectives in different grammatical forms
  • configurable training options - you can be involved in the model fine tuning, decide to boost some features or define stop words

The “smart” thing about the smart search engine is that it works perfectly even though no perfect search terms are provided (or incomplete information in the query is provided) - it is able to catch synonyms, to observe specific attributes are more important than others, and to propose most adequate alternatives when no match is found.

This engine is efficiently incorporated and served through Touchpoint studio in order to ensure intelligent search on digital platforms.

Back to top