Frequently Asked Questions (FAQs) about maya.ai

maya.ai is a personalization platform designed to help Credit Card, Debit Card and Wallet businesses engage their customers more effectively with personalized consumer choices. Our platform currently offers personalized engagement capabilities across multiple consumer categories, such as shopping, retail, dining, travel, etc.

maya.ai enables card and wallet businesses to become their customers’ top of mind and eventually their top of wallet choice of payment option with ease and confidence.

maya.ai can operate across all digital engagement channels from offer pages on websites, mobile sites, mobile apps to more traditional channels like email and SMS

We have a white-labelled app (Engage App) on which personalized offers can be shown.

Maya.ai also offers Engage App, a ready-to-deploy white-labelled app for clients without an existing digital engagement asset, to display personalized recommendations and engage their customers with.

To an extent, it is. For clients with existing offers, maya.ai can personalize them for their customers. If maya.ai detects that the existing offers are not effective in engaging the customers, it can offer suggestions on what offer merchants to acquire and onboard to improve the engagement level. Additionally, maya.ai Bazaar is our marketplace offering that connects your business with a supply of merchant offers from around the world and across consumer categories.

While maya.ai, personalization can be used for digital acquisition, the platform has been used primarily for driving engagement, transactions and reactivation. In the past we have seen 1.3x improvement over control on reactivation, 3x-10x improvement over control on response rate, 20%-50% increase in click rates, 3%-7% increase in spends

maya.ai adopts a unique approach to generating recommendations leveraging structured, unstructured, historical as well as real-time data. maya.ai deduces a customer’s taste by looking at her consumption patterns aided by our deep understanding of each merchant the customer transacts with, as well as signals we receive through real time engagement.

To arrive at a detailed customer profile, maya.ai adopts a three-step approach when analysing transaction data:

1. Deduplication of merchant names in transaction data
The same merchant in transaction data may appear under different names or even different entities, resulting in difficulties in understanding the customer transactions. With a number of propriety algorithms, maya.ai can clean > 1M transaction in under 4 hours, with 80%+ accuracy in attributing the correct merchants to the transactions

2. Merchant data enrichment at scale
maya.ai leverages our patented TasteGraph, which contains in-depth knowledge (+300 attributes) for over 6 million merchants around the globe, as well as other propriety algorithms to analyse unstructured and structured data from first- and third-party sources.

maya.ai uses the TasteGraph to enrich each merchant entry with hundreds of attributes. For example, type of product carried, price point, rating of the merchant where the transaction is made. This enables maya.ai to understand the meaning behind each customer transaction more accurately.

3. TasteMatch calculation to identify the best recommendation for a customer
Using proprietary & patented methods based on the graph theory (TasteGraph), maya.ai calculates the affinity between the products or merchants in the enormous transaction data set. Combine this affinity with the outputs of a few other algorithms, mixed with real-time contextual signals, such as location, device, intent, weather, occasion, events, and influence signals, such as trends, trusted data sources, a TasteMatch score is used to identify the most relevant choices for a customer.

maya.ai profiles the customer – understanding the customer’s “tastes” — through deduplication of the transaction data, enriching transaction data with attributes, and calculating the taste match score between every customer and merchant/ product

• Deduplication of transaction data: The same merchant in transaction data may appear under different names or even different entities, resulting in difficulties in understanding the customer transactions. With a number of propriety algorithms, maya.ai can clean > 1M transaction in under 4 hours, with 80%+ accuracy in attributing the correct merchants to the transactions
• Enriching transaction data with attributes: maya.ai enriches transactions with hundreds of attributes, for example, type of product carried, price point, rating of the merchant where the transaction is made To do this, maya.ai leverages our patented TasteGraph, which contains in-depth knowledge (+300 attributes) for over 6 million merchants around the globe, as well as a number of propriety algorithms to analyse unstructured and structured data from first and third party sources.
• Calculating Taste Match across the products and offers in the portfolio: through patented methods based on graph theory, maya.ai is able to calculate the affinity between the product/merchant in the customer’s transactions, and other product/merchant. This affinity, combined with context, such as location, device, intent, weather, occasion, events, and influences, such as trends, trusted data sources, is used to refine the most relevant choices for a customer

• Marrying internal and external data, structured and unstructured data, to have a comprehensive understanding of the customer: less than 20% of the vendors in personalization space utilize external data to enhance the profiling and recommendation, and even when they do, they typically bring customer tracking data – which is usually in violation of customer privacy laws
  • maya.ai brings the vast understanding of products and merchants from outside world. These insights are not limited to structured data. The NLP algorithms allow maya.ai to extract insights from unstructured data, such as text reviews. With internal, external, structured, and unstructured data combined, maya.ai creates a much deeper customer taste profiles than do competitors
  • Almost every other vendor relies on segmentation-led algorithms to profile customers, which by definition treats clusters of individuals alike. This defeats the purpose and premise of personalization Marrying internal and external data, structured and unstructured data, to have a comprehensive understanding of the customer: less than 20% of the vendors in personalization space utilize external data to enhance the profiling and recommendation, and even when they do, they typically bring customer tracking data – which is usually in violation of customer privacy laws

• Privacy-sensitive: by looking at customer’s tastes instead of tracking customer’s identity, maya.ai is able to offer personalization without any PII, offering end customers complete privacy
• Work with sparse data set: maya.ai requires only a few transactions per customer to start recommending choices. Even when there’s no transaction, maya.ai can still build a customer taste profile via a few simple questions/test choices. With machine learning, maya.ai can overtime enhance the customer profiles by learning users’ interactions with the recommended choices.

• The interactions made by the customer on the recommendations shown on a digital asset, inform maya.ai of the following:
  • Current lifestyles need of the customer – what category is the customer looking to spend in or is he/she just browsing at random?
  • Profile of merchant which appeals to the customer – what is the avg ticket size? How popular is the merchant? Is the merchant new to the customer or has the customer spent money at this merchant before?
  • What attributes are driving their current needs – what are the common attributes underlying the merchants a customer is interacting with?
• All this information enhances the understanding of customer context at the given moment
• From here a few things happen:
  • In-session the recommendation lists get re-prioritised to show customers more relevant recommendations in a bid to drive a purchase
  • In-session anticipate merchant combinations that go well together and attempt to prioritize recommendations in a bid to increase “cart value”
  • Across sessions, identify next best recommendations based on predicted customer behaviour
  • Store interaction data to send notifications to customer

• There are over 4.5Mn merchants globally that have been tagged and enriched within maya.ai’s repository
• A merchant, according to a bank is:
  • An MCC code
  • A ticket-size
  • A city / country where it is available
• The above is limited to the most-popular merchants, mid & long tail merchants are unrecognizable in the bank’s datasets owing to poor quality of data stored in the transaction description
• To maya.ai, a merchant is:
  • The category it belongs to
  • The attributes that define it
    • A category can have up to 10 attributes
      • Each attribute can take up to 10-15 unique values
        • A merchant contains at least 2-3 values per attribute
  • The ratings given to it by customers on public platforms
    • Ratings are typically given across 4-5 parameters such as budget, service levels, etc.
  • The reviews given to it by customers
    • Free-text reviews from which key discussion points are extracted using NLP algorithms
      • What is special / unique about the merchant
      • What is it that people love about the merchant
      • What is it that people dislike about the merchant
  • In an essence, a merchant is close to 75-100 attributes in the maya.ai repository

We have global partners and can augment offers from them on a need basis

We follow a subscription model (monthly licensing fee). Contact us to know more.

We use the Test vs Control approach for measuring ROI. The test group will keep receiving personalized communication while the control group will get the regular BAU communication

• maya.ai engages customers across different channels, from email/sms, to mobile app and web portal, to high-touch/high-value channels, such as concierge, using one powerful Choice API.
  1. For email/sms campaigns, maya.ai allow clients to identify, explore opportunities, curate and execute campaigns based on customers tastes
  2. For mobile app/web portal: Choice API can power personalization for customers’ existing assets such as reward page and internet banking logout page
• maya.ai also come with white-label assets to accelerate client’s personalization journeys
  1. Choice API can even be used to drive personalization on high-touch/high-value channel. For example, in a global card issuer, maya.ai powers the concierge for high-spending customers. In this case, maya.ai deliver the personalized choices to concierge staffs, who then recommend the optimal choices for customers on the phone.
• With this comprehensive approach, maya.ai can reach the majority of the customers in the portfolio (75%-80%), and lift many metrics:
  • Reduce bounce rates from industry-standard of 50% to 15%
  • Increase open rates by 50%
  • Increase CTRs by up to 70%
  • Increase response rates by 1.6x-2.4x for campaigns (Test vs Control)

• Typically, maya.ai penetrates 85-90% of the eligible customer base within a client’s portfolio.
• On average, 50-60% of customers are contacted each week through campaigns.
• 20-30% of the customers transact on recommendations made by maya.ai in the observation window
• 10-15% customers show repeat visits to the portal each month

• This is on the conservative side of the assumption
• maya.ai has shown value delivery of 5-10% with our clients
• For example, maya.ai has delivered 3-5% spend increase with email and SMS campaigns in 2019 for a Largest Private Sector Bank in India

• maya.ai drives customer engagement and transactions for enterprises powering the age of relevance
• maya.ai has driven portfolio growth with personalized campaigns for India’s largest credit card issuer
  • Increased campaign spends by ~7%
  • Increased click rates on digital channel by 20%

We can go from data to digital in 7 days. Results start showing in 3 months and in 12 months we get to know the full impact of personalization with maya.ai 

We can drive traffic through campaigns. Our APIs can be integrated with marketing automation tools of banks and cadence can be decided jointly with bank on the kinds of campaigns and choice of channels 

We can initially show popular offers based on demographics, location and other contextual information. After we start getting digital interactions on the offer pages, our AI models will start capturing the preferences of the customers and use it for recommending the right offers 

We have a single API that can be integrated with as many channels and digital assets as required 

• Daily transaction data feed  
• API led offer refresh  
• maya.ai integration across digital assets  
• Cross functional collaboration with digital, marketing, analytics teams on periodic basis  

• Clean images, creatives, push campaigns for new offer seasons  

• maya.ai’s key differentiator and one of its key IPs lies in the external data curated by maya.ai
• This external data is hosted in a multi-tenanted cloud environment
• maya.ai’s algorithms work in conjunction with this data & data accessed with a dedicated client VPC
• It is not economically feasible to deploy maya.ai and its complex architecture within the bank datacentres as:
  • Setup takes too long
  • Running costs become high and affect the engagement ROI
• Having said that, on-prem option can be discussed on a case to case basis.

Bazaar is an AI led personalized commerce marketplace that benefits consumer, merchants and banks.

• Bazaar makes the consumer journey personal and relevant by recommending the right merchants as well as products, with discovery to fulfilment in one portal.
• Bazaar helps merchants with a portal to identify and reach the right customers, drive traffic from within the bank’s ecosystem and get consumer insights.
• Bazaar provides the bank with a scalable model for offer acquisition, easy configurability of consumer conversations and automated tracking of performance.

Merchants don’t have visibility to customer insights or the ability to reach the right customers
They see ZERO traction from current bank sites. Also, on-boarding is still manual. Merchant engagement portals have been designed to address these concerns of merchants

Please contact us to know more.

Coming Soon

• Crayon Data can do image and other content curation as an additional service
• However, it is not a scalable process if Crayon Data has to do this in an ad-hoc, services manner
• This process scales quickly if the responsibility for sourcing high quality content rests with the merchant who owns the brand
• The maya.ai engine can automatically detect image quality and reject images that do not meet quality standards for display to end-customers
• This is one of the reasons why merchant-facing portals are so important

• All merchants store product catalogues at an SKU level
• The SKUs have a well-defined hierarchy to capture product attributes
• This is very similar, in construct, to how maya.ai stores merchant attributes
• The existing merchant recommendation algos would work as is for product-recommendations

Yes, we can bring you offers. Please contact us to know more.

• The engine provides a 24×7 self-serve interface where merchant recommendations are available on-demand
• The question is, how often does the bank need new offers?
  • Each month?
  • Each season?
• Anytime the bank wishes to identify how their current offer portfolio is performing and how best to improve it, they simply need to login to maya.ai

The maya.ai & Bazaar platform has been designed to plug into the existing technology ecosystem banks operate in, and works with minimal disruption. Below are the different types of integration the proposition comes with:

• Integration with Bank’s customer and transaction data lake
• APIs and Widgets that integrate into Bank’s digital assets – websites, mobile apps, etc.
• Integration with Marketing Execution systems to enable Email & SMS campaigning
• Integration with Transaction Alert system for trigger-based engagement

There is some support required from banks at different stages of the maya.ai & Bazaar journey:

• Contracting Stage – Support in getting sign-offs from security groups on proposed architecture of maya.ai & Bazaar
• Setup Stage – Some development effort (1-2 weeks) is needed to automate the data refresh from the bank’s data lake into the maya.ai VPC
• Portfolio Management – Support in identifying bank’s priorities as pertains to portfolio growth, definitions of customer segments as identified by the bank (if any) – this helps fine-tune the recommendation engine to cater to the Bank’s objectives
• Personalization API Integration – Some development effort (1-2 weeks per asset) is needed from the Bank’s end to integrate maya.ai’s APIs and Widgets into the desired digital assets – webpages, mobile apps, etc.
• Execution Stage – 1 SPOC is needed on a continuous basis (1 day per week) to agree upon campaign cadence and to ensure execution of email and sms campaigns on third party execution systems incl. collecting templates from the Marketing team
• Cashback Disbursal – 1 SPOC is needed once a month to reconcile the Cashback disbursals into Customer accounts basis transactions driven

• Infrastructure
• • maya.ai’s capabilities are deployed on a Virtual Private Cloud – we currently support deployment on AWS & Azure
  • Bazaar is deployed on a Crayon owned multi-tenant cloud set up and is connected to Crayon’s Global Merchant Repository
• Execution – maya.ai & Bazaar’s capabilities are made available to our clients in three modes:
  • Fully Assisted – Once set up is complete, Crayon will take on complete responsibility for execution & portfolio management
  • Assisted Self-serve – Once set up is complete, Crayon’s Customer Science team will work closely with the bank to co-own execution and portfolio management
  • Fully Self-serve – Once set up is complete, Crayon will hand over all studios, features and execution capabilities into the hands of the bank and will only provide support when needed