Overview

Next Best Action (NBA) is a new artificial intelligence and machine learning tool built directly into Atlas that helps Care specialists identify customers’ needs. NBA provides specialists with the “next best action” to discuss or recommend offers and promotions that are tailored to a customer’s specific needs.

  • NBA anticipates customer’s needs in real-time based on hundreds of data points, such as their account set up, network experience, and tenure
  • NBA analyzes these account data points along with customer behaviors and patterns to provide the right solution for the right customer at the right time
  • NBA is designed to save experts time on visual audits and provide more opportunities to upsell or save a customer

Defining the Problem

There is a R1 version developed and now the product team is looking to expand functionality. Design and research will be used to inform a redesign of the tool to account for a regrouping of offers by recommendation rather than by account or line.

Goals

In collaboration with internal design and research teams, product management, business, and development partners:

  • Enable proactive recommendations for the most relevant actions and ensure the customer gets the right solution at the right time
  • By levering AI and machine-learning capabilities, systematically infer customers needs (based on hundreds of attributes and signals), and use that inference to prescribe actions (e.g. new offers) to positively impact key Retention and GA metrics
  • Identify and promote Nex Best Action to be taken for a customer in real-time while the specialist is engaging with Care or Digital Channels
  • Ensure only eligible offers are being promoted
  • Present new recommendations in real time during the interaction session if the context changes considerably
  • Automate provisioning process to reduce customer and specialist effort

Requirements

Need different options for prioritization of NBA offers:
  • Customer who’s calling in first, then by priority
  • Offers across BAN and not top of each line (max 3 offers per interactions)
  • Show all recommendations at once and let the specialist guide their own conversation to bring up offers when they want to in the order they feel is more natural
  • Show only top offer and show the next one after saving the previous one
  • Correct the info bar to make it more noticeable when there are unsaved offers

Guiding Principles and Metrics

Guiding Principles:

  • Right customer: target elevated and relevant opportunities         
                    Elevated – targets customers who are likely to perform a certain behavior (e.g. add a line)
  • Right solution: drive business value that’s aligned with Care priorities     
                    NBA considers the value of each individual recommendation in terms of unit economics  value to improve the customer’s value
  • Right time: identify the right action in real-time NBA reads account and  customer data attributes at the moment the call is handled, ensuring up-to-the-date context. NBA uses hundreds of data points that go beyond
  •  

Metrics:

  • Maintain and improve site score on performance and SEO
  • Increase revenue, customer base, and basket size
  • Improve site accessibility

Use Case Study

Concepting and Brainstorming  
Regrouping of multiple offers by recommendation rather than by account or line

Initial Drafts
Interaction model, selection mechanism, toggle atribute

Research

With a R1 proof of concept already developed, the product team looked to expand functionality. Research was used to inform a redesign of the tool and to account for a regrouping of offers by recommendation rather than by account or line.

The design and research teams conducted a research test + prototype during the project to iterate and improve the design, and to understand the users’ needs:

  • Atlas – Next Best Action – redesign exploration
    3 use case/8 pages = BAU w/list, matrix w/blade, matrix w/pan handle
  • Investigate the interaction design of the NBA tool (panhandle vs accordions)
  • Determine optimal selection interactions (matrix vs radio button list)
  • Explore toggle options between view by subscriber and view by recommendation
Link to Next Best Action research readout

Storyboards
Single offer – account and line level

Final Design
Single offer by account or line – NBA icon

UX Quality Assurance Testing
Next Best Action R2

Solving the Problem

An iterative agile process, where our design was continually revised and informed by feedback every step of the way.

  • Redesigned, overhauled and launched tool in 5 months
  • Improved the disposition content
  • Updated the navigation/tab NBA icon to display a numeric badge when an opportunity is identified
  • Leveraged draft wireframes for user research prototype
  • Performed user testing for 3 design concepts with 4 experts
  • Pan handle preferred with matrix selection – more organized and requires less scrolling
  • Enhanced AI/ML in selling with opportunities to upsell
  • Improved customer support with relevant recommendations
  • Grouped non unique offers to save customer’s response with less clicks
  • Drove business value

Outcome

Research findings determined that Care specialists believed NBA would help improve their audit and call out opportunities for the customer. They liked the value of the suggested wording section, saying that it would help in transitioning the conversation to a new topic. The pan handle interaction with matrix selection was preferred however further R2 feedback deemed the design too complex with having multiple actions to go through on one call. For simplicity the final R2 version supported one offer by account or line rather than by multiple offers.

Reflection

All of this was accomplished by using AI to enhance the Expert desktop experience to automate repetitive tasks and provide actionable insights to better serve our customers.

I learned how machine learning can bring simplicity and humanity to the customer experience. The work we’ve done is purposeful and meaningful and presents capabilities which delight customers and unleash efficiency.

Testimonial

I want to express a huge ‘thank you’ for the work that UX and research teams have been doing for Atlas, and NBA in particular! I’ve been tremendously impressed with the effort that Karen and Thomas put into NBA Release 2 project. Their input was exactly (or even more) what a Product Manager can hope for:

  • High quality and quick turnaround
  • Bringing pain points that weren’t thought through yet
  • Think about ‘what’s next’
  • Think about end-to-end experience