ROIex - A platform using predictive analytics to quantify the ROI on CX!

Recently, we put pen to paper to further articulate our thinking around ROIex.


In the spirit of sharing, we have documented these notes below.


What does ROI on CX even mean?


Businesses want to be able to quantify the impact of a positive or negative customer experience.


Hypothesis


Customers who have more great experiences with a company are more likely to stay longer and spend more. Conversely, customers who have crap experiences will leave.


We are looking to prove this hypothesis by building a robust data model which correlates the (positive and negative) impacts of customer experience on the financial performance of an organization - a term we refer to as CX-ROI.


First-Principle Thinking


Customer > Pays money to a company > Company interacts with the customer across multiple touchpoints > Company captures data regarding touch points > Interactions influence the customers perception of business > Customer’s emotions influence spending decisions.


We should be able to build a data model which calculates customer sentiment (CX Score) after every interaction, no matter whether the customer has told us directly, or by inferring it via operational and/or behavioural data variables. This CX score will be correlated back to the Lifetime Value (LTV) of a customer in order to segment customer spending habits according to the experience they have with a business.


Assumptions


In order for CX-ROI to be calculated, these interactions must be scored and somehow linked to operational (transactional) data utilizing a primary key.


When building this model, we should assume that all datasets can be linked utilising a primary key (UUID).


How can we demonstrate ROI on CX?

  • Customers generate value back to a business, usually in the form of payment

  • Customers also interact with businesses, via a number of different channels

  • Phone Calls

  • Emails

  • Webchat

  • Video Calls

  • Instant Messaging

  • Social Media

  • Website

  • Surveys

In order for CX-ROI to be calculated, each interaction that a customer has with a business must be scored and somehow linked to transactional (financial) data utilizing a primary key.


What primary keys could we use to link the data?


There are a number of primary keys that we could utilise:

  • Email address

  • Phone number

  • First+Last Name

  • Social Media Name

  • Existing UUID

  • Customer Number

  • Shopper ID

  • Policy Number

  • Claims Number

Fuzzy matching techniques could be used to link incomplete or partially accurate data.


How might we define a positive CX?

  • Subjective → When the customer has told you that it was positive.

  • Objective → when the business has data to imply that it was a good experience i.e. faster approval processes, quick call answers

Mapping & measuring multiple touch points across the customer journey


The customer will likely have a different experience each time they interact with an organisation. They may, or may not tell us about that interaction directly however operational / behavioural data should be used to:

  • Identify that the interaction has occurred

  • Infer how the customer felt about that interaction (if there is no experiential data captured on this interaction)


How would a CX-ROI tool even work?


For this concept to work, each dataset (experiential, operational and transactional) would need to have two mandatory fields:

  1. Date of interaction

  2. Primary key (UUID) associated with the customer

From that point, an algorithm would digest relevant variables within all available data sources:

  1. Financial / Transactional Data - to calculate Customer Lifetime Value

  2. Operational & Behavioral Data - used to calculate a (predicted) CX score

  3. Experiential Data - used to calculate an actual CX score

The “CX score” concept


It would be necessary to standardise the CX Score used across all types of interactions in order to effectively calculate the ROI on CX.


For example, an 11-point NPS rating, 7-point customer effort rating, 5-point CSAT rating and customer complaint should be transformed into a consistent CX score scale.


CX Score Scale (Proposed):

  • 5 - Great Experience

  • 4 - Good Experience

  • 3 - Neutral Experience

  • 2 - Bad Experience

  • 1 - Worst Experience

The CX score will be calculated using a bespoke algorithm which analyses each different dataset:

  • Objective Data: Behavioural & Operational data points

  • Subjective Data: Customer feedback via surveys and interactions

This model will utilise machine learning and predictive analytics to gauge customer sentiment even if the customer doesn’t provide feedback directly. These CX scores will then be correlated with business transaction (financial) data to demonstrate the impact that CX is having on the spending habits of a customer.


What data will be used to calculate a CX score?


A combination of O & X data sources could be used to calculate this CX Score:

  • Call Centre Data - Average Call Waiting, Average Handling Time, Disposition Code

  • Surveys - NPS, CES, CSAT, Verbatim

  • CRM - Complaints, Compliment, Request

  • Website - Bounce Rates, Thumbs Up / Thumbs Down on content

  • Chatbots - Transcripts and feedback obtained within interactions

  • Email & Social Media Sentiment - Sentiment derived from open-text analytics


Correlating CX Score with Lifetime Value of Customer


Once we have determined the CX Score for each customer after every interaction, we will longitudinally map each interaction and overlay this with transactional (spend) data.


At an organisational level, we will be able to track the lifetime value of each CX Segment:

CX Scores

Lifetime Value

Customer Length

1 - Worst Experience

​$15

0.2 years

​2 - Bad Experience

$45

0.6 years

3 - Neutral Experience

$62

0.9 years

4 - Good Experience

$90

1 year

5 - Great Experience

$120

2 years

Considerations / Problems to Solve

  • Ability to tie interactions back to a customer if the primary UUID is not available

  • Ability to link a survey to a particular interaction if the survey response is received on a later day

  • Ability for algorithm to automatically attribute a CX score based off of data indicators or open-text verbatim

If you'd like to get involved on this concept, please drop me a note.


Cheers,

Ben the Bearded Man







16 views0 comments