Big Data Application in Effective Customer Marketing & Satisfaction
Depending upon the overall Importance Level of the given customer and/or by predicting upcoming call priority in advance, incoming call can be “Automatically & Intelligently Routed” by “Recommended Call Action System”; which can decide what treatment should be given to this upcoming call. Based on complex factors , it can either leave it in queue or route to one of the “specific agents” which are ranked in increasing level specialty in handling customer calls in pro-active manner or do some other action.
Driving Principle for Method
-Best /Priority Treatment to customers who brings maximum value to business
-Avoid waiting time for valuable customers or customers who are calling again for same issue.
-Treatment to be predicted prior to answering the call to customers by call center agents.
Call Center is front for modern Business marketing, the treatment received at call center greatly influence the relationship customer keeps with the Business. (1).
Call Center is key to all Customer Interactions but it normally works on First Come First Served mechanism or based upon network bandwidth or available resources. This is huge challenge since not all customers are equal. With limited resources & practically impossible goal to satisfy every customer for call Center; However as noted in article (2) by Mark Smith that “8 out of 10 organizations reported that customer satisfaction is a top-three business issue.’ . Also as noted by Elena Dobre in article (3), the most important metric for call center resolution is “First call resolution Rate”
So there is critical need for Big Data Analytics based solution to Intelligently & Automatically Route Incoming Calls
The biggest challenge is to respond to each incoming call of customer in unique way & in real time & pro-active manner BEFORE the call is answered by agent.(Prior Arts1/4) The analytics system need to be dynamic enough to handle varied CRM challenges like….
Not All customers are equally important
Not All Calls of Important Customers deserves same priority every time
Not All customers react same way for waiting in queue
Not All the times Call Center is loaded with equal call traffic.
Not All call center agents are same, only few trained enough to handle high profile customers
Customer may be important but his current call may not urgent or a customer may be average but his current call may be very important. The expected solution need to be taking into account all above factors before recommending Decisions’ for Call Action.
A “prescriptive analytics” for call routing solution means to decide the treatment for call viz. which call should be queued and/or priority level for queuing out a call. Also preferably the selective treatment is to be given to the call before an agent of the call center answers the call.
The typical outcome could be forward call to a recorded IVR system or give busy tone or attended by Call Center agent or if call is too important then responded by Personalized Relationship Manager.
- Customer will call up the call center and based on call “Importance Level” the call will be either routed to IVR or call center.
The importance level of incoming Calls can be the output of this method. The method shall give the importance level of customer in the business context as “Call Priority” as below
- Inside the call center the call will be routed by the skill-set of call center agents. The three kind of call agents could be
- Call Center Agent (Level1)
- Call center Supervisor (Level2)
- Personal Relationship Manager (Level3)
- The possible recommendation for Call Actions by Call Routing Analytics system could be(in 1-1 correspondence with call priority of (I) & II )
- Normal Call Processing (Process Call in Queue)
- Refer /Escalate Call Processing (Take call out of queue & respond)
- Expedited Call Processing
- Depending upon the overall customer Importance Level & the predicted call priority , incoming call will be “Automatically & Intelligently Routed” by “Recommended Call Action” to either of the agents with which are ranked in increasing level specialty in handling customer calls in pro-active manner.
– Technical Goals
- Overall Call priority is calculated by below equation which takes into account various deterministic factors & predictive models.
- Call Priority = ( ∑Customer Importance Score * ∑ Call Volatility Ratio * Risk Propensity to Leave )
Call Priority Equation= ( ∑ CIS * ∑ CVR * $RAP )
All the numerator inputs of the equations are normalized in the range of 0-1. The end result of equation hence will be normalized in the range 0-1 which 1 indicating utmost urgency of the call and 0 being lowest priority.
1.Customer Importance Score
- Customer Importance Score is Metric derived by below Predictive models. The weighted average is left for configuration of Business user.
- Customer Satisfaction Model
- The customer satisfaction model is text book concept & its definition varies from company to company & can be customized by parameters specific to a company. For example (6) ACSI defines customer satisfaction index as per industry but most companies prefer to construct their own feature vectors to arrive at this matrix.
This is a classification model built upon the customer profile by taking input such as, net promoter score of customer, customer responses to past offers & demography of customer. The classification method used can be any classification algorithm. (CRT/C5 etc)
- Customer Profitability Model
- This is a classification model built using customer past payment standing with company. This may take into account factors such as customer plan, cost to business, profit to business, income level of customer & any past delinquency from the customer. Customer profitability is again a company specific concept as different factors can determine its exact definition. At very basic level its always the gap between cost & revenue & all companies wants to maximize it. So we know that this factor needs to be maximized. Many more factors can be used as described in this article http://www.destinationcrm.com/Articles/Editorial/Magazine-Features/Predicting-Profitability-43446.aspx
- The classification method used can be any appropriate classification algorithm. (c5/CRT).
- LTV (Life Time Value) Model of Customer
This can be a Cox regression (7) model built upon the customer profile, to provide the life time value of the customer to the business. The life time value of the customer is the concept widely used in Telecom sector to determine how much customer is worth over a long period of time & what’s the value that one can get over his survival lifespan. It’s being described at detail here http://jsr.sagepub.com/content/9/2/139.short
2. Risk Propensity Score
- Not all customers react same way to waiting in queue. Customer Behavior for waiting to be attended is different for each customer. Some are have higher risk propensity to leave/ attrition/churn the business as opposed to others.
- Based on past customer responses to call center predict the risk propensity of customer likely to leave business if he is not attended within the given time-span. This can also take into account previous call center logs which has information related to this customer & similar customer like this. This will lead to customer profiling information that can be used to match current customer with them.
- The algorithm used can be Neural Net algorithm to calculate propensity score as it will need to be adjusted to a number of random variables which may or may not have any clear relationships between the variables. Below paper describes how neural net can be a good choice for customer profiling based risk prediction http://link.springer.com/chapter/10.1007%2F978-3-642-12189-0_31
3. Call Volatility Ratio – CVR
- Not all calls by all customers (even if they are important customers) are of equal significance each time, few calls needs more urgent attention over others. Predicting individual call priority by anticipating its motive is thus the important factor.
- Customer making calls 3 times in same day or multiple times in same week for same complaint then such is serious call as compared to others.
- A Relative Age of Recent Calls with weighted average & time inverse functions need to take into account recent calls made by given customer in recent past.
An algorithm for “Finding Frequent Items” which takes into account “Relative Age” of calls can be used. Counter based “Lossy Counting” algorithm is good candidate for this. The algorithms is described here http://dl.acm.org/citation.cfm?id=1341433
4. High Level Technical Approach:
Diagram for Technical Approach
- Related Prior Arts:-
- USPAT5828742-The above art discriminates between the different callers based on a recognized ringing pattern. This art is not related to propose one since it concerns only personal systems and not a call center mechanism.
- USPAT5999613 -The above art routes the incoming calls to the free operator, this is not related to proposed art as routing system is just first pass the post and neither predicts importance of the call not the customer.
- USPAT6160877-The above art if just a call forwarding system and hence not related to proposed art
- US5684872–The above art tried to do “Prediction of a caller’s motivation as a basis for selecting treatment of an incoming call” The above art does not use any Analytics or Predictive Analytics method in the derivation of its caller’s motivation but uses a static flowchart rules to determine caller motivation points. This is huge drawback of system as its not scalable, not usable for call center operations and not accurate as it does not use any prior analytical methods.