Best Next Action: predicting the future

By Philip Lacey. August 2018

Best Next Action is designed to improve sales and customer retention performance. It uses data analytics to help you make the right offer, to the right person, at the right time.

Definition and background

Best Next Action (BNA) is an example of marketing exploiting Big Data by using large volumes of information to predict what should happen next. It is a marketing model that considers the different actions that could be taken for a specific customer and decides on the ‘best’ one, based on historical data.

Best Next Action in action

Essentially Best Next Action uses statistics to define the most appropriate product, service or remedy for the customer and the more data available, the more accurate the response.

How it works

A good, integrated system will identify an inbound customer, push the appropriate record and then display the options available for the agent to offer the caller, however, if no information is held on the customer, the agent should be prompted to gather the data needed to start building a client profile. Best Next Action makes it possible to make informed decisions based on general rules from past experiences but there are no guaranteed results.

Data

Access to the right data is key and the more complete the available data, the more accurate the models will be. Usually, modelling is undertaken in a Data Warehouse where data from across the organisation is collated and cleaned. There must be a proper understanding of why models are the way they are, if not, misinterpretation may go unnoticed with consequent damage to the sales or customer care process.

Creating models

The approach outlined below uses existing data and mathematical models to generate statistical models.

• Case    A collection of facts
• Facts   Raw material for the creation of knowledge
• Knowledge      Patterns identified within the data

• Patterns           Regularities that can be expressed as a rule

For each rule there should be Support (how often this correlation appears) which provides Confidence.  With modern systems capable of reacting to thousands of rules simultaneously it is possible to build models that can adapt instantly by using data mining in real time to see if the rules have changed.  Artificial intelligence uses statistical tolerances which can change rules when if finds better statistical results, this is machine learning.

BNA in operation

There are three ways that rules can be created and applied in reaction to new data, each with their own merits and drawbacks.

Out of hours

New data is processed as a batch, at the end of the business day

Pros

• Useful if complex rules need generating
• Reduces operational overhead
• Batch processing is suitable where systems do not have an API

Cons

• Doesn’t adapt to live information
• May make sub-optimal recommendations

Fast acting

As the call arrives, BNA looks up the key data and runs that data through a set of fast acting rules which present before the agent speaks to the customer.

Pros

• Rules update in real time
• New data added to the record immediately
• Can read from multiple sources (agents, locations)

Cons

• Increased operational overhead
• Potential for error due to inaccurate data
• Database must be available

In-call calculation

Best Next Action re-calculates as the agent talks to the customer. This means that the agent sees the most up to date information possible. New data is run through the rules and BNA calculates live.

Pros

• Rules change in real time
• New data added immediately
• Ensures BNA is the best available

Cons

• Increases operational overhead of real time calculation