Ask the DM Expert: David Kelly
Predictive Analytics key to getting to the right customers according to David Kelly, Managing Director of IAS
Predictive analytics seeks to identify the behavioural trends from transactional data in the past to predict what might happen in the future. Or to put it another way, to find clues as to what the future might look like, by looking in the consumer transaction data rear view mirror.
Here is what David Kelly, Managing Director of IAS Ltd had to say about Predictive Analytics as part of our Ask the DM Expert series:
Predictive analytics involves using a wide range of mathematical formula to make the best decision for improved customer targeting. It comprises traditional statistics such as regression, cluster, classification and factor analysis including machine learning and genetic algorithms.
It answers the following key questions most organisations need to know:
- Who will respond to my offer and who will convert
- Who will defect/churn
- What are my consumers willing to spend over their lifetime with my brand
- Who are my potential bad debt customers in the future
- What resource allocation required for different customer segments, channels
- What transactions will be fraudulent ones
It provides the following benefits:
- Identifies most profitable customers
- Segments your customers on dimensions such as high value and high risk or low risk
- Help identify needs and improve customer service
- Detects abnormal patterns in behaviour
- Helps you make better faster and better decisions
Successful companies at predictive analytics only use it as part of an overall customer management framework and are focused on gaining a 360 degree customer understanding that facilitates the RDA (Retain, Develop and Acquire) cycle of customer relationship management.
So, where does a company start when deciding to go down the route of datamining or data analytics as it’s also known?
The start position is auditing where you are now (tools, skills, processes) and where you want to get to in your overall integrated customer understanding or closed loop marketing capability. For example, if you want to know your customers potential value, customer profitability, customer lifetime value, customer churn potential, customer propensity to purchase. Having the right tools and skills/resources in place will deliver the ROI required when targeting key customer segments.
There are many tools available to help organisations on the predictive journey and the “Gartner magic quadrant in predictive analytics” is a good place to start in accessing what’s available and what’s works best for my business and budget. However, it worth noting that the choice of technology and the technique used depends on:
- the business problem
- whether the problem is predictive or descriptive
- underlying data environment
- types of variables to be predicted or described
- ability of organisation to implement a solution
- whether key statistical assumptions hold
In summary, it all about the problem trying to be solved, never about the technique or the tools
My experience is that it’s a combination of approaches that often works best.
Below I have outlined the steps (CRISP) involved in a typical predictive/datamining project.
Figure 1.0 CRISP Model: The lifecycle of a Predictive/Data Mining project
Having worked with a number of large organisations in Ireland and UK, over the past 10 years many have already implemented CRM front-office capabilities to provide immediate customer interaction feedback but are missing the value generated by analyzing new and existing customer information. Without predictive analytics companies lack the ability to discover meaningful correlations, patterns and trends required to recognise the ROI from existing front office investments (CRM) and maximising the value in the existing customer database.
Guide to having a successful analytical capability within your organisation:
- Undertake a CRM audit: this will look at your current CRM capabilities in-house from needs based segmentation, campaign planning through to insight generation to predictive tools deployed and identify the gaps in where your organisation is and best practice, and where it wants to be in the future.
- Benchmark your business analytical capabilities against your competitor set: Find out what your competitors are doing in terms of push vs. event driven marketing, customer campaign planning integration, social network analysis, predictive analytics, Customer lifetime value modeling and join up the organisations goals in having a sustainable customer insight generation setup.
- Customer Insight vs. Customer Research vs. Community Insight: Know the structure and skills required for these roles because they are different and with the advent of social media it is never been more important to understand both what customers think (research) and what customers actually do (listen to conversations without interruption, look at their ideas and interact and evolve your product or brand with them). This new channel is a rich data source for marketers and has introduced new datamining techniques such as “text mining” and “sentiment engines” but typically require a large or high level of human input and are very labour intensive. Analytics has a role to play here to decipher if the buzz or sentiment been expressed about a product/brand actually is linked to real world behaviour?
- Integrated Marketing Capability Review: As previously stated analytics forms part of an overall structure from Reporting, Analytics, Marketing Planning & Execution, Next best Action. Find out what tools are best in class in each of these categories based on complexity of data within your organisation, business problem identified and budget (capex or opex) available.
- Engage with the experts: the development or deployment of predictive analytics as part of an overall customer management framework is key to implementing it successfully within your organisation. It requires a vision and detailed roadmap outlining phases and deliverables.
- Outsource or In-house Capability: you can start earlier in the cycle by deciding to outsource the building of churn models or response models to a third party. There are a number of players with various skills and costs but it does get you up and running with an output (a targeted DM campaign) within 2-3 weeks depending on the quality of your data. It also avoids the immediate upfront costs for software or hiring of skilled analyst to run models.
Why predictive analytics is so important
Industry best practice for predictive analytics employs a holistic approach and combines churn models and response models to prioritise customers to target. It segments customers for understanding and uncovers triggers for input into campaign strategy. Finally it calculates contribution or CLTV to help optimise the right offer to the right person.
Want to ask David a question?
Just email your question to email@example.com