Predictive Data Analysis and Modelling: How does it enhance marketing and market research?

Predictive marketing analytics would have been adopted years ago – if only the computing power were more omnipresent, the data would be more accessible and the software would be more user friendly. Now "predictive analytics" itself is almost a buzzword, after almost 30 years of marketing tracking with a backward glance.

Even medium-sized enterprises often still operate their marketing "scoreboards" in Google Sheets or One Drive ... "throw it into a spreadsheet" still works today, well over 30 years after the founding of Lotus Software.

But companies looking to the future tend to know more than just what happened in the past. "Scoreboards" (most analytics and tracking tools) do not tell you what the score is going to be. Some of our recent articles entitled "AI for marketing" have gained readership because more and more executives are looking for ways to look forward with their numbers, not just back. SAS defines the term correctly:

Based on historical data, predictive analytics is the use of data, statistical algorithms and machine learning techniques to determine the probability of future outcomes. The aim is to go beyond knowing what happened to provide the best assessment of what's going to happen in the future.

Five Current Predictive Analytics Applications for Marketing

While a full list (and subsists) could extrapolate 20 or more individual use cases, we have highlighted 5 current predictive analytics applications that marketers should be familiar with today:

Predictive Modelling for Customer Behavior

Predicting consumer behaviour and expectations is the hallmark of businesses such as Amazon and eBay (see our eBay machine learning interview here), but the technology is becoming increasingly available and important to smaller firms too.

It would be an extensive and cumbersome process to create a complete catalo of predictive models, but there are a number of relatively simple types of models that apply well in the marketing domain. Ailene, a predictive marketing firm based in Silicon Valley, identifies three primary classes of predictive models:

·         Cluster models (segments) – Used to classify customers; algorithms segment target groups based on various factors, ranging from demographics to average total order. Common cluster models include behavioural clustering, product-based clustering (also known as category-based clustering), and clustering based on brands.

·         Models of tendency (predictions)-Used to offer "real" predictions of customer behaviour. Popular models include lifetime predictive value; probability of engagement; tendency to unsubscribe; tendency to convert; tendency to buy; and tendency to churn.

·         Collaborative filtering (recommendations)-Used to recommend products, services and customer advertising based on a variety of variables, including past buying behaviour. Popular models (such as those used by Amazon and Netflix) provide guidelines for up-sell, cross-sell, and next-sell.

The main methodology used by companies for predictive analytics is regression analysis in its different forms. Simply established, the analyst performs a regression analysis to determine the intensity of the associations between different consumer variables with the purchase of a particular product; they can then use the "regression coefficients" (i.e. , the degree to which each variable influences the purchasing behaviour) to construct a score for the probability of future purchases.

Results for predictive modelling, like so many predictive analytics approaches, are highly dependent on proprietary data, but as outlined in the next four applications, there are several common ways in which this information can be transformed to results.

Qualify and Prioritize Leads

A recent study published by Forrester identified three categories of cases for B2B marketing use reflecting early predictive success and laying the foundation for a more complex use of predictive marketing analytics:

1.       Predictive Scoring: giving priority to known opportunities, leads, and accounts based on their probability of taking action.

2.       Identification Models: Identifying and recruiting opportunities who have similar characteristics to current clients.

3.       Automated segmentation: Segmenting leads to custom messaging.

Each of the above issues leads in the preparation and prioritization, and doing so groundwork allows teams to apply the following techniques. Sales personnel who can prioritize are more likely to buy (or more likely to move the sale forward with specific next steps) will be in a better position to close more often.

It should be noted that while predictive marketing capabilities are becoming increasingly accessible to start-ups and small businesses, these techniques require a high volume of sales in order to adequately build and train a predictive model. While even small companies can generate trillions of impressions or millions of clicks and low-ticket eCommerce product purchases, face-to - face sales data is more difficult to collect for smaller or younger companies. This potentially places larger firms in favour of successfully yielding a return on investment from techniques involving data on sales.

Bringing Right Product / Services to Market

Data visualization is a valuable tool which not only appeals to the eye but can also be used to inform, inspire and guide customer behaviour-based actions (and other business information).

A brick-and - mortar marketing department, for example, should use all the consumer information available to make data-based decisions on which goods and services are best to bring to market. By using data visualization to show what types of customers live in the neighbourhood of a store, teams can hone important guiding questions: are they buying more hard or soft goods? Is there a density of age range showing what should be stocked? Does the desired product makeup change as you move in or out of competitor locations?

Targeting the Right Customers at Right Time with Right Content 

Targeting the right clients with the best offer links back to customer segmentation at the right moment. This may be the most common marketing application for predictive analytics, because it is one of the "simplest" and most direct ways to optimize a marketing offer and see a quick turnaround on better ROI.

According to an Aberdeen Community report, consumers of predictive analytics are twice as likely to recognize high-value customers and market the proper bid. Your data sets issues, and best practices dictate that you use historical data to segment and target the behaviour of existing customers, and use that same data to create personalized messages.

A variety of predictive analytical models can be used in this application, including affinity analysis, response modelling, and churn analysis, all of which can tell you, for example, whether it's a good idea to merge digital and print subscriptions or keep them separate, or help you determine content that should be charged a subscription fee versus content that should be given a one-time purchase price.

Many vendors are offering a marketing cloud platform, like Salesforce, through which marketing teams can build audience profiles by combining data from multiple avenues, from CRM to offline data. Feeding appropriate data into the system and monitoring actions over time creates a behavioural model that helps teams to make long-term, evidence-based decisions in real-time.

Driving Marketing Strategies Based on Predictive Analytics Insights

Certain drilled-down applications for predictive marketing research, in addition to those described above, include:

·         Accessing structured, internal data

·         Accessing data on social media

·         Applying scoring behaviour to customer data

Predictive analytical insights provide an effective tool for dealing with "channel proliferation and changing buyer behaviour;" all of the above applications could be used to determine whether a social media marketing campaign will have a greater impact, or whether a mobile marketing campaign is more appropriate for the target public.

Applied to social media data, text analysis and sentiment analysis is another example of capturing insights which can be used to help drive marketing campaigns and future product creation.

Existing Market Research on Predictive Analytics for Marketing

Predictive analytics, like machine learning, is getting a lot of attention. While predictive analytics have been kicking around in the business and marketing world for longer than ML, this year only seems to have been when teams using some form of marketing analytics platform are beginning to outpace those who still choose to go without. A few years ago, Forrester conducted a report detailing two major problems facing modern marketers:

·         Make messages more personalized and appropriate for increasingly selective buyers

·         Create tailored marketing strategies that include a wider spectrum of key decision-makers and influencers in the buying process sooner

The figures are in, and predictive analytics tend to have the ability to double indicators of marketing performance in consumer engagement and targeted revenue in the B2B and B2C industries. Analysis by Forrester yielded three main findings:

·         Marketing predictive analytics use correlations for improved market outcomes and metrics

·         Predictive marketing analytics helps marketers play an important role in their businesses

·         Predictive marketers use innovative tactics to further effect across the consumer life cycle