Predictive analytics goes a step beyond using data just to ascertain the cold, hard facts, and instead delves a little deeper. The aim of predictive marketing is to use this data to learn how customers are behaving and to figure out which marketing actions are going to be the most successful to engage customers in future.
“Essentially, it’s the use of computers to help spot patterns that exist within data, typically sourced from multiple disparate or inter-connected systems,” explains Jon Clarke, CEO of Cyance. “These can then be used to help provide insights or predict future outcomes, based on historical behaviours or transactions. The best way to look at predictive analytics is to see it as a better use of data that provides real-time or near-time actionable insights to drive more effective marketing.”
As Glen Westlake, CEO and co-founder of BrightTarget, puts it: “The first step to predicting a customer’s behaviour is understanding what their needs are and what their response will be to your marketing actions.”
What are the biggest misconceptions around predictive analytics?
According to our experts, many marketers believe that predictive is simply too challenging an approach to implement, but with the variety of tools available, rolling out a predictive plan is actually pretty doable. “People assume it’s too advanced for them to get started or that it’s this unattainable ideal, which isn’t true,” explains Jennifer Roubaud, UK and Ireland country manager for Dataiku. “There are many tools out there today that make predictive analytics really accessible, even for small teams; it’s just a matter of making that change and aligning the organisation around the need and execution.”
What are the potential benefits?
Predictive analytics enables marketers to see a customer painpoint before it has created a negative result. And seeing the problem can allow marketers to act quickly. “Marketers are able to see how certain ‘actions’ such as the customer support steps taken, service experiences, or specific answers to questions, affect outcomes”, says Sid Banerjee, executive vice chairman, founder and chief strategy officer at Clarabridge. “They can then train their employees to better respond to customers to create the best action, and ultimately, achieve the best result.”
What are its limitations?
Jon argues that some predictive analytics models rely too heavily on single-source, and very narrow, third-party buying behaviour data sources, which discriminate where buyers go to do their research. “They are often limited by a very small arrangement of behaviour topics which can skew the results from predictive models and segments,” he says. “Some predictive models also only take a snapshot of behaviour within a very small timeframe. It’s important to spot trends over time and assess what’s just normal activity versus behaviour relating to a buying journey.”
How can predictive analytics be applied successfully in B2B?
- Boosting personalised nurture campaigns
Cyance has seen success with predictive analytics after testing its feasibility for achieving better results from demand generation campaigns. It did this by generating more high-quality leads for less effort. “We’ve used predictive to connect web intent signals with our customers’ own marketing campaign event and sales CRM data outcomes and have been able to help them build personalised nurture campaigns and customer journeys,” says Jon.
Cyance used its experiences from more than a decade as a B2B demand generation agency to help it join up the dots between how people in B2B organisations progress through the customer buying journey, where they go to do their research and using these insights to build better predictive models.
“We tested this hypothesis over many months, using different data sets, marketing channels and sources of customer CRM and third-party behaviour data to establish the best combination,” Jon adds. “We recognised the need for very tight integration with all demand generation activity once the data was passed over and the need to ensure all outcomes are factored into the future models.
“The results have shown a typical 300% increase in MQLs with campaigns, reduced churn by as much as 40%, and increased conversion from marketing leads to sales pipeline by up to 35%. In many cases, these results have been achieved while reducing associated costs by as much as 66%.”
- Using predictive to drive a master data solution
Data company Dun & Bradstreet recently used predictive analytics to target attendees for its Master Data forum events. “Our customer analytics and insights team used Dun & Bradstreet’s own B2B data, customer behaviour, product propensity models, lookalike models, and location analytics to support the campaign,” explains Rishi Dave, CMO, Dun & Bradstreet. “These variables were used to identify the best (highest propensity) companies to target for this campaign and helped locate where buying power exists within those companies.”
Dun & Bradstreet was able to pinpoint and prioritise a list of companies with a definitive need for a master data solution, as well as provide a geographical view of market concentration (the density of target companies within a particular geographic area). “For example, there may be 47 companies who need a master data solution in London, but only 15 potential targets in Dublin,” explains Rishi. “This affects how we approach various marketing strategies, like where to host key events, for instance. The results were positive and we saw a 29-fold increase in pipeline return from the campaign. But it’s still too early to determine the full ROI.”
How should marketers select a platform/provider?
One of the most important starting points is to determine how providers build their predictive models and which data points they use. “Think about where your customers go to do their research and ensure the vendor can factor that into their modelling,” says Jon. “Check for evidence of results from customers to ensure they can deliver on the promise. Find out if they truly understand B2B buying behaviours and customer journeys. Have they any experience and history with B2B marketing, for instance? And how has this been used to help build better predictive models that improve what you are currently doing today?”
In short, as long as marketers avoid relying too much on single-source and very narrow third-party buying behaviour data sources, predictive can deliver promising results. Select a vendor that uses the correct data points and understands where your customer carries out their product research, and you’ll be well on your way to predictive success.
