As the technology edges closer to maturity in digital, here’s an introduction to predictive analytics.
What is predictive analytics?
Predictive analytics is the use of models to predict future outcomes. In business and marketing, this entails the use data and statistics to stereotype customers.
A user’s profile and behaviours should allow companies to predict how that user will behave, by looking at the historic behaviour of other users.
This means a company can change how it communicates with that user or discount them altogether, based on the user’s propensity to undertake a particular action.
Machine learning continually refines a predictive analytics model as more data (such as the success of its outputs) is fed into it.
It’s a relatively simple concept but one that has many implementations. I’ve listed a few below.
How can I use predictive analytics?
Okay, this isn’t a marketing implementation but it is one of the early uses of predictive analytics.
Assigning a customer a credit score is a prediction of the likelihood of repayment.
Increasing the accuracy of the prediction allows for more revenue to be made as fewer customers default.
Predicting customer lifetime value (CLV)
Predictive analytics can help to gauge the likelihood of repeat purchase or even CLV.
Whilst a simple RFM matrix (recency, frequency, monetary value) is the basis of prioritising customers, so much more data is now collected by businesses (and available from third parties), that CLV can be based on much wider metrics.
Predictive analytics greatly augments a simple RFM matrix
Estimating churn propensity
This is the other side of the CLV coin. Companies such as telcos want to know if customers are set to cancel their contracts or jump to another provider.
Knowing this allows these companies to incentivise such customers to stay, and to forecast more accurately.
This may be the most infamous example of predictive analytics in marketing and retail, thanks to the apocryphal (?) tale of a teenage girl being sent maternity offers by Target before her own father even suspected she was pregnant.
Supermarkets that use loyalty cards have so much transactional data on their customers that they can create incremental revenue with targeted offers, encouraging more frequent shopping.
Display ad targeting
Just like with ‘old-fashioned’ paper vouchers, programmatic advertisers want their display ads to achieve a high redemption rate (clickthrough and eventual conversion).
Because programmatic advertising allows targeting by demographic and behaviour, predictive analytics can be used to optimise this targeting.
An oldie but a goodie – ascertaining which of the leads delivered to your sales team should be prioritised.
Where tech and data is being integrated and shared across Sales and Marketing, these departments are working closer together to maximise efficiency.
Analysing reviews, social media commentary, call centre scripts etc. can allow companies to produce sentiment analyses which in turn can be used to suggest improvements to communications.
30% of Amazon revenue reportedly comes from its recommendation engine (various sources).
Netflix and Spotify generate vast engagement through recommending content. This is all predictive analytics.
Optimising PPC campaigns
How does the weather affect clickthrough rate might be one question to ask.
There are, of course, many other variables that affect keyword volume, clickthrough and conversions, which can be optimised for.
For big retailers and manufacturers, being able to forecast demand and therefore also the requisite inventory will prevent popular items from selling out and impacting revenue.
A broad area of predictive analytics which encapsulates many of the other examples in this post.
When predictive analytics is plugged into CRM and marketing automation tools it helps to divide up your audience into segments, to which tailored messages can be delivered.