Marketing resembles more a numbers game these days where fast and reliable analytics replaces guesswork and gut-feeling – pretty much in the same way as it has been happening in stock market trading, risks management in banks or insurance underwriting.
Trying to figure out how to convert the Predictive Analytics buzz into increasing sales and retaining customers?
Here goes a “cheat sheet” with the core most frequently implemented applications:
“It’s about Insight”
Before we dive into the Roadmap in more detail – we need to clarify one thing:
Predictive Analytics – it’s not just Modelling
In fact – there are more “non-modelling” Predictive Analytics projects delivered than actual models. Why? Because for every Model that you need, there are several more questions that Predictive Analytics can answer but which will not require a deployed Model at the end of it.
It’s about the “insight” – getting an understanding of why our customers leave us, where are the design flaws in our onboarding process, what customer profiles convert and so on.
A Predictive Model follows if you want to implement tactical reactions to customer’s propensities, e.g. in a form of a proactive retention campaign – but it’s not always the end-goal.
If you have a long list of potential customers and you lack data-based prioritisation logic of who to call – a Predictive Model for Lead Conversion is your answer. It will attribute a “conversion propensity score” to each Lead and all you need to do is to work the list from top down focusing on the high propensities and ignoring the low ones.
For more mature sales process you may need to analyse and model each step of the sales funnel.
If you have a problem with customers signing up and cancelling soon-after you may want to understand why. Predictive Analytics is a perfect weapon for this task. The algorithms will slice and dice the data automatically, mix demographics, sales process, product usage patterns and will help to understand why the cancellations happen.
For some products and services you can observe customers paying but not using them. It’s a customer churn waiting to happen. You don’t need a Model to take action but you will probably want to know what kind of customers do that.
You are likely to have different versions of your product/service with different
Up-sell models will show you who is actually interested in paying more for better.
They are easy to build, quick to implement and almost always very successful.
If you have several products that can be consumed at the same time – Cross-sell models can segment your customers into “interested” and “not interested”.
Working hand-in-hand with Up-sell Models, they are a powerful tactics for increasing revenues and customer loyalty.
“Some customers decide months before the renewal date that they will cancel”
Retention Analytics spans much wider than just Predictive Models. This is a major decision-making space where you need quick answers to many questions and not all of them will convert into deployed models.
- What drives our customers to cancellations?
- Why some of them renew and some don’t?
- What can we do to retain them?
- Why some of them lower their spend?
- Do our retention activities need to be different in special seasons – like Christmas?
- Have customer churn patterns changed in the wake of the recent competition launch?
..and so on.
While you are gaining more insight, you can start deploying tactical models for different churn types:
- Subscription churn (cancellation)
- Early churn
- Usage churn
When trying to retain customers – the single most important factor is timing. If your models predict churn for tomorrow – it is already too late for any retention action. Some customers decide months (!) before the renewal date that they will cancel.
How do you deal with that?
By building Models with much longer time horizon and analysing what leads to such early decisions.
Armed with that insight you can improve your offers and processes to keep your customers happy and away from the danger zone.
New Era for Marketing
Predictive Analytics enters a new level of automation allowing to get fresh Models & Insights within few days instead of months.
This creates a new playing field for marketeers where they can ask tens of questions instead one or two and get up-to-date answers quickly without the pain of scoping every time a multi-month project.
It’s a numbers game and we hope the Predictive Analytics Roadmap will help you to navigate through it.
Xpanse AI Team