How CRM Predictive Analytics Create Competitive Advantages
- Data becomes much more actionable when it advances from historical to predictive.
- Without CRM predictive analytics, the view and information for every person in your company is entirely backward looking.
- The ability to convert data into actionable insights is one of only four sustainable competitive advantages.
Predictive analytics analyze historical and real-time data to show patterns, relationships, trends and anomalies and create simulation, propensity and predictive models. These Customer Relationship Management analytics orchestrate data and extend the trajectory of data to deliver forecasts. This changes reporting from backward to forward looking.
The ability to apply data for 'what-if' scenarios or pro forma modeling is a powerful lever to compare competing alternatives and allocate budget and scarce resources.
In fact, as shown in the below chart, Gartner research data found that Predictive Analytics is the single greatest technology capable of delivering competitive advantage.
Forward looking reporting is also a sustainable competitive advantage because unlike products or services which deteriorate over time, accurately predicting the future never loses its value.
The path to the most effective customer-focused predictive reporting starts by identifying the use cases that deliver the greatest business impact. It's been our experience that when helping clients create CRM predictive analytics it's best to start with some real-world examples that are easily relatable and likely repeatable.
For example, creating an offer propensity model is valuable to most companies. Knowing what types of offers to deliver to what types of customers will achieve the most profitable conversions.
We start by merging and analyzing customer data from online behaviors, product purchase history, look-alike segments, loyalty program tracking and other sources such as customer intelligence. We then surface the variables that impact the desired results, test them as coefficients to understand their individual impact and apply uplift modeling to determine which customer segments will respond to offers in a way that maximizes the company's margin. For example, consider the below four uplift modeling segments.
- Offer-induced customers. This target audience is sometimes called the persuadables, as these customers make incremental purchases if the offers are relevant and personally motivating. Identifying this customer segment offers the greatest margin and revenue upside.
- Offer-unnecessary customers. This target audience responds positively to offers, however, would have purchased anyway without the offers. Although marketers often include these sales results in their campaign payback calculation, these purchases actually represent a margin reduction to the company. Examining customer purchase history and maturation patterns can identify expected product lifecycle sales which can then be removed from promotions. The best strategy here is to only offer new or complimentary products from which the customer has never made a purchase.
- Offer-denied customers. This target segment declines all offers and instead only makes purchases when they have an explicit need. Offering incentives where there is already a predictable need, or the absence of alternatives lowers the company's margin.
- Offer-adverse customers. This target audience not only discards offers, but also reacts negatively and may unsubscribe, complain or even cancel their loyalty membership. These customers are often loyal, but simply want to be left alone, which is why they are sometimes referred to as sleeping dogs.
Next best offers are another predictive model we've created for several clients.
Bayesian algorithms examine customer segments and cohorts such as each customer's existing products and products acquired by like customers. We compare the cohorts to customer segments and metrics such as Customer Lifetime Value (CLV.) For consumer industries we also include household data, social data and append the consumer record with third-party aggregate data such as products procured elsewhere.
Testing and measuring each cohort for its contribution to an offer-acceptance creates a predictable model. You can then insert the most relevant offer into a digital campaign or during the next customer interaction and achieve a predictable response rate.
The concept is similar to Amazon and its recommendation of books and other products based on people that purchased similar books or demonstrated similar interests or behaviors. This technique also increases the bottom line as up-sell products are typically higher margin sales. Recommendation engines also constantly refine their algorithms in real-time based on each customer interaction which increases the frequency of offers which convert and decreases those that don't.
Over time the technology develops predictive conversion confidence levels based largely on the combinations of customer profiles and product attributes. This sophistication makes up-sell and cross-sell predictable based on forecasted customer traffic and types of interactions.
We typically categorize predictive analytics to identify options and go-to models for different business objectives.
There's no shortage of predictive analytics use cases, so prioritization is needed. The fastest route to identify the highest impact predictive analytics is to perform a Design Thinking workshop with a multi-functional team.