Real-Time Customer Value Management – Sales and Churn Triggers

Traditional monthly churn prediction or sales potential models help companies identify only some of the opportunities they can take action on to grow and retain their customer base; real-time sales and churn triggers help identify the rest…


Customer value management-related data mining models have been around for decades now, used by companies the world over to help them acquire, grow, and retain customers. Over time, these models have been refined, enhanced, and fine-tuned, now able to much more accurately predict which customer may close his or her accounts, which holds financial potential, which may be good for using for referrals, etc., than ever before. What these models are not so good at, still, is around being able to understand the human nature. The fact that people can react very fast under certain circumstances (such as disappointment with customer service, a good offer from a competitor, etc) makes common predictive models imperfect. In order to respond to opportunities and risks due to this reactive nature of customers, companies need to analyze and utilize events that can trigger sales or churn within a time span ranging from days to seconds.


Although people are assumed to be rational and utilitarian in consumer behavior theory, in reality, they hardly are. Incorrectly overbill a customer for the second time and not resolve it immediately with a refund and an apology, or have a customer service representative snap at a customer when he or she has called in with a complaint, the customer is likely to immediately close his or her accounts (regardless of how excellent the value proposition in place may be). No preceding decrease in spending, no other warning sign, just an unpredictable churn event.

The same goes for many sales opportunities. If a customer is not pitched when they are in the “zone,” then the opportunity may be gone forever. The “zone” could be during an on-boarding process when they are acclimating themselves to the company and are open to new offers, or when they are in high spirits because of a positive contact center experience with the company. Not that sales cannot occur outside of these windows, but the probability decreases, leading to a lower return on investment around sales efforts.

What this means for companies that have monthly (rather than real-time) cross-sales, up-sales, and churn prediction models in place is that they are too late to capitalize on the opportunities that present themselves for ever so brief a moment. For many customers, there is a window of time in which they are open to an offer (for selling more to them, or for retaining them) – miss it and the opportunity is gone for good. Such opportunities are preceded by events (like complaints), and are the ideal time for taking action – a recent study found that event-driven actions generate four to five times higher response rates compared to traditional marketing activities.

Many companies are waking up to this reality, realizing that monthly predictive data mining models are not enough. According to a recent survey by Micros, 76% of companies have already implemented or plan to implement real-time analytics in the near future. Buying some new hardware and software is not enough, however, to tap into the value lying within real-time analytics. Doing so requires companies conduct a comprehensive triggers analysis.


From business discovery to taking real-time actions, managing opportunity and risk triggers follows a 5-step approach:

1. Hypothesize Potential Triggers: The first step in trigger analysis is, as in all analytics modeling activities, business discovery. The main objective of this phase is to enlist all potential triggers that could result in an increased likelihood of customer churn or offer uptake. Companies should involve employees from touch-points during this process, as they experience customer reactions to various events daily.


2. Compile Trigger Data: Once the potential triggers are listed, the next step is to collect and prepare related data, as well as churn and sales information on each customer. As the list of potential triggers is defined based on an “ideal world,” not all triggers may be identifiable with the existing systems and data sources. In such cases, based on the anticipated criticality of these triggers, IT changes may be required.


3. Analyze Trigger Impact: Compiled trigger data and business hypotheses on their potential impact need next be analyzed, based on former customer reactions to such triggers, as well as acceptance rates in test scenarios. From the above example, a telecom operator would need to analyze the impact of certain triggers on each customer segment based on the percent of customers cancelling their line afterwards (after the world “cancel” has been said during a complaint,” whereas the bank would need to test credit card sales campaigns targeting customers with hypothesized triggers to identify their impact on sales. The impact of multiple trigger-related incidents happening to a customer should be examined here as well, as there could be an exponential increase in sales or churn being triggered – an example from the banking sector:


4. Define Trigger Actions: For each trigger or trigger set identified as critical, the next step is to define corrective or proactive actions (such as promotion of the relevant offer, courtesy action to resolve dissatisfaction, etc.). For these actions, the impact of timing should be also studied, in order to maximize acceptance rates (i.e. sending an SMS within 5 minutes after a critical complaint versus 15 minutes later when the customer might have cooled off).

5. Automate Event Management: Unlike monthly campaign plans, trigger based actions require continuous and rapid actions be taken on the identified customers, necessitating a real-time marketing system integrated with operational and communication systems be deployed. Manual management of such efforts is simply not practical – an illustration of how such a system would work:



Pilot efforts are the natural next step once the above steps have been completed. With successful pilots conducted, a full-scale ramp up would come next, allowing for significant benefits to be obtained in the immediate short-term (be it through an increase in sales or a decrease in assets lost). These five steps should, however, be viewed as a cyclical process; companies need to monitor performance of these triggers and actions, and repeat this cycle to refresh them regularly.

Tags : analytics, business intelligence, churn, consulting, cross-sales, customer analytics, data mining, decision making, decision management, fact-based, intelligence, marketing, marketing campaigns, real-time, real-time marketing, real-time triggers, retention, sales, value management