Successful mobile marketing is all about 1:1 customer engagement—delivering a relevant message to an individual customer in the right context—with the goal of delighting the customer. But how do you determine the best way to engage with each customer—especially when you have thousands or even millions of customers? Is it possible to understand what’s “right” for each individual when behaviors, such as shopping patterns, location, purchases, activity, usage, and social interactions are constantly changing?
This requires a shift in marketing techniques and technologies. Customers aren’t satisfied with what’s deemed “good” for the majority—they want what’s great for them.
The challenge of delivering a personalized experience has ignited a search for marketing technologies that involve complex data capture, sophisticated analysis, and real-time decisioning to determine the best context for engaging each customer and the most relevant message—whether it be promotional, educational, informative, etc.
This rings especially true among mobile operators. In-base marketing teams are charged with retaining customers and restoring revenues. But with millions of customers and limited resources, how can they possibly analyze the massive amounts of data to determine what’s right for each customer, while also doing so in a way that’s manageable, scalable and accurate?
This is where machine learning comes into play. Machine learning has the ability to automatically improve with experience. Rather than being programmed by their users to solve a specific task or problem, computers come up with their own programs that become more intelligent over time as a result of the input that they are fed.
With this technology, operators are removing the constrains of A/B testing and therefore no longer need to wait for days or weeks to determine what’s working and what’s not. Machine learning allows brands to run every iteration and do all of the complex analysis to sort out what works best for whom, when and where and automatically identify which factors drive a certain behavior or outcome, thus eliminating the guesswork. As a result, marketers have gained the ability to test an infinite number of combinations of offers and contexts, get quick learnings, and then, based on the insights, continually iterate and optimize.
So, how is machine learning different from a traditional knowledge-based approach? It’s simple. The knowledge-based approach relies on manually defined rules, while machine learning does not. This introduces a risk of being influenced by either a lack of knowledge, by a bias, or both. Often the rules are complex and have multiple conditions—for example, if a mobile subscriber has just placed a call and is a pre-paid customer with a balance that is under $2.50 and a tenure that is over six months, then send message A with a recharge offer of B.
There are a number of problems with this approach:
1. The marketer must develop the rules and conditions upfront. (In the example above, should tenure be included or not?)
2. The marketer creates messages with no way to determine the efficacy of each message by customer (Should the customer get Message A or Message B, C, or D?)
3. When evaluating results, the marketer can’t sort out which rules and conditions are helpful and which are hurting the campaign (If the marketer removes the tenure condition, will the results improve, weaken, or not change?)
4. It largely assumes that all customers who trigger an event are the same and doesn’t take into consideration rich behavioral customer profiles (and, to the extent it does, it does so in the inexact rules described above.)
Machine learning techniques do not need predefined rules. Instead they derive rules themselves by analyzing the streams of data. As a result, and in contrast to a knowledge-based approach, machine learning does not suffer from bias. Using machine learning is like having a “blank” intelligence that through learning automatically gets smart over time.
Consider all of the reasons behind a certain behavior or outcome for an individual customer. What machine learning does is automatically identify which factors among hundreds of behavioral and demographic attributes drive that behavior or outcome. It virtually eliminates the manual nature of analyzing campaign performance.
With the help of machine learning and automation, have mobile marketers transitioned from manning the clutch of a five-speed stick-shift to enjoying the ride on autopilot? Not entirely. But they are getting better results. By using technology to target relevant offers in the right context to each customer, mobile operators, for example, are seeing a more than 10% improvement in customer revenues and retention. There are no theories or rules or intuition to credit for helping to delight the customer—instead modern computer programs getting smarter and better with time.
Dr. Olly Downs is SVP of data sciences at Globys, a Big Data analytics company that specializes in contextual marketing for mobile operators.
