Netflix, Google and Facebook were all spawned with machine learning built into their DNA. Think about Netflix, and its ability to continually learn about your viewing patterns to then deliver highly personalized recommendations. By keeping current customers engaged while marketing to new ones, Netflix viewership is predicted to surpass ABC, CBS Fox and NBC by 2016.
Companies like these have invested heavily in machine learning technology from the moment they were founded. Vigilantly gathered data is constantly fed into algorithms that drive the delivery of highly targeted content, emails and even phone conversations to consumers. For these companies, machine learning is a strategic asset that's absolutely core to their business.
But what if your company isn't a tech behemoth? What if your company grapples with a legacy data system that's nowhere near capable of uncovering the types of discoveries critical to being able to formulate and target optimal customer experiences? What if machine learning capabilities may be valuable for your company, and strategic, but aren't accessible?
Opening new doors
The good news is that if you weren't born a fish, you can still learn to swim. The same approach can apply to deploying machine learning. For startups and large, established brands alike, machine learning can deliver impressive payoffs, and sooner than some companies think.
So why now is the tremendous value of machine learning capabilities coming to light? Because recently machine learning has been getting to the point of near-human recognition quality on some interesting problems, where humans have typically been best, and at massive scale. We all know that an individual can track a single long-term customer relationship and provide regular feedback to keep that customer happy. But what about doing that for thousands or even millions of customers as their behavior dynamically changes? That's the scale of automated machine learning that's now possible.
Common barriers
Many marketers will say they've tried machine learning and it didn't work. However, it's more likely that all the right places weren't in place. Usually, the company has one - or both - of the following issues:
- The company's data isn't set up in a way that's amenable to applying machine learning techniques. Today, most companies are still focused on data warehouse initiatives where data is aggregated and the richness is in the form of averages and scores. The primary motivation is efficient data storage rather than maximizing the fidelity of the data retained. The issue is that there's then no ability to investigate nuanced customer behavior in order to discover and leverage new insights.
- Machine learning is not integrated into the company's business operations. Too often sophisticated modeling takes place but is not part of a closed loop process. In other words the process for learning is not tightly connected to the process for taking action. As a result, getting business impacting results is very difficult.
What companies can do to grow into machine learning
By leveraging well designed algorithms and the computational power of today, companies of all sizes and types can realize the transformational effect of machine learning - even if they weren't born a machine-learning company.
Take mobile operators, for example. Not exactly an industry that was born using machine learning. What's remarkable is seeing operators dig deep into their data and apply dynamic machine learning to discover patterns in their customer's behavior that would otherwise go undetected. And furthermore, leverage these discoveries to power algorithmic marketing of incomparable relevance.
Operators generate billions of data points per day: customer's voice, sms and data usage, payment transactions, balance information, recharges, rate plan changes, purchases, device migrations, spend patterns, churn propensity, profile, social connectedness, marketing interactions and responses, customer service and support interactions, and more. By being able to analyze these raw data feeds, operators can develop a rich understanding of each individual customer's journey which then fuels the ability to determine optimal customer experiences. The power of the machine learning is in automatically i.e., without human intervention, discovering the highly granular rules for who gets presented what and when.
The key to success for these operators, as seen by the impact on business results, has been the integration of machine learning capabilities into standard operations. Machine learning works best when it is used in a closed-loop fashion. Constant feedback to the algorithms allows continuous refinement of the learnings about currently-optimal marketing.
Growing into your own
Operationalizing machine-learning within existing companies will become easier as technology evolves. Companies don't need to be born as Google or Netflix to take advantage of machine learning.
Today, major industries have been able to deploy big data platforms that are designed from the ground up to support machine learning techniques. As a result, companies are able to test thousands of marketing messages and automatically determine and execute what's best for each customer at the right time. What's exciting is seeing these companies grow into their own - and make use of machine learning to strengthen relationships with consumers and set new benchmarks for business results.
Dr. Olly Downs is the Chief Scientist behind Amplero, a self-learning personalization platform built by Globys.