Why is Machine Learning better than Human Analysis?
Glen Westlake – For humans to analyze data they need to simplify it into a few dimensions to analyze a few measures, so unless they know what to analyze and how to compare them, they cannot find anything significant. So they just keep reporting the same information as they have always done, typically reporting historical results, for which it is too late to do anything about.
A machine can quickly analyze and compare thousands of variables across thousands of dimensions until it finds something that correlates to an outcome. So it can consider far more data and find genuine correlations rather than prejudiced feelings about why things happen. This enables the machine to start to forecast and predict future outcomes with a high degree of accuracy.
How can this technology apply to Marketing?
G.W. – Consumer sales and marketing was actually the first commercial area to really benefit from this technology. In this field, social signals and lots of data are continually analyzed by machines to make the most appropriate customer offer. Personalized adverts on the Internet and product recommendations are probably the ones we are most familiar with.
The forward thinking marketers are beginning to embrace new software and data analysis techniques to enable more dynamic, relevant and usable customer segmentation, using all the data available. This can be used at the heart of customer acquisition (lead generation), retention and growth campaigns (upsell/cross-sell) to improve marketing performance. Dynamic customer profiling is something all marketers should get excited about.
Could machines take over human functions?
G.W. – With the previous track record of human performance with data, it is inevitable that machines will take over human jobs in this area. The question now is how long before the default is a machine over a human? My guess, within 5 years machines will become standard and humans will pick up the more emotional and creative functions only.
Fortunately, I don’t believe it is all over for humans quite yet: companies only do things better when they take some action, such as make an offer to a customer or perform a task. What action you take and how it is executed is still very much a human decision, often requiring some creative and emotional intelligence to influence another human, which machines are not currently very good at. Extracting ‘actionable insight’ is what machines are now very good at, especially when fed lots of data, far more data than any human processing could ever contemplate handling.
But the key question is “Can you afford not to let a machine loose on your data and how much value is hidden in your data?”