Machine Learning and Data Analytics
Today many organisations collect data from their online transactions. Server logs. Basket contents. User sessions. Much of this is collected and stored automatically, without a clear rationale as to how it will be used. The thinking being that as long as all data is collected, then one day it can be exploited – and monetized. It’s the promise of future revenue that supports the cost of collection.
But this faith is perhaps misplaced. Without understanding the specific use cases that the data will inform, how can anyone be sure that the right data is being collected? Are the right unique identifiers being collected? The key types of data that will identify something that can then be matched to other data. And is the right complete set of data being collected?
Help is at hand. These days, with the advent of Big Data, this has led to another technology capability: Machine Learning. The concept that powerful computers can process large amounts of data to discover meaningful patterns automatically, and to improve on their ability to do so through learning – or corrective feedback.
With machine learning, computers can now analyse large sets of data, and from the mass of different data types and values can start to derive insights, without having to be told what to look for. What was the key correlation between several million users of a particular system that drove them to X? Is it, as the traditional persona-based approach would suggest, because they are from a particular socio-economic group, of a certain age and sex? The correlation identified through machine learning could be very different. It could be that the common attribute shared by the majority of people rich and poor, young and old, male and female, is that they all enjoy 80s music. Sounds too far-fetched? Perhaps so, but data and technology are conspiring to discern insights that were previously not possible to anticipate. By creating clusters of common attributes, derived from large sets of data that are constantly aligned to changing real-world behaviours, machine learning is creating new types of auto-personas that have a lot more relevance in the digital economy.
However this in itself is of no value if it is just treated as another set of data to be analysed and reported on. The critical link to break – or make – is to take these data clusters and apply them in real time to the systems of engagement that they came from. If cluster A describes the characteristics of people most likely to want X, and another person of type A accesses the system, then personalise the experience for person A to serve them X. It seems so obvious, but requires a paradigm shift from marketers preparing data and reports and business acting on these in next Monday’s review, to an almost immediate and automatic system re-alignment that will enable people of type A to get to X as fast as possible.
Data Analytics can mean different things to different people in the digital economy. However by leveraging machine learning to operationalise data analytics, businesses can gain competitive advantage through increased conversions.
