Skip to content
All posts

Demystifying machine learning


When you hear the term machine learning, there is a misconception that only large businesses with multi-million pound budgets have access to these methods to increase their business performance. This is not the case.

The way I like to explain machine learning to clients and colleagues, is that these processes do the heavy lifting that was once required manually. Such techniques can also spot patterns and likely outcomes in data that the human eye or brain simply cannot.

Machine learning is a huge field of expertise with extensive applications.

Taking a definition of machine learning we like from DeepAI, “Machine learning is a field of computer science that aims to teach computers how to learn and act without being explicitly programmed. More specifically, machine learning is an approach to data analysis that involves building and adapting models, which allow programs to "learn" through experience. Machine learning involves the construction of algorithms that adapt their models to improve their ability to make predictions.”

Let me explain this another way, in the content of TwentyCi & our data capability:

If I was to look back at how we might select data, say four years ago, without machine learning methods, we would have applied a layered data selection which would have looked something like this:

  • Audience type (Homemovers or Buyhaviour)
  • Mosaic profile
  • Income
  • Lifestage
  • Number of children
  • Property price
  • UK locations
  • Disposable income
  • Number of bedrooms
  • And so on...

The Mosaic, income, property and location details would have been based on a customer profile we would have either received from the client, or from a profile we created from our client's customer data. We would then take this information and build a data selection for a clients' marketing campaign in our campaign management software.

The issue with this approach is that it assumes 'one size fits all'. It’s highly likely that from within the audience not selected, there are pockets of audiences which would perform well in the campaign. And there are most likely parts of the selected audience you would be better not targeting - this is where machine learning comes in, to do this automatically for you!

Jumping forward to the current time, and our data selection using machine learning techniques now looks something like this:

  • Audience type (Homemovers or Buyhaviour)
  • Client model (Created from previous transactional & performance data)
  • Any client specific hard exclusions (for example locations where a business doesn’t service)

From here, we use campaign management software to score all possible prospects on their likelihood to transact, selecting the top number of records for the campaign.

The key difference with machine learning is that it very quickly ramps up the marketing programme; the targeting is more granular and the ROIs will be higher without you needing to spend time building individual selections.

Talk to us today about how to increase your marketing performance by using machine learning to maximise return on your marketing spend.



Would you like to find out more?

TwentyCi | | 01908 829300