Understanding Behaviours with Machine Learning

Understanding the behaviour of your customers and staff will help you create value and growth, but how do you access this important information?

In our previous article that you can read here, we discussed topic identification and how this is important in providing context to the conversations your business has with customers.  Behaviour identification is the next step in this process because it tells you what’s happening in relation to a specific topic.  First you need to know what is being discussed, then you can identify what behaviours or actions are occurring.  This information will demonstrate how the behaviours and actions are being carried out and lead to understanding what patterns are exhibited in relation to particular business activities.

 

So how do you go about identifying this information?  Behaviours or actions are identified by taking a particular topic, and then adding ‘wake’ words or terms to be recognised by a system.  For example, the ‘wake’ word for the Amazon Echo is Alexa which consumers need to say out loud before stating a demand or asking a question.  One of our clients asked us to identify certain behaviours carried out by their staff including when they were giving advice.  The action of giving advice is a behaviour so we constructed a list of ‘wake’ words and terms to ‘listen’ for.  Keep in mind that when creating ‘wake’ terms you should be careful to avoid making the terms too long or complex so as to avoid missing the core behaviour that is of interest.  These lists are specific to a behaviour or company and can be trained using ML (machine learning), to remove irrelevant terms and add new ones.

In the area of giving advice, key wake terms may include ‘I think’, ‘I believe’ and ‘I suggest’ which would then be analysed against the topic of a conversation.  Once you have a list of wake words or terms compiled, a system can flag these with appropriate actions depending on what has occurred in the conversations.

Here is an example of the topics and behaviours occurring in a conversation between a sales person and their client:

  1. A sales person calls their client
  2. During the call both football and business are being discussed
  3. The client is a fan of the Manchester United team, for their sins, and watched the last game played
  4. The sales person suggests the next game that the team plays will be a win for them and that football is great for betting
  5. The conversation then moves on to the client indicating an interest in purchasing shares
  6. The sales person suggests client invests in Vodafone shares

As you can clearly see, the two topics occurring within the conversation are football and purchasing shares.  You can also see that the context of the conversation is critical when applying these ‘wake’ terms, which is why topic identification is still the vital first step.  The wake term ‘I suggest’ has hugely different meaning when applied to purchasing shares vs suggesting football is great for betting.

The behaviours identified include giving advice in both topics, but the football instance is a non-business incident and is therefore not relevant from a business perspective.  On the other hand, the second incident of advice is business-related and could result in non-compliance to certain regulations if no action is taken, therefore of business interest.

ML is then used to flag this incident to the appropriate people with suggestions on the actions that should be taken.  For example, the system identifies that this is the third incident where that sales person has given advice in the last 6 weeks and actions previously taken included a warning then re-training.  This means the business can more effectively flag for advice and monitor the actions that are taken afterwards to ensure they remain compliant.

Another beneficial example of behaviour identification would be setting up a system for clients to register an interest in a certain product or service.  So the behaviour is “I would be interested in purchasing 1 million GBP of Russian Ruble at 82 or less around April”.  You would then use ML to monitor the rates, dates and identify who would be interested in that information combination, enabling the setting of flags so if the scenario occurs then the client details would be passed onto the relevant sales person for follow up.

 

So is this everything you need to know to effectively analyse your communications? A simple answer might be yes, but to realise the full benefits there’s a third aspect of identification that you should understand called Sentiment Analysis.  Keep your eyes peeled because we’ll be discussing this in our next article!

 

Synetec is an Agile solutions provider with expertise in diverse development technologies, such as Angular, the .Net Framework, SQL Server and other cloud friendly data stores. We are certified and have successfully delivered projects across different cloud technology stacks such as Microsoft Azure and AWS, delivering integration and development solutions since 2000.

We work with a number of the UK’s most respected financial institutions to deliver a range of innovative solutions. We have expertise in working with both established businesses as well as start-ups and extreme growth businesses.

Leave a Reply

RECENT POSTS

RECENT JOBS

ADDRESS

509 The Print Rooms
164-180 Union Street
London, SE1 0LH
Phone: 0208 1444 206
Website: synetec.co.uk
Email: info@synetec.co.uk

DISCLAIMER

Important: The information contained in this website is for general information purposes only. Any reliance you place on such information is therefore strictly at your own risk. Synetec Ltd endeavour to keep it up to date and correct.
All images are copyrighted to their respective owners.
Bitnami