What do we find interesting & why?

SAIL – Using AI to “find the interesting” in data 

 

By Matt Beckett, Managing Director, CSL

 

We all know that the key to a good suite of dashboards is to ensure it tells a story; enabling the user to follow a chain of thought to support a decision or action. The challenge is, the average suite of dashboards contain so much data they are more akin to a library than a book.  

When we want to find out what’s happening in the world we don’t turn to a verbatim script of everything that happened everywhere yesterday, we look at newspapers or websites that filter out what journalists think we will find interesting.  Why should analytics be any different?  

So just over 18 months ago we set up a research and development team with a brief to “find the interesting” in everyday data; to see how the latest technologies can be used to apply human thought patterns to vast quantities of information. The aim – to produce headlines that can be integrated into existing reporting systems or consumed via a stand-alone mobile app.  It has been a fascinating experience.

 

Most interesting was discovering that the challenge was only partly technical. The bit that required the most oily fish to be consumed was learning what we (humans) find interesting and why, so that the technology could be trained to look for it.  

We decided early on to use our specialist knowledge in healthcare to our advantage and to concentrate on areas that are of interest to the industry.  After nearly two years of development, testing and refinement we are launching SAIL this Autumn with some exiting initial insight engines:  

  • Anomaly detection – SAIL sifts through the latest data and highlights events that are unexpected, using intelligent algorithms to weight those we think are most significant  
  • Predictions for next period – Forecasting is normally only attempted at national or perhaps territory level, as increased noise at lower levels (such as account) mean there is too much variation for regression analysis to work effectively.  In SAIL we have taken a different approach by comparing patterns in the data to similar patterns seen before to gauge the most likely outcomes in the next period.  This allows us to make predictions at the account / product level, and only show the predictions we are confident in 
  • Changes in policy – Does the data indicate a coordinated change of your drug or a competitor’s drug at a CCG or PCN level?  SAIL appraises each organisation for the typical coherence of its actions and then uses that to assess changes in prescribing across the GP practices that make up the organisation.  If the data indicates a change in policy, SAIL will bring it to your attention 
  • Unusual buying patterns – SAIL can keep an eye out for buying groups that are purchasing products in unusual ways, which might indicate parallel exporting for instance.  The data is analysed for the patterns in which stock is purchased as well as assessing the appropriateness of purchases given local demand 

We will continue to develop and add new insight engines over time. We can even write bespoke engines for your particular needs.  We’re excited about moving the analytics story on and would love the opportunity to show you how SAIL could take your existing analytics to the next level.  

If you would like to find out more contact us on info@cls-uk.com or call Dean Williams on 01483 528323