2 minute read

How well do you (need to) know your models? – A DataIQ member report

Name: 2025 Data & AI Awards – Entry Deadline
Start: 2025-01-10
End: 2025-01-10
Location: Nashville, USA

ML and AI tools have exploded in use for data organisations, but what more can be done to improve quality, compliance and efficiency with the models? DataIQ member Robert Bates, head of decision sciences, Currys, provides expert insight in this report.

DataIQ member Robert Bates, head of decision sciences, Currys, has created a report that delves into the ways in which machine learning (ML) and artificial intelligence (AI) models are being utilised and what more needs to be done to ensure quality, compliance and efficient use with these new tools and technologies. His expert understanding of the topic and extensive industry experience provide unparalleled insights into how data professions need to examine their AI and ML model approaches in an age of exponential technological expansion.

Over the past five years there has been a massive growth within retailers looking to use ML and AI to improve performance, simplify the customer journey and increase innovation. But should we always trust the algorithms, and are they even required? Is AI really the next level of retail optimisation, or simply another solution still searching for the problem it is meant to fix?

As the UK opened up after Covid-19 I found myself once again having to defend my preference for explainable machine learning over artificial intelligence within retail. It was a strange conversation (they always are) with a sales rep for a supplier of AI tools which would, they assured me (they always do), increase my speed to production of forecasting and predictive models more accurate than I’d be able to create myself. I asked the usual questions (how does it work? What checks are in place? Is the output explainable?) and was a bit surprised by the response – “Why do you need to know how the models work or what drives them? If they work, they work, don’t they?”

I don’t know what concerned me more – the blind faith in processes without any real understanding of how the business I work in operated, or the matter-of-fact way it was presented; resonating with many pitches I have sat through and how this has permeated through many non-analytical (and even analytical) business functions.

Despite their apparent sophistication, ML and AI are not the solution to every problem on their own. They are part of a set of tools and processes which can be used in partnership with the business operators to identify drivers of performance and predict what may, or may not, happen within a given set of circumstances. Sometimes it is fine to trust the models and output 100%, at other times they simply provide guidance (a range of certainty) where you can apply judgement and decision making on top.

Click here to download the full report for the finer details on AI and ML modelling presented by DataIQ member Robert Bates and discover how it can be implemented within your own organisation.