Retail intelligence advances through technology

Advanced technologies, such as image recognition, machine learning and artificial intelligence, are having a significant impact on the retail sector.

Johannesburg, 11 May 2018
Read time 3min 10sec
Andrew Dawson, Commercial Director, Solutions in Hand.
Andrew Dawson, Commercial Director, Solutions in Hand.

In retail, there's an age-old adage: 'Live on shelf at all times'. What it means, is that if your product isn't available for customers to purchase, then you aren't going to make any money. It's that simple. And while the supply chain comes with its own set of challenges, having consistent merchandising and product availability in-store is vital.

Andrew Dawson, commercial director of Solutions in Hand, says: "Unlike Amazon, suppliers to brick-and-mortar retailers are very reliant on human intervention to get their product onto the shelves where they can be sold. Whether it's merchandisers or aisle stockists, the relevant role players are predominantly employed by a third party, so may not necessarily have the same brand loyalty as a direct employee of the supplier."

All too often, the ethos of 'Live on shelf at all times' becomes a bit of a hit-and-miss affair, wholly reliant on the quality of the merchandising force. As a result, some fast-moving consumer goods (FMCG) companies have employed independent audit firms to try and get snapshots of the state of 'shelf health' in various outlets, in an attempt to get an indication of their product availability. It also enables them to penalise the responsible merchandising team for non-performance, should that be necessary.

"Unfortunately," continues Dawson, "a large portion of any shelf audit still requires human intervention and it is possible for the results to be clouded, skewed or even not representative. The good news is that by automating the audit part of the process, we can make it more objective."

He goes on to break down the three core activities that contribute to the automation process:

Image recognition: the objective is to be able to use a mobile phone to take photo of a shelf or cooler and for the application to be able to automatically differentiate between brands and products and their facing quantities without having to physically count the items. This reduces the time burden on the auditor and takes counting and finger trouble out of the equation.

Machine learning: the aim is to detect, segment and classify objects based on images captured. The images form part of a larger dataset against which new images are compared before being categorised. Specific products are classified by the image being compared to a set of images of that product. The model can determine whether a random image contains the required product with a probabilistic measure of accuracy.

Artificial intelligence: image recognition and machine learning processes are able to record datasets of images in a live database environment from multiple retail points. A combination of rules and algorithms that make up the perpetual and accumulative review of these sets of data (AI) can start to identify patterns in stock movement, which can be shared with relevant stakeholders.

This proactive approach ensures that the stakeholders have a finger on the pulse of events at store level and can make informed decisions based on identified trends. With time, trend detection can potentially change into trend prediction. Dawson explains: "Trend prediction will result in a process of refining algorithm development that accelerates auto ordering, correct 'days of cover' and ensure 'live on shelf at all times'. All of these will impact on the cost of availability, which in turn influences profitability."

He concludes by saying: "Automation and intelligent services are here to stay and, if used correctly, can enhance job creation and security while protecting and growing the bottom line."

Login with