What Is Retail Analytics?

Retail analytics is the practice of collecting data across a retail business and analysing it to decide what to stock, where to put it, how to price and promote it and how many staff to roster. It draws on sales transactions, loyalty schemes, stock systems and online behaviour, and in physical stores on what cameras and sensors see on the shop floor.

Online shops have always known which pages a visitor viewed, what they added to the basket and where they gave up. In-store analytics gives physical shops a similar view: how many people came in, which aisles they walked, how long they waited at the tills and which shelves ran empty. Combined with point-of-sale data, it turns gut feel about the shop floor into evidence.

Analysts usually split retail analytics into four types, each answering a harder question than the last.

Most retailers start with descriptive reporting. The value grows as live in-store data feeds the diagnostic and prescriptive stages, because staff can act while the customer is still in the shop.

Each source adds a different piece of the picture. Few stores use all of them, and the cheapest wins usually come from combining footfall with sales.

Retail analytics produces dozens of numbers, but a handful drive most decisions:

• Conversion rate: transactions divided by visitors. It needs footfall data, which is why people counting is usually the first in-store system fitted.
• Average transaction value: takings divided by transactions, tracked against promotions and layout changes.
• Dwell time and zone visits: how long shoppers spend in each area and which displays attract them.
• Queue length and wait time: the trigger for opening tills or redeploying staff.
• On-shelf availability: the share of the range actually on the shelf, not just in the stock system.
• Staff-to-shopper ratio: rostered staff against live occupancy, hour by hour.

Store camera analytics usually runs on a small computer in the shop rather than in the cloud. Streaming every camera off site needs bandwidth most stores do not have, and a queue alert is only useful if it arrives within seconds. Processing video on site also helps with privacy: the computer turns video into anonymous counts, paths and alerts, and only those numbers leave the store. Our edge AI explainer covers the approach in general.

The retail edge AI gateway case study shows a deployment in practice. Where a store already has screens, our digital signage players can double as the in-store edge computer, running content and analytics on one system. Our guide to PoE power budgeting for multi-camera edge AI helps size the network for the cameras.

In the UK and EU, camera analytics falls under UK GDPR and the EU GDPR, and the Information Commissioner’s Office expects a data protection impact assessment for most systems that monitor people. In practice, retail analytics is usually designed to avoid personal data: counts, heat maps and queue lengths do not need to know who anyone is. Processing at the edge, keeping only aggregate results and showing clear signage all help. Facial recognition is a different matter and carries much stricter legal requirements.

Retention matters too. Where recorded video is kept at all, set a retention period that matches the purpose, limit who can view it and document both in the impact assessment.

We build the in-store edge computers as standard: signage players and edge AI systems supplied with the operating system, drivers and AI runtime installed and tested with your cameras and screens before they ship.