The Case for AI Analytics in Retail

AI video analytics have traditionally been used for well-known security applications such as shoplifting and violence detection but now retail analytics open opportunities beyond security. Retail AI video analytics is now being used to complement traditional data such as sales metrics to add new dimensions to retail analytics and give a more complete view of customer behavior. These valuable insights enable retailers not only to increase customer satisfaction but also grow revenue and profits.

AI video analytics – overview

AI video analytics rely on computer vision and machine learning algorithms to extract meaning from video camera feeds. Advanced analytics do not just analyze static frames but instead identify movement through a space over time. Video analytics can detect objects and humans and track individuals through an area. In security applications, this is used to detect criminal behavior or weapons, whilst in retail analytics, the same approach is used to give insights into how customers interact with a retail space.

Retail problem definition

Customers can purchase either online or in person at a retail store. Customers buy in person for one of three reasons:

● Necessity - the customer can’t find the article elsewhere or they may want it immediately.

● Convenience - it may paradoxically be easier to purchase in person than online, or they may want to see the item in person.

● Experience - shopping becomes a form of entertainment due to the retail experience, such as special entertainment, promotions, the store layout or service.

Retail analytics must support these three drivers by removing any friction from the customer’s buying experience, whilst improving customer service. When friction is removed and the customer is happy about their buying experience, they are more likely to spend money with the retailer.

Retailers can measure performance with:

● net promoter store – promoters of the store vs detractors of the store;

● retail sales per square meter – a measure of efficiency;

● numbers of items purchased by each customer;

● dollar value of items purchased by each customer.

Retail analytics

As discussed previously, AI video analytics can identify movement through a space. This can be used in a retail context to understand how customers and staff move through the store.

Retail video analytics solutions can distinguish between humans and background objects, track movement through an area, count occupancy within specified areas and measure dwell times. In a retail environment, this can answer these questions:

● What path do people take through a store?

● How long are they in a store?

● How many people enter the store?

● How many people leave without buying?

● What areas of the store are they most interested in?

● How long do they wait to be served?

● What areas of the store cause congestion?

Simple observations, such as informal reports or sales and stock figures, do give some insights into retail performance. Adding behavioral data, such as pathway analysis and dwell time, gives a far richer understanding of customer behavior. This richer data also allows retailers to identify trends to give an early indication of customer responses.

Benefits of retail analytics

Retailers can benefit from retail video analytics in several ways.

Store layout optimization

Store layout optimization means developing the store layout that makes it the easiest for customers to find and buy what they want, and to move easily through the store. Reductions in frustrations result in increased likelihood to buy because it improves the customer convenience.

Another factor is the “street appeal” of the store, which is a measure of how likely passersby are to walk in. This can be measured and optimized to improve the customer’s shopping experience. For example, how does a pilot store with a different layout perform against the current design in terms of street appeal?

The metrics needed to help optimize the store layout are:

● Line crossing - to learn the number of people entering the store;

● Pathway analysis - to understand the directions taken by visitors once they have entered;

● Heatmaps - to identify areas with more traffic or choke points.

Staff resource planning and training

Staff resource planning and training means making sure that there are the right mix of people to provide customer service and that they are properly trained to meet customer service needs.

Reduced checkout time is strongly correlated with reductions in potential customers simply abandoning their purchase and not returning.

The appropriate metrics to plan staff resourcing and identify training needs are:

● Number of customers in the store by time, which indicates staffing requirements through the day;

● Dwell time, which shows how long it takes for customers to check out with their purchases (a long time may require more checkout staff or better training);

● Heatmaps may indicate a need to increase product training to help with enquiries, for example, when customers spent more time near certain products.

Measurement of the effectiveness of campaigns

It’s one thing to have a campaign that increases sales, but how can you get more insights into the second order impacts of the campaign, such as customers attracted by the campaign who don’t buy, or impact on staffing? Analytics such as Scylla allow you to dig deeper by understanding:

● how many customers enter, and leave without buying;

● what customers spend the most time looking at;

● how long they spend in the store.