Artificial Intelligence and its adoption is at the point where businesses are either in the beginning stage of implementation, or some departments and individuals in companies are using machine learning or predictive models to optimize their processes. Optimizations can be part of their existing software solution or can be an API plug-in which is available on the market.
We all know that AI is not just hype. But where in Retail can we use it and which organizational department or sector should be the “AI main adopter”? After all, embedding AI throughout the business requires a significant investment in talent and upgrades on existing software, because the use of AI functions should improve the user experience of end consumers with better decision making and faster operations in crucial processes with faster calculations (taking into consideration all the variables). So, the answer to the question where to start implementing AI features in Retail is relatively easy – go with common sense and “just follow the money”.
The business areas that traditionally provide most value to companies tend to be the areas in which AI can have the biggest impact.
In Retail there are two departments which are biggest “money spenders”. The first is Sales & Marketing and the second is Value Chain Operations (how goods arrive from the production line to the end consumer’s home or use).
By using AI in Marketing & Sales with personalized promotions and optimization of customer relationship with basic AI features, Retailers can increase sales by 1-2% with incremental sales. In Marketing & Sales the machine learning and clustering of similar types of customers shows patterns which are usable and can bring incremental sales if they have the proper user experience for end consumers.
Today, the most commonly used AI applications are:
Use of predictive models in sales promotions
Weather is one of the variables that can influence sales. With proper weather projection models, which are combined with past search on the web and sales data, the predictive models can be trained to spot optimal timings for sales promotions in the future.
Projects like EW-Shopp, where AI is used to determine how the weather and events are influencing Customer Journey, are just one example of AI building predictive models on how weather and different events are influencing sales.
What does that mean in real life?
I’ll try to explain it on a practical example: If we take the category of drying machines for clothes. By using AI – predictive models, AI determines that there is one variable that has a higher input on search for dryers on the web and on sales in online and offline sales channels than other variables. This variable is Precipitation, and predictive models may show that the category of dryers is a seasonal business. The season lasts from August till December (seasonality is shown from past sales data). From past data the predictive model calculated that the demand for search queries online and sales in retail was higher when Precipitation was 70% or higher than when Precipitation was below 70%. With this knowledge the optimization of sales promotions can be done by producers, distributers or retailers.
In Slovenia in October 2017, a test was conducted and one retailer using AI prediction models increased the sale of dryers by 260% compared to the weekend before. He made use of the knowledge that the conditions were optimal (Precipitation above 70%) and he launched a sales promotion at the right time. This kind of input can be done with the weather or with different events (concerts, product launch, new releases, seasonality, etc.)
E-mail marketing optimization with personalized offers
Marketing automatization tools (Sales Manago, SharpSpring, Watson Marketing and many others) can use machine learning to analyse the past purchase and search data, use segmentation for different clusters of potential customers, and provide optimized offers for potential clusters of customers. This is all automated and can be done if the architecture of data sets is done correctly.
Personalized webpage with offers and filters which are customer-specific
With the same marketing automatization tools described in e-mail marketing optimization and with the same data sets it’s also possible to customize a webpage for a specific user.
By employing recognition of customer’s last picks and purchases, it’s optimized to offer cross-sell and up-sell promotions which are beneficial to the end consumer. If your last purchase was a Samsung Gear S3 Frontier watch, and you teach the marketing automatization tool to show the same customer 2 months after purchase the leather strap for Gear S3 on the index page, then cross-sell incremental sales is possible. Same goes for the optimization of filters on the category and product page – with machine learning and clustering the same segments of customers the same filters are shown to the same segments of customers. With this optimization the user experience and cleanness of the category and product page is improved.
Chatbots and video bots for optimizing interaction with end consumers
In the case of customer care or customer experience (CX) it’s vital that the response to customers’ wishes, requests and questions happens within a few seconds (24/7). This is costly in the case of internal or external call centres. Here AI kicks in; by using AI-driven chatbots, the first 2-4 questions can be answered by the bot which is segmenting the nature of the customer and problem. With such a solution the response time is much quicker, and 80% of requests can be answered by the bot (most of the questions are generic and can be answered by AI). Where traditional analytics techniques are used, there is the potential of AI optimization in Sales and Marketing.
The second large department in Retail where the functions and benefits of AI use are projected is Value Chain Optimizations.
In Retail, stock optimization and sales projections are the key success factors. By using Artificial Intelligence techniques that fall under the umbrella term “deep learning”, which utilizes multiple layers of artificial neural networks, current analytical processes inside SKU-level forecasting systems can be upgraded. AI can enable forecasting based on the underlying casual drivers of demand rather than prior outcomes, thus improving forecasting accuracy by 10-20%.
Replenishment systems inside a retailer’s value chain can reduce inventory costs by 5% and increase revenue by 2-3%.
While AI applications cover a full range of functional areas, it’s precisely in these two cross-cutting areas in Supply Chain Management and in Sales & Marketing that retailers should start using AI, which can have the biggest impact.
For leaders and managers to calculate the economic potential of using AI techniques they have to take care of some limitations.
- First, the company needs to do the data architecture. Retailers should obtain data sets that are sufficiently large and comprehensive enough to feed deep learning;
- should have expertise in and mastery of human power;
- and should have a vision of where AI techniques will be used first with the MVP (Minimum Viable Product) as a case study, which will also be implemented in the difficult “last mile” integration into regular workflows.
For all of us who are trying to learn how to use AI techniques and where to start using them in retail, the time is now. Try to implement them where you’re using most of the money.