Being able to accurately predict the future is understandably fraught with difficulty: the Danish physicist Niels Bohr arguably summed it up perfectly when he said, “predictions are hazardous, especially about the future”. However, as we look forward to what the next 12 months have in store, one thing looks set to be a safe bet: that 2020 will be the year that puts the ‘AI’ into retail.
It’s no secret that it’s an exciting time for Artificial Intelligence (AI). Advancements in technology are providing increasingly powerful machines, removing many of the computational barriers that were present up until even just a couple of years ago. As with any aspect of tech, this trend of constant development and improvement will surely continue in the coming year.
There’s already been an increase in the number of tech companies offering AI-based solutions, and one would expect to see progressively more of these alongside a greater willingness to experiment with these new solutions on retailers’ part. The result will be a perfect scenario for AI to flourish. More specifically, it may well usher in a wiser use of AI in e-commerce and encourage more widespread adoption. This is especially true as competition for consumers’ attention and purchasing power shows no signs of easing off.
When talking about an increased prevalence of AI in retail, the great enabler in this regard will undoubtedly be the adoption of graph database technology. Every decade has had its database technology and 2020 will mark the definitive start of the graph database era. The reason is that graphs unlock the immense potential AI and machine learning hold because the technology incorporates all the context and connections needed to make AI more broadly applicable.
🏆
The 2024 Creative Retail Awards are open for entries.
The Creative Retail Awards are much more than a mere accolade; they represent the pinnacle of achievement in the retail industry. Garnering a nomination or winning one of these awards is a testament to innovation, excellence, and leadership.
www.creativeretailawards.com
For e-commerce, the surge of graph database technologies will enable the use of more AI, and in particular, to more real-time predictions. This not only pertains to what a user may like but also about what a user’s intentions are, whether browsing or buying, with purchase intent prediction. Expect to see an increasing number of pioneering brands deploying solutions to know whether a customer is buying or just browsing on an e-commerce application.
This prediction will be obtained in real time by analysing customers’ actions and behaviours on e-commerce web sites. For example, it may be noticed that some customers browse the product comments for a long time before they place an order. By acknowledging this and other corroborating patterns, businesses will be able to better understand users’ actions, intents – and ultimately – to offer better services to their customers.
When it comes to email, the benefit of AI will be even more apparent with the widespread introduction of real-time product recommendations. To date, the majority of solutions currently available for email provide product recommendations at the time of send. In other words, the products shown to each recipient are determined before the email campaign gets sent.
However, starting next year you’ll see these recommendations will be generated ‘on-the-fly’ at the time of opening, when recipients read their emails. This development is crucial, because real time generation is vital to improve the quality and effectiveness of product recommendations. Graph-based AI systems will be able to include more variables (the context and connections mentioned earlier) including factors such as stock availability, latest price changes, promotions, the time of the day, users’ approximate locations as well as even the weather conditions in that area. By incorporating all these extra conditions into the recommendation, they can be tailored even more precisely to each user and move one step closer to the ever-elusive level of on-to-one personalization everyone is aiming for.
The dramatic success in machine learning has led to a torrent of AI applications. AI systems are taking over a vast array of tasks that previously depended on human expertise and judgment. Often, however, the “reasoning” behind how these systems make decisions is not clear and can cause a lack of trust. As AI is applied more broadly, it will be crucial to understand how it reaches its conclusions. XAI—especially explainable machine learning—will be essential to understand, appropriately trust, and effectively manage an emerging generation of artificially intelligent “machines”. Especially in fields such as health-care or finance, where mistakes can have severe consequences, the black box aspect of AI makes it difficult to trust. Furthermore, data protection regulations that have recently entered into force emphasize the “principle of transparency” of intelligent algorithms and imply the “right to explanation” of algorithmic decisions.
It should be noted that some (perhaps naive) explainable AI has long been familiar in e-commerce as part of online recommender systems. Some types of recommender systems adopt intuitive yet easily explainable models to generate recommendations, such as memory-based collaborative filtering, which provides recommendations based on ‘similar users’ or items, with Amazon perhaps being the most famous example.
However, as the machine learning algorithm adopted to making predictions increases its accuracy, the explainability of the result often becomes harder. State-of-the-art recommendation and search models extensively rely on complex machine learning and latent factor models (such as matrix factorization or even deep neural networks), and they work with various types of information sources. The complex nature of these state-of-the-art models makes search and recommender systems black-boxes for end-users, and the lack of explainability arguably weakens the persuasiveness and trustworthiness of the system. As such, with the growth in availability and utilisation of AI and machine learning in e-commerce and retail marketing, there will be a parallel development of XAI to complement it.
We’ve all seen how AI systems have steadily taken over responsibility for a vast array of tasks previously dependent on human expertise and judgment, and online retail will be no different. But the difference to some other sectors where AI has been used is that in a retail context you have to always factor in the end user (or customer) and their understanding of how and why they are being recommended products. However, with better XAI and more AI on the horizon in general, it’s likely 2020 really will be the year that puts the AI in retail – and likely for good.
By Gabri Corti, CPO and Ricky Saccomandi, CTO for Kickdynamic, a pioneer in email marketing trusted by over 200 of the world’s leading retail, fashion and travel brands. Their open time technology and its suite of email enhancement tools are used to help marketers move from sending manual, static campaigns to highly personalized live email, at scale.
