From hoarding toilet paper to shops shutting their doors, it’s fair to say 2020 was one of the most disruptive years for retailers. According to the ONS, online sales grew by 46% in 2020, compared to the previous year. With consumer behaviour changing rapidly in response to the pandemic, retailers have had to adapt and keep pace with demand. This has been the perfect use case for automation, which has proved invaluable as retailers navigate a path through disruption.
The scene was set for automation to save the day, but with familiar trends, patterns and demands disappearing in front of their eyes, the initial reaction for many retailers was: “We need to turn our AI systems off.”
This was an understandable reaction. Within retail, AI and machine learning (ML) are often used to make ordering processes more accurate – eliminating the chances of being over or under stocked. The key to their accuracy often depends on historic data, but given history and ‘do it like last year’ wasn’t applicable anymore, many retailers went with their instincts and concluded that AI and ML were not up to the job during the pandemic. Desperate to bounce back from more than £3 billion in lost sales, many panicked and reverted back to manual decision-making.
Ignoring individuality
At this time, many retailers made ‘broad brush’ decisions and simply slashed prices across the board, which could have cause lasting damage such as eroding both brand equity and profits.
This trend was especially common in fashion retail, with seasonal trends up against the clock and the fear of waste looming.
Across sectors, many retailers making decisions manually failed to take individuality into account. ML, on the other hand, individualises each item and makes recommendations on a range of factors including availability and consumer demand. Reacting with panic and ignoring these individual strands could cause serious long-term damage to business.
Causal reconstruction
The power of AI and ML should never be underestimated. Despite so much uncertainty and forward planning seeming near impossible, when trusted, systems adapted quickly and made better predictions that any human.
This is because AI and ML adaptivity was always intended to embrace changes, fluctuations and anomalies.
It all stems from the systems’ causal reconstruction capabilities, which are made up of three phases:
1. The causal model: This includes analysis of those aforementioned, individual properties of item, location and calendar. It dictates the calculation and prediction of consumer demand.
2. Analysis: Analysing the success of those calculations in their short-term aftermath.
3. Adaptation: The ability to calculate recent deviations (residuals) and then adapt the output appropriately.
This final phase was critical in 2020. In just a few days, the systems had realised that demand variables weren’t the same and that corrections were required. Those corrections occurred far more quickly and were far more accurate than anyone could have anticipated.
AI > human emotions
When shops suddenly ran out of toilet paper, it took humans a few days — if not weeks — to notice there was a purchasing phenomenon. ML had spotted it within two days. The end of the hoarding also passed humans by, but not AI systems – who quickly brought ordering levels back down again. As we enter potential future lockdowns, this technology will beat us to the punch every single time.
As Covid-19 struck, some retailers were worried that AI ‘would be found out’. However, what was really found out is just how adaptive AI can be. Those who tried to override AI manually soon turned back. Those who opted for a hybrid of adding their own procurement decisions on top of the systems’ veered even further away from eventual stock demand levels.
It is human instinct to want to fix problems, and that notion has taken on extra poignancy this year. It is entirely understandable. However, if you really think about it, it made sense to take those emotions, feelings and panics out of the equation. Even though it felt like the world was completely shifting before our eyes, from AI’s point of view, not much had changed.
The immediate demographics of customers were largely unchanged, the seasons still followed one after the other, and some items are non-negotiable as society’s demands around education, health, hygiene and nutrition were not stopped by the pandemic.
Learning to cope with change
2020 and 2021 have undoubtably been full of change, but the only difference from a machine’s perspective is the pace of that change. It is understandable that humans were doubtful of how AL and ML would react given their reliance on data history. However, life and industry are constantly changing, and that is what they are built to recognise. From fashion trends to dietary trends, consumer behaviour changes throughout the years.
2020 came with lots of challenges, lots of changes and lots of uncertainty. However, if it’s taught retailers anything, it’s that AI and ML’s ability to adapt in disruptive circumstances is much better than any humans.
By putting trust in AI, retailers will be able to reap the benefits. Whether that’s by helping to set the right price, manage inventory or streamline processes on the shop floor, AI is proven to make predictions through times of uncertainty.
Photo by Parker Burchfield on Unsplash
Interested in hearing leading global brands discuss subjects like this in person?
Find out more about Digital Marketing World Forum (#DMWF) Europe, London, North America, and Singapore.
Thank you so much for this amazing article