We are worried about commerce. Things are not going as well as they could ─ and should ─ with digitalization. Even though the industry would be predestined to use data-based machine learning models to… But first things first.
For the EHI study “Technology Trends in Retail 2021”, AI is by far (63 percent) the “most important technological development of the next three years”. Though it’s funny that “analytics”, with only 26 percent of the mentions in the list finds its place among the “farthest-flung”, but on the other hand is in the lead among the “most important projects” with 41 percent.
To say the least, this gives an inconsistent picture of how widespread the topic of data use with AI is in retail. And it also begs the question, what is retail doing with AI, if not analytics?
Part of the effort, and that’s only part of the answer, is apparently going into the (further) development of so-called smart stores, which use artificial intelligence to organize the filling of shelves and self-checkout to minimize the use of personnel.
At least the industry is building a bridge to AI-supported analytics, and dynamic pricing is playing an increasingly important role here, something that has so far been an issue primarily in the fully digitalized online retail sector. In stores, which have hardly been technically equipped for this, electronic price labels are being used across the board, and according to an EHI study, 79 percent of grocery stores already use them. But without analytics, a digital label remains just a digital label.
What can AI do?
Artificial intelligence, or, as the experts prefer to call it, Machine Learning (ML), primarily helps with processing extensive and real-time data. Algorithms trained and modeled on data via machine learning can classify, sort and search data for anomalies and deviations to derive information. This information can support decisions or directly trigger actions along the entire value chain.
If this is too abstract for you, let’s try an illustrative example: The Internet tells the beautiful story (the online medium One-to-One calls it a “modern legend”, but we don’t know exactly) of a retailer who once took a close look at the contents of his customers’ shopping carts, adding the time of purchase and demographic data.
The retailer came to the astonishing conclusion that, for example, men between the ages of 30 and 40 preferred to buy beer and diapers in a bundle on Saturday. It probably had something to do with their weekend plans, but the Internet doesn’t know anything about that either. However, the shopping cart analysis led to the fact that the retailer simply put the beer in the baby section on weekends and achieved a notable increase in sales.
True or not, the story is a good illustration of what data-driven analytics can be useful for in retail. The fact that it is still one of the best stories about the use of artificial intelligence in retail is not something we can blame on AI, nor on the amount of data available ─ it is the responsibility of retailers.
The technical side of AI
Those who want to use AI to first understand their customers better, and then serve them better and do better business overall, first and foremost need a digital mindset, the willingness of transforming what is into what will be by digital means.
This results in a bunch of tasks, which are discussed in this trend report with alternating initial letters: The organization of e- and o- as well as q- and t-commerce makes sense and is necessary, but all these variants of a commerce encompassing all channels depend on the digitalization of the value chain as a “conditio sine qua non.”
By the way, the technology-driven part of AI-supported Big Data analyses has long been done: Today probably every corner store already has a digital checkout system that at least processes all transactions digitally and ─ even more important in this context ─ preserves them in the form of digital data. Larger retailers have larger amounts of data, even today: master data, product data, and transaction data are often already available, but they are also often stored in silos that are not connected to each other. So, the data helps in single parts of the value chain, but not in comprehensive and cross-process data analysis.
Where to start?
We didn’t choose the example of data analytics connecting beer and diapers randomly, we chose it because it shows the way retailers of all sizes can get started with AI and analytics. Basically, shopping cart analytics based on existing data is open to any retailer.
A shopping cart analysis is nothing more than the analysis of existing data about customers and their shopping behavior, as well as other data such as age and gender, and perhaps external factors such as date, weather, or events. A store that also sells via the Internet or one of the major retail platforms has additional data at its disposal: How much time do customers spend in the online store, which route do they take, where do they possibly get out before buying something, and where do they go from there?
All this data helps to critically evaluate your own offering and identify potential for improvement, which could be optimized shopping routes, seasonally offered products, bundles of frequently requested goods. But they can also be services that increase the value of a product: Training offers, insurances, maintenance, spare parts… The sky is the limit, at least in terms of data and the technologies used to process it.
Don’t take the second step before the first one
By the way, we made a small faux pas at this point and went into practical implementation via the second step. We often see this behavior among retailers as well, who would really like to skip the first step.
Before starting the shopping cart analysis, the gods of AI want you to sweat the sweat of data work. If the data is insufficient or simply too bad, then any analysis will fail or provide incorrect results and recommendations for action. Consolidating data, checking it for completeness and consistency, must come first, because without complete and qualitatively consistent data, analysis is impossible.
But then the doors will be spread wide open to retailers: They can use their own and perhaps external data to do more than just shopping cart analyses, which are just a quick door opener. They can use it to buy and stock products on demand, which can very quickly lead to cost reductions and less capital tied up. They can move into dynamic pricing because they will know exactly which customers prefer to buy which products and when. They can move into predictive forecasting because, in the best case, the data can tell them exactly which products will be needed in which quantities and when in advance. And they can use cross-selling offers to ensure that their customers have even better options for their purchases. Again, the sky’s the limit when it comes to imagination.
What now, retailers?
Is it really that simple? Some may rub their eyes in wonder. But yes, it is that simple: just get started! The technology for machine learning algorithms is available, the IT infrastructure from the cloud does not pose any major challenges, the data can be consolidated in a manageable period of time ─ truth spoken: you might need a little help. And then get started ─ ask your very personal questions to your data, the ones you always wanted to ask, but never dared to ask. We can help you with each of these steps, except the first one: getting started with digital thinking, that’s a step you’ll have to take on your own. What are you waiting for?