For The Battered Retail Industry, Machine Learning Has To Be About Sales
We are ultimately trying to sell more. I think that is the beauty of retail in that the motive is very simple.
More and more, the forecasters of technology and the global economy are saying that the road to economic empowerment will be led by machine learning and artificial intelligence.
Accenture and Frontier Economics predict artificial intelligence will double the economic growth of the world’s 12 biggest economies by 2025. In a different report, Accenture forecasts the unlocking of $2.95 trillion in economic value by retailers and consumer goods companies over the next decade by adopting digital solutions and adapted business models.
For much of the retail industry, this is great news …
If it can stay afloat long enough to make the evolution a reality.
2017 is the year that the crows have come for retail. Almost every week we hear about new layoffs, massive store closing and weak earnings. Years of neglect of backend technology systems, lack of direction in an omnichannel world and pressure from outside forces have cut back the already slim margins of retail to the point where the industry is breaking, cracked open like an overripe melon.
The persistence and growth of ecommerce, which now ranges between 12% and 19% of all retail sales depending on time of year and industry segment, has been the final straw, vaporizing the margins of retailers and forcing some difficult decisions.
The great hope is that artificial intelligence is here to help. And it will, if incrementally at first.
Machine Learning Is Coming To Retail
To many retailers, the notion of artificial intelligence is science fiction. The mass of small, brick and mortar single owner retail stores are not going to find a lot of use with machine learning today.
What about the biggest retailers, those that have a little room to maneuver and build new systems to take advantage of the changing consumer behavior?
“At the end of the day, it is happening. Since we deal with a lot of the larger retailers, you see first hand that it is happening,” said Grant Ingersoll, chief technology officer of Lucidwords at the Internet Retailer 2017 conference in Chicago.
“There are a surprising amount of companies between the $500 million and $2 billion market where once they find that right niche they will do well for themselves, make a name for themselves. That is a big market. The real losers are the brick and mortar ones that haven’t figured out the omnichannel model.”
The question of machine learning for retailers is multifaceted. It stems from the omnichannel world that has been built over the last decade. It stems from updated backend systems and new sources of data. Machine learning branches from the furthest reaches of the warehouse all the way to the floor associate.
Lucidworks, through its Fusion engine, uses machine learning and natural language processing to help companies improve enterprise search (for both external, consumer-facing entities and internal). Search, almost by definition, implements a sort of artificial intelligence to help produce the most accurate queries. The lessons of machine learning taken from search can apply to any variety of retailer features and functions, such as recommendations, pricing, consumer history and so on.
And yet, it is well known that retailers have had difficulty identifying technology needs and hiring the right people to fulfill those needs. Even the biggest retailers might only have a couple of data scientists on their team.
“And maybe what they really have is an engineer that happens to be good at math,” said Ingersoll. “Or a business analyst that happens to be good at math. So therefore that person is now the company data scientist. They just can’t keep up with all the different ways they want to do it. They want to apply all of these different things, not just across search, but across supply chain and all aspects of it to figure out what really matters in the business.”
The trick for retailers in implementing machine learning is to aggregate their data and be able to apply it to individual consumer intent. Is a customer searching for a hammer because she wants a hammer? Or is she building a tree house and knows that a hammer will be an essential item in that project.
“Being able do understand the user’s intent up front is where a lot of these retailers want help,” said Ingersoll. “The solution there is often a machine learning-based approach. Analyzing all of your prior interactions, those kinds of things. It is essentially a classification problem.”
Fusion helps companies through these problems by implementing machine learning either on premises (on a retailers own servers or locations) or through the cloud. Fusion can help aggregate all the varying sources of data—email, search history, recommendations, email campaigns etc.—through hooks and APIs and apply machine learning models on the data for more actionable results.
Machine Learning Has To Be About Sales
The type of machine learning that companies like Lucidworks offer has a variety of benefits. By loading different silos of data into one model, the mess of backend data that most retailers face can be cleaned up. Internal decisions based on the predictive and intelligent analytics machine learning provide can help companies understand and act on trends or issues quicker.
But, in the end, it has to come down to sales.
“Ultimately, a lot of these conversations come out as we are trying to get better results, we are trying to get better recommendations,” said Keith Messick, chief marketing officer of Lucidworks. “We are ultimately trying to sell more. I think that is the beauty of retail in that the motive is very simple. If you are doing search intra(company), it is about productivity, it could be a whatever kinds of stuff. [In retail] it is about how do we sell more stuff, how do we sell more units per transactions.”
Companies like Oracle tend to agree. In a white paper on artificial intelligence and machine learning in retail for its Bronto commerce engine, Oracle states: “machine learning offers the greatest revenue generation potential for commerce marketers. Software developers are applying it to build tools that can analyze millions of data points about shoppers’ preferences and actions to create a more personal experience.”
Messick and Oracle are both right: for retailers, it has to be about sales. Unlike any other industry, retail is all about moving product off shelves. Machine learning is about making that process as efficient as possible, connecting the consumer with the product as seamlessly as possible.
“It is really about how they make that more efficient, how they improve that, how do that make it faster, how do they maintain infrastructure, how do they make it work seamlessly with search,” said Ingersoll.