Can a machine know a customer better than they know themselves? The answer is, for the purposes of shopping, yes it can.
First of all, Artificial Intelligence (AI) takes a dispassionate view of customers and their behavior online, while in research, consumers will often give contradictory answers, answers that then change over time, depending largely on how they are feeling at that particular moment. As an indicator of how those consumers are then likely to behave in terms of what they buy, this has been proven to be unreliable.
AI on the other hand, supported by machine learning to deliver better and better outcomes over time, operates without emotions and simply reacts to – and learns from – what it is being told.
In online retail, AI is set to revolutionize the world of search. If revolutionize sounds too big a word for it, bear in mind that search technology has barely changed in 10 or more years. While brands have invested heavily in making their websites look amazing and optimized them to steer the customer easily to the checkout, they have generally used out of the box search technology ever since the first commercial engine was launched by Altavista back in 1995.
- Promising artificial intelligence startup ideas for 2020 (opens in new tab)
Traditional vs agile online search
Given that typical conversion rates on retail websites is 2-3 percent, then there is everything to play for in making search easier and more rewarding for shoppers. Retailers invest heavily in SEO and PPC to get customers from Google to their site but too often think the job is done once they get there.
Products are then displayed to their best advantage on the site; email or newsletter sign up is offered; online chat is offered; promotions pop up; a list of nearby stores is offered; and so on. But at no point is the customer offered or given any help, apart from the online chat window which follows them around.
At this point, the customer may well start to follow the journey laid out for them by the retailer; they get distracted and end up somewhere entirely different from what they intended. Some customers like to wander, but those that already knew what they were looking for do not.
Meanwhile, what has the retailer learned from all the precious time the customer has spent on their site? Only that the customer has not bought anything, and it is only at this point that an offer pops up or the online chat box appears. But none of these actions are based on any knowledge of the customer other than which pages they have looked at.
Turning search to understanding with AI
The search engine is not very good at learning; it may be able to refer the customer back to a page they looked at before because of the consumer’s digital footprint or due to the cookie the site left behind, but if that webpage was not useful, then the search process has actually gone backwards. So, the customer continues to end up where they never wanted to go in the first place – ever decreasing circles displaying a choice of unwanted products.
These on-site search functions can be compared to stubborn school children who simply refuse to learn, whatever they are taught. The customer searching online tries to make their query as accurate and intelligent as possible while the search engine simply responds by sharing everything it knows, but without actually answering the question. AI by contrast can spot what the customer intends and gives answers based on that intent, depending where an individual shopper is in their own personal buying journey.
It then returns increasingly accurate results because it is learning from what the customer is telling it. Search thus becomes understanding because it is looking at behavior not just keywords, which is the current limit of conventional search engines. The AI can also create the best shopping experience beyond basic search, including navigation, to seamlessly and speedily advance a customer to the checkout.
This is really what delivering personalized journeys is all about – the site understands the customer, knows what they want and how they want it. For instance, when a shopper is very clear about what they want, the AI can plot the quickest route through the site to the payment page, while customers looking for inspiration can be given a slower and more immersive experience, with lots of hand-holding as required, such as links to online chat to help them with their decision or curated content to inspire browsing.
- How to effectively deliver an AI transformation strategy (opens in new tab)
Move from keyword search to results based on consumers’ buying intent
AI in ecommerce assumes a character all of its own, essentially a digital assistant that is trusted by the customer to help them find what they want. Retailers can personalize AI in any way they choose, while the processing and intelligence that sits behind it continues to work unseen.
AI in action of course creates a huge amount of interactional and behavioral data that the retailer can use to make improvements over time to base search, navigation, merchandising, display, promotions and checkout experience. It delivers good results for individual customers as well as all customers as their online behavior continues to evolve.
Our view is that customers want help when they are on a website. They want to be able to ask questions using natural rather than search language and they want the search function to learn based on those answers. By ensuring that their search strategy is underpinned by AI, retailers can then introduce more dynamic search enablers, such as visual and voice. But rather than simply adding commands, the customer is able to hold conversations with the digital assistant using natural language. Search then turns into discovery and it is this that leads to higher customer conversions, repeat visits and long-term loyalty.
To date, a lot of the conversation around AI has focused on the technology rather than what it enables in the real world. And there has been some reticence to adopt it for fear that it will replace human jobs; however, in the case of online search, one automated process is simply complementing another and all in all, doing a much better job. Check out your own search function now. How is that working for you?
- Artificial intelligence trends of 2019 (opens in new tab)
Jamie Challis is UK Director, Findologic (opens in new tab)