Emma Herrod looks at how artificial intelligence is being made more human online.
Research shows that brands and retailers believe artificial intelligence (AI) will give rise to significant opportunities across retail, creating more meaningful relationships with customers.
While almost 60% of businesses would trust fully automated AI-driven marketing campaigns to deliver meaningful content to their customers, only 45% of customers believe personalised adverts displayed to them are relevant, according to a study by international law firm CMS, in partnership with Retail Economics.
Those same consumers are not convinced that retailers are able to handle their sensitive data either. In fact, retailers and brands are trusted only slightly more than social networks – a rather scathing indictment given recent scandals.
However, the study does show that most organisations acknowledge that AI ethics must be a focus for them – 66% of organisations surveyed believe they will require roles managing data ethics.
Once we start to look at more subtle areas of personalisation, such as site navigation, studies show that consumers don’t realise that the content being shown to them is personalised.
The latest version of Apptus eSales merchandising solution, for example, has been optimised for the fashion industry so that its AI engine can understand nuances such as differences in colour – light blue is a two-word way of describing a shade of blue rather than showing a customer a mix of dresses and blue table lamps. It also knows that burgundy is a shade of red and that teal isn’t just a type of duck. All of this is determined automatically through image analysis rather than tags. Sizes are optimised too rather than being a muddle of UK and international sizes.
ESales Fashion combines on-site search, category listings and recommendations which adapt in real time to deliver an experience that’s relevant to each consumer. Each component in the single intelligent merchandising solution learns from and informs the others. The results from the search auto-complete, for example, are optimised by colour, size, availability, site rules etc.
While the search results aren’t personalised to the individual user, since this has been found to harm conversion, the AI engine does optimise for the site’s rules. This results in less products being shown but they are more relevant and more likely to be what the customer is looking for, explains Apptus UK Country Manager Andrew Fowler.
The overall system is ‘personalised’ for the fashion vertical and much of the workings incorporate personalisation such as recommendations being based on past history or the first row of products returned.
For the retailer, the AI engine takes away a lot of the manual labour and makes processes scalable, freeing up merchandisers’ time as well as enabling the site to be optimised according to the business objective likely to deliver the best results whether that’s conversion, revenue or profit. Predictive analytics inform likely outcomes.
One Swedish retailer switches to optimising its site for revenue for six days following pay day before changing it back to being optimised for conversion. This has proved to be a successful strategy according to Fowler but A/B testing could be used to see the different impact.
A further change with eSales Fashion is that the UX has been optimised to how consumers shop for fast fashion. The web component includes the best UX experience for the end consumer, explains Fowler. “We found that people were using our technology to get great intelligence back but losing some of the benefit of using our AI by implementing the UX experience badly,” he says.
Apptus therefore has moved to combining a good UX interface along with AI intelligence, he explains. “We want to present the UX for fashion,” says Fowler and bring in UX capability alongside the intelligence. “This delivers better results for the shopper and better results for the shop,” he says. “And a better experience all round”.
This is showing results for the first customers which includes Joe Browns. “We substantially overachieved Joe Browns’ success criteria for their A/B test,” says Fowler.
Adobe is another company aiming to give shoppers a better experience on retail sites through AI. New innovations within the latest version of its Adobe Target solution enables marketers to deliver “more impactful, personalised experiences to their customers faster, and at a greater scale than ever before,” the company says.
These capabilities include AI-powered personalised recommendations, AI-enhanced reporting, mobile app personalisation and what it says is an industry-first solution for single page applications increasing the speed and responsiveness of personalised areas making sites faster to load.
Building on last year’s launch of personalisation insights reports, the new solution enables retailers to assess the effectiveness of any campaign seeing the actual relative lift that each offer or recommendation is delivering against the default page. Amongst the customers for its mobile app personalisation solution, which aims to enable retailers to improve app retention, engagement and conversion, is a pharmacy which is automating the ranked-order of the navigational icons in its app based on customer behaviour. Another example showcases how location-based personalisation is being used by a major hotel chain to trigger personalised screens with recommendations of amenities and upgrades.
Test and evaluation have been key for retailers in recent years so it’s no surprise that systems based on AI are incorporating ways for the AI engine to be questioned and tested too. RichRelevance, for example, enables retailers to see easily how and why the strategies and models are working in a certain way and leverage their own models alongside the AI strategies to perform tests in real time.
Its latest release has also been enhanced with natural language processing so that the solution can understand unstructured data, auto discovery of behavioural shifts so that customers who change behaviour such as an increase in lifetime value can be handled differently. A visual component has been added allowing easier analysis of campaigns and interrogation of the AI element.
The company aims to push the boundaries of one-to-one personalisation further across all touchpoints including web (site search, navigation, products), mobile, email and contact centre turning every digital interaction into a personalised experience.
Carl Theobald, CEO of RichRelevance explains that having an open AI solution means that retailers can see easily how and why their strategies and models are working in a certain way and are able to leverage their own models alongside the AI strategies to test them in real-time.
As the Apptus solution makes searching for specific colours more natural, so the ability to understand natural language by RichRelevance’s latest release shows how the industry is moving from algorithms and rules to a more human and natural experience. Information in unstructured data, such as ratings and reviews, can be used to help map products to customers, while an understanding of behavioural shifts means a marketing team can be better informed. New campaigns can be designed quickly and implemented with content optimised automatically.
AI is enabling one-to-one personalisation and merchandising in a way that human marketers would never be able to achieve and with a shift to greater trust being put on AI by retailers it is being humanised too. It can work out where customers are behaving in new ways so campaigns need to be created – by the human marketing team – and optimised by AI for what the retailer wants to achieve.
This may be revenue, conversion, profit or to increase engagement on the site to increase learning and understanding of certain customers. Whatever the rules, trust and transparency between system, retailer and customer is growing.