In a recent InternetRetailing webinar, The journey from data to shopping experiences: making AI work for you, Peter Thomas, CTO and Chintan Gupta, product manager at Attraqt, shared insights and advice on how retailers could reassess and reset their AI automation strategies. Here’s a bulletpoint overview of the presentation.
• Opening the webinar, the presenters asked how useful AI was to those taking part in the event.
Poll: Does AI assist you in your job today: 28% said they didn’t think so, 39% said not as much as I would like.
In 2017 45% of marketers said marketing mostly involved real-time personalisation
By 2019, 90% expected to put marketing personalisation into use, moving to smart personalisation in 2020.
By 2020 expectations of very substantial use of AI in digital commerce organisations.
By 2022 this will control interactions to a larger degree
• Using IoT
• Digital experience extending in store
• Mobile: as much as 85% of commerce traffic already on mobile. Thomas: “From a retailer’s point of view you’re living in your customer’s pocket.”
• Social: about exchanging views
• Data privacy and security one of the most important for consumers.
• Changing behaviours for retailers and other websites
• Transparency: Rights don’t necessarily mean engagement and trust
• Cross-channel shopping: wouldn’t it be great if you were connected from online experience to the in-store experience? - but that doesn’t happen much at the moment.
• Increasing amounts of choice – and data. As merchandising teams it’s hard to cope with that volume to provide a quality merchandising result.
• Search, personalisation through AI.
• Voice, image, all important ways of being able to enrich the conversation to make merchandising more productive and deliver more engaging experience.
“Most of the devices we have currently are already to equipped to deal with additional signals” - tone of voice, scanning of items through the camera, mood, reflexes and heart frequency are also “going to be influences in the algorithms that actually drive experiences in the future and it’s important to start thinking about that now”.
Merchandiser defines the strategy, AI executes it, learning and adjusting towards goals that have been set.
Factors in the shopping experience include trends, social, profile, context, intent
“It’s about making sure you’re understanding the shopper and the meaning being conveyed through actions.”
Simple ordering of search and navigation. We need more than just a basic global ranking, or only one or two things sell. Useful when little contextual or personal data, but prone to self-fulfilling results.
• Not collecting direct information but enough information in terms of their location, context, time of day that might enable you to give the a digital fingerprint. “If you match that to other shoppers that have characteristics to suggest to your new shopper what they may be interested in.”
• Automatic segmentation to infer shopper preferences - apparent personalisation, appears to be personalised.
• The navigation on your site is just a suggestion for what might happen. Seen them before, see them again. Collect information in general terms, recommendations while still remain anonymous.
• One of the problems with relevancy is looking for a silver bullet - but let’s look at the things reducing relevancy: you may use knowledge of past transaction history if logged in on the site to say you’ve already bought that, may not want to buy it again.
• If you have the data why force your shopper to make that choice again, remove the choice. If you have personal measurements, companies specialise in matching those measurements to common brand sizes) “You can remove choice because you’re not presenting, for example, sizes of clothes that are irrelevant.”
• You may find you have more data than you realised in terms of being able to improve relevance just by removing those irrelevant data choices.
• Adapts to current behaviour.
• Use NLP to understand synonyms. Update search and navigation, recent activity quickly outweighs past behaviour.
• Refining navigation to confirm understanding.
• Poll: 28% said used test and learn with external analytics and reports prepared by other teams, 50% used test and learn with reports built into the tools, 22% not yet.
Process of testing and learning
“It’s very important to adopt a team methodology that reflects review and adaptation. It’s a valuable source of information to have within the team much faster than you would have.
“The important thing is that you develop a methodology within the team so that you know what you’re looking at, and you can adapt”
• Great feedback mechanism.
• Privacy now a major issue.
“A number of clients we’ve been working with are really tuned into this connection with their shoppers because it’s an important part of their brand. They are deliberately designing their shopper experience to be very explicit about consent. They present up front exactly what information is collected, what it’s for and make it very clear how they can revise their choices going forward. It generates a lot of trust on the part of the user.”
Watch the webinar in full below: