AI and Personalisation: explanations, limits and drives
Guided by the reasoning “offer the right product to the right customer at the right time!”, retail professionals have moved to personalise marketing strategies.
Always transform better, constantly increase the average basket and consistently maintain customer loyalty while responding to relationship and emotional challenges and offering each customer a bespoke experience: given this never-ending quest for personalisation, let’s examine how AI has boosted the rise and conditions of one-to-one marketing.
Personalisation or Hyper-personalisation: what do we mean?
Marketing efforts focused solely on the product are over; the client-first (or customer-centric) wave has overrun everything in its path, repositioning business and branding issues with the customer in mind. Buyers are flattered to find themselves the subject of so much attention, and 80% of them say they are more likely to buy from a brand that offers them a personalised experience.1 Let’s take a simple approach: if we look at it from a minimalist angle, personalisation consists in integrating personal and transactional information (name, purchase history, etc.) into communications.
This data only represents 1.6% of the data that purchasers generate. Nearly all the "Digital Body Language"2 resides in behaviour data through the millions of signals that customers send to us on an online sales website. Hyper-personalisation is conceived as the exploitation of often deeply immersed data, in particular to use it in real time to create extremely contextual and relevant communication for the website visitor. A quick example based on email targeting:
- With Personalisation, the user will receive a flash sales email with their first name in the subject line for shoes belonging to the same category (sports) in which they bought a pair about six months ago.
- With Hyper-Personalisation, the level of contextualisation is far more advanced: this user, after browsing the site and not purchasing anything but leaving traces of their visit, will receive a message built around the trainers they saw, their search and purchase history for this range of products, and at a time of day that matches the time they usually make a purchase.
Capturing and processing data within the imposed framework
Generally, we separate hot data from cold data. Cold data comes from a CRM or DMP, a pre-existing set of data such as age, gender, purchase profile, etc. Hot data is captured as the visitor browses the website, such as behavioural data on the content viewed or searched, the time spent in various categories, technical data (such as device and browser), and contextual data (time, geolocation, etc.). Exploiting this data is naturally at the heart of how personalisation works.
This processing must be executed precisely and immediately; the customer must be influenced and persuaded in real time! It reaches a point that we come up against certain paradoxes. Data collection combined with algorithmic processing allows for content to be adjusted in real time, which could result in limiting discovery and "free", or at least non-imposed, browsing. It’s similar to the criticism about tribalism that was made of Facebook a few years ago. Consumers, reassured by the implementation of GDPR principles, now require more transparency and want to know what data brands use to optimise their personalised systems.
According to a recent study, 83% of respondents are ready to share their data for personalisation purposes3, but only insofar as brands are transparent about how they use it, and users maintain control over the information sent. As you have seen, the customer experience must be built around consent, transparency and intelligence in the use of data. ?
Driving AI to serve marketing
Given the boom in shared data and the proliferation of targeting criteria to be triggered in real time, Artificial Intelligence has naturally found its place in this context of the automation and optimisation of the customer experience. Let’s remember that Machine Learning and AI already have extensive scopes of application. For our interests here, the algorithms used will be able to offer buyers the most relevant product suggestions and recommendations (similar to pioneer Amazon) and more.
How? Based on the more numerous customer segments, the algorithm will make predictions founded on its calculations and the correlations observed between those who bought the same offer or those who found the same message appealing. All while processing an ever-greater volume of data and with unparalleled reliability. On the market, a wide-ranging ecosystem of solutions has grown around AI combined with hyper-personalisation. For example, we can cite Kameleoon, Sparkow, Nosto, Nuukik, Untie Nots and Target2Sell, each of which is working to improve its predictive algorithm to optimise merchandising, content management, search, promotions/loyalty systems, and omni-channel targeting (email, social media).
See the SAP Commerce Cloud solution [/caption] For now, let’s take a closer look at the Context-Driven Services solution, an add-on to the SAP Commerce Cloud offering, which includes:
- A principle of consent for any use of customer insights (Consent Management within the Store Front),
- The definition of segments in the Segment Builder by managing attributes and conditions in relation with the orders, the basket content, and the customer,
- The ability to dynamically assign customers to the segments dynamically so as to present them with specific content (content, promotions, search) in real time, whether the user is logged in or not,
- As the visitor browses, the service enriches the profile, constantly making new calculations in the customer/segment assignments to drive a personalised experience on the site's front office, notably via the SmartEdit CMS.
As we have seen, personalisation brings ever more value to customers given its advanced attention to contextualisation, whose perceived benefits spill over into the relationship between the customer and the brand...as long as the brand has built a climate of trust in how data is processed and if the consent to share can be withdrawn at any time.