Top 10 use cases to get started on your AI-powered personalisation journey

Personalisation in retail is at another inflection point, driven by data, AI and automation at scale. The amount of data in the datasphere is expected to grow from 175 zettabytes in 2025 to 393 zettabytes in 2028 – a growth rate of >30% year over year¹. This data, if harnessed, can give retailers more ‘signal’ to build profiles of their customers. Advances in AI mean that signals can be deciphered and applied to customer profiles, making those profiles richer, with the ability to make highly accurate behavioural predictions. When combined with intelligent automation, where business rules are applied in conjunction with relevant predictions defined by AI, retailers can step into the new era of personalisation. Retailers that take steps into this future will outcompete all others, achieving the dual aims of improving customer experience while growing sales and profitability.

Taking one giant leap into the future can be seen as a giant leap of faith – right for some, not right for others. One approach is to start small, learn then expand while building the organisational confidence to do more and increase the pace of change.

At dunnhumby, we have been serving major retailers with personalisation services for more than 30 years. Every week we work with over 1,300 brands and 50 retailers, serving more than 12bn recommendations to over 90 million customers across 25 countries. To retailers that want to get started on this journey, what should they do first?

From our experience, these are the top 10 entry points to ‘start small’ and build the organisational confidence to embrace a bigger, ambitious AI-powered personalisation transformation.

1. Onboarding. Relationships start at the beginning. For customers, it is the first time they browse a website or enter a store. These moments are opportunities for retailers to get to know their customers and invest in the relationship. Retailers can incentivise customers to self-identify or declare their preferences. And when signing up to a loyalty programme, retailers can send a targeted campaign of communications, to create a relevant and compelling dialogue, to inspire customers to start shopping. Prompts for personal information, the logic for welcome comms and the AI for relevant content can create an engaging start to the customer/retailer relationship.
2. Lapsing customer reactivation. Customers trade down when their finances take a turn for the worse. They may also split their shopping with other retailers due to convenience (e.g. a store is close to their office) or because they have found something unique at a competitor’s store that they really like. Retailers can see these early patterns in the data, understand the motive, and respond with inspiring, informational messages or help customers out with personalised discounts. All sent through the most relevant channel (e.g. SMS, email) at the right time of day for best engagement.
3. Upsell at point of sale. When you get to that point on an online shopping journey where you’ve made up your mind, there are opportunities to present new, relevant opportunities to customers that they might like. Missed a promotion? Let them know. Offer better value with a bigger pack size (from beverages to socks)?Let them know. What is presented to customers – through a pop up or panel in a digital journey – must be relevant. A personalised prediction will find that sweet spot of something the customer wants (i.e. They will find it enriching) and has a high likelihood of conversion to drive sales for the retailer.
4. Cross-sell at point of sale. This is very similar to upsell, the difference being that it relates to products that complement those in the basket. One point of difference is the additional option of using visual search for general merchandise or fashion. Chosen a yellow dress, how about some white sandals? Picked a garden table, how about these matching chairs?
5. Footfall boosts. A competitor opens a store nearby. There are adverse weather conditions that impact sales. People are shopping more online. These, and other causes, may lead to a decline in footfall into a particular store, leading to decreases in sales. Predictive models can help nudge customers back into named stores, by issuing time bound personalised offers, relevant and inspiring messages or a mixture of both. The science is in finding those customers with the highest propensity to make the visit, that wouldn’t normally, then sending them the most meaningful outreach at the right time of day through the most relevant channel. Footfall goes up reversing the unforeseen woes.
6. Retention of loyal customers. We can’t take loyal customers for granted; the maintenance of their loyalty is as important as its acquisition. Retention is easier and cheaper than winning back a customer, and can be done through Surprise & Delight initiatives, for example, in which the customer is rewarded without a condition asked – just for being as they are. A relevant recommendation of a product that is important for that customer can also do the trick without necessarily impacting the margin.
7. Product alerts. Customers give us several signs of their interest in a product: browsing, clicking, buying, buying similar, upgrading from something else... but sometimes the customer needs a little push to finalise the purchase. A price drop alert is a good example that drives conversion without the need of an additional discount. Customers that were not sure about it might just need that to finally click on ‘order’ or go to the store.
8. Basket abandonment & saved for later. Whether items are left in the basket, with a view to saving them to purchase at a later date or a full order has just been abandoned, the retailer can send helpful nudges and communications to the customer. The secret is in maximising engagement with that customer. Some customers prefer email, some SMS. Some customers read messages first thing in the morning, others in the evening. Some customers respond well to reminders; some jump into action with time bound offers. To help nudge customers to make that purchase, predictive models can send the right type of message, through the right channel at the best time, aiming for optimal engagement. As a fast start, pick a channel and learn, before optimising across channels.
9. Substitutes. What happens if a customer wants something that is unavailable? Rather than disappoint, recommend something that they will most probably like – something similar, for a similar price, from a similar brand. This is most applicable to eCommerce journeys where these recommendations must be made in real time, reflecting predictions of how the customer will respond to available inventory.
10. Surprise & Delight. Retailing used to be personal. Now it can feel anonymous and customers feel unseen. How can technology bring back that personal touch? Perhaps a triggered, personalised message with a gift on a customer's birthday or a message to celebrate the anniversary of your loyalty programme membership? Perhaps a thank you for leaving a review? Or if there is an adverse experience of some sort, an apologetic message with a personalised discount.

The secret, to all of this, is capturing vast amounts of data, of different types, attributing it to individual profiles which are updated in real time. This way, behavioural predictions are always fresh and accurate. But it doesn’t need to happen all at once.

The transition to a fully personalised perimeter, powered by data and AI, is a big step for most retailers to take – especially those with a strong bricks and mortar presence. The upside can be considerable, growing profits by up to 40%. However, the technology changes can be profound, and too far to go in one step, for some. The important thing is to acknowledge that this future is inevitable, to get started and not get left behind.