The Problem with Hyper-Personalisation
We are still getting it wrong.
Background
As far as I can tell, the term “hyper-personalisation” started to appear in marketing literature some 5+ years ago. It felt like marketeers wanting to ride on the AI bandwagon by creating new jargons and buzzwords. Early approaches to hyper-personalisation didn’t really have any meaningful substance; they were re-labelling what already existed as maturing recommendation engines in the consumer space, citing use cases like Spotify and Netflix. They were bringing up the “holy grail” of marketing which is 1-to-1 segmentation, a highfalutin concept that is not rooted in evidence, and that I can’t agree with.
With the rise of AI-enabled products and services, hyper-personalisation continues to be on the strategy radar; it continues to be a major trend in 2025 and the near-future beyond. However, many organisations still can’t execute against it, and even when they do, they don’t get it right.
And so my 108th article is a discourse on why organisations simply don’t understand what is hyper-personalisation, what it takes to execute it well, and how to sustain it.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
Back to Basics
Personalisation is a marketing concept that proposes the use of customer data to tailor more relevant marketing messages, content, and experiences. What distinguishes hyper-personalisation from personalisation is the dealing-up on the use of data, often with the help of AI, to make the targeting more customised and more pervasive and integrated across multiple customer touch-points. Other buzzwords are also thrown into the mix, like “dynamic” and “customer journey” to make the new term sound more sophisticated. The reality is that the line that separates hyper-personalisation from personalisation is often not clear, given this definition of simply going deeper and wider. It’s also the reason organisations can’t seem to implement hyper-personalisation well. They often act on assumptions instead of understanding, they over-personalise and become intrusive, they cannot maintain coherence across channels; all these dilute the organisation’s brand identity.
There’s a real need to sharpen the definition of hyper-personalisation for us to make it truly usable. I propose defining hyper-personalisation through the shifting of 3 dimensional attributes of personalisation:
- From Reactive to Proactive
- From Behaviour to Intent
- From Dynamic to Adaptive
Reactive to Proactive
Most hyper-personalisations are reactive. The marketing actions are triggered by and in response to a user’s explicit action. No action, no targeting, no offer. These organisations tend to see hyper-personalisation through the lens of recommendations; i.e. recommendation engines. And they are obsessed with making better recommendations (to increase conversions) by incorporating more and more customer data. Reactive actions do not build durable customer relationships. A reactive mindset is a transactional mindset.
Good hyper-personalisation needs to be proactive. It needs to anticipate what the customer is trying to solve for, and then present a frictionless path forward for them. Being proactive significantly expands our range of context awareness. For example, a coffee bean company notices that a customer has been consuming coffee beans at a consistent rate, and sends a notification that an order for a new bag of coffee beans has been pre-filled before it runs out. For example, the organisation is planning to maintain its website and understands that you, as a customer, consistently logs in at that time, and so the organisation sends a notification to let you know to log in earlier instead.
Behaviour to Intent
Most hyper-personalisation approaches are set up to monitor customers’ behaviours, sometimes leading to customers experiencing surveillance paranoia. While behaviour data is essential for hyper-personalisation, it misses the deeper abstraction of intent — i.e. what is it the customer is solving for? In some cases, it is obvious what the customer’s intentions are. And in those cases where they are not obvious, organisations need to develop analytical solutions to predict it, rather than relying on explicit surveys. Organisations need to build “intentions prediction engines” because understanding intentions is a key reinforcing ingredient to enabling proactivity in hyper-personalisation.
Dynamic to Adaptive
Many organisations boast that their hyper-personalisation execution is dynamic, able to react according to how the customer responds. It’s a glorified set of “if-then” actions. True hyper-personalisation is adaptive, meaning that every interaction, whether resulting in the desired outcome or not, is an opportunity to learn something about the customer. Rather than just ignoring it, a non-response to a targeting action is a chance to update our baseline assumptions for that customer. Smart organisations would be instrumenting intelligent feedback loops. This is a much more sophisticated process that requires serious architecture design to explicitly record the underlying assumptions, and how it is being continuously estimated and re-calibrated through the feedback loop — what data would be valid or invalid to support that. And how those changes in underlying assumptions trigger off a new set of hyper-personalised actions. It’s a marriage of data science and behaviour science.
Conclusion
When I first started exploring this topic of hyper-personalisation, it was with a view to debunk it as marketing fluff. Observationally, I knew it was not living up to its original expectations. But as I dove deeper into the practice, I realise the half-successes were due to a lack of rigour in both definition and design. Moving the hyper-personalisation needle is predicated on the need to understand intent, preferences, context and historic behaviour, all coming together as an integrated whole. There is every opportunity for hyper-personalisation to be a competitive advantage for the select few organisations that can master it.
