True 1:1 Personalisation Is Here. Here Is What It Looks Like.

There is a version of "personalisation" that has been declared imminent since roughly 2010. It usually involves a customer receiving an experience precisely calibrated to their individual context: not their demographic segment, not a behaviour cluster, but genuinely them. And responding with something like delight rather than the vaguely uncomfortable feeling of being surveilled by a system that knows slightly too much.

The gap between that vision and reality has been wide enough, for long enough, that experienced people in the industry have developed a healthy scepticism reflex when the word "personalisation" comes up in a presentation.

So let me try to be specific about what has changed in 2026, and what has not.

What Is Actually Different

The infrastructure has crossed a threshold. Large language model APIs, real-time customer data platforms, and multi-channel orchestration tools have combined to make it technically feasible to generate a genuinely individual experience for every person in a database of hundreds of thousands. For email, this means subject lines, body copy, product recommendations, and send-time optimisation, all generated per individual, not per segment. Instead of A/B testing two subject line variants, you generate a variant specifically for each customer's profile and history.

Klaviyo and Salesforce Einstein are among the most widely used enterprise platforms doing this in 2026. For UK retailers, Marks and Spencer, Tesco, and Boots are the most publicly discussed implementations. All three are using their rich first-party loyalty data (M&S Sparks, Tesco Clubcard, Boots Advantage) as the training signal for personalisation systems that operate at the individual rather than segment level.

The results are real. Early trials of AI-powered individualised campaigns have shown 10–25% increases in return on ad spend. Practitioners report cases where ROAS has improved alongside lower wasted impressions, even as click-through rates fell slightly. That divergence matters: improved downstream conversion quality, not just more clicks. (These figures are drawn from US and global market studies; directly comparable UK benchmarks are harder to find published, though practitioners report directionally consistent results.)

The Honest Limitations

The thing I keep pushing back on when this topic comes up is the gap between "generating individual messages" and "understanding an individual's context."

The former is a solved problem in 2026. The latter is much harder and depends almost entirely on data quality. If your customer data platform has a clean, complete, consented view of a customer's purchase history, browse behaviour, returns pattern, promotional response, and channel preferences, you can generate something genuinely useful to them as an individual. If your customer data has the gaps and inconsistencies that most retailers' data actually has (multiple accounts for the same person, sparse behavioural signals, consent holes, data that is months out of date) the "personalisation" ends up being confidently wrong, which is worse than being generic.

Tesco Clubcard was launched in 1995 and represents more than thirty years of individual-level purchase data. That is not a typical starting point. Most retailers trying to stand up individualised personalisation in 2026 are working with considerably patchier data, and the AI cannot manufacture signal from noise.

There is also the uncanny valley problem. The experience of receiving a message clearly generated specifically for you, referencing your recent purchases and anticipating what you are about to need, sits differently with different consumers. Some find it useful. Some find it mildly uncomfortable. Some find it actively off-putting. Research on consumer attitudes to AI personalisation consistently shows that acceptability is context-dependent and consent-sensitive: people are more comfortable with personalisation they have explicitly opted into than with personalisation that appears to use data they did not know was being collected.

The Privacy Question Is Real

Dynamic one-to-one pricing and promotions, where incentives are tailored to individual behaviour and price sensitivity, is the personalisation application that generates the most regulatory and consumer trust concern. It sits close to price discrimination, and the line between "offering you a discount because AI predicted you would churn without one" and "charging you more because AI predicted you would pay more" is thinner than most marketing communications around the technology acknowledge.

In the UK, the ICO's guidance on AI and data protection alongside the government's AI Opportunities Action Plan sets out the domestic framework. UK retailers serving EU customers are also in scope for the EU AI Act. The practical upshot: automated personalisation decisions need to be explainable, and the data used to drive them needs to be lawfully collected under UK GDPR. Retailers doing individualised pricing and promotion need to have clear answers to questions about the data used, the decision logic, and the human oversight mechanisms, because those questions are coming.

Where This Is Heading

True one-to-one personalisation is real in 2026. It is producing measurable results. The data infrastructure to support it well is unevenly distributed, and the retailers doing it best are mostly the ones who have been building first-party data assets for years. The gap between those retailers and the rest is getting wider, not narrower.

The philosophical question of at what point an AI system knowing you better than your loyalty card does becomes something consumers are genuinely uncomfortable with does not have a settled answer. It is probably not the same answer for all consumers, all product categories, or all relationships between brand and customer.

That ambiguity is worth leaving unresolved rather than resolving it prematurely in either direction. The "personalisation is always welcome" framing of vendor marketing and the "personalisation is inherently surveillance" framing of the most alarmed critics are both too clean for the messy reality of how people actually feel about it.

---

Data sources: Bain and Company, Retail Personalisation · CX Today, UK Retail Loyalty AI · ICO AI and Data Protection Guidance · KPMG UK Attitudes to AI