The Personalisation Line

A shopping app recommends running shoes after you browse sportswear.

Helpful.

It recommends shoes in your size and budget.

Still helpful.

It sends a message: “Based on your late-night searches, recent weight-loss content and walking patterns, these may suit your new fitness goals.”

The same personalisation has crossed an invisible line.

The customer moves from “You understand me” to “Why do you know that?”

Relevance and intrusion travel together

AI-enabled personalisation improves relevance by using more information about the person.

The same information creates privacy risk.

A review of AI-enabled personalisation describes this as the personalisation-privacy paradox: customers can value tailored experiences while resisting the data collection required to produce them.

A 2025 systematic review of privacy concerns across AI, IoT, augmented reality and big-data environments similarly emphasised that privacy reactions depend heavily on context, perceived control, risk and willingness to disclose.

There is no single amount of personalisation that everyone welcomes.

The surprise test matters

Customers often tolerate data use they remember giving permission for.

The discomfort rises when the inference surprises them.

“I told the app my shoe size” feels different from “the app inferred I may be pregnant.”

“I searched for hotels” feels different from “my bank used location, purchase history and family data to predict travel intent.”

The deeper the inference, the more the company needs to think about explanation, consent and benefit.

Sensitivity is contextual

The same person may happily share food preferences and strongly resist sharing health or financial details.

People may tolerate personalisation from a platform they use daily but reject it from a brand they barely know.

They may accept data use that produces obvious value and resent it when the benefit is trivial.

This makes privacy a relationship question, not merely a compliance checkbox.

Build personalisation around permission and proportionality

A useful principle is:

Use the least surprising data necessary to create the promised value.

Explain why the recommendation exists.

Let customers adjust personalisation settings.

Avoid using sensitive inferred information simply because technology allows it.

Make “no” easy.

And remember that data can be legally usable yet psychologically invasive.

Trust operates above the regulatory minimum.

The long-term commercial issue

Aggressive personalisation can improve a click and still weaken a relationship.

A brand that feels observant can feel useful.

A brand that feels watchful can feel unsafe.

The difference often appears in one sentence:

“Because you told us…” feels very different from “We noticed…”

Questions worth sitting with

• Which customer data would feel surprising if you described its use plainly on the screen?
• Does your personalisation create enough value to justify the intimacy of the data used?
• Are you optimising only conversion, or also the customer's sense of control?

Leave points

• Personalisation improves relevance by consuming information about the customer.
• The same mechanism that creates usefulness can create intrusion.
• Surprise is a powerful predictor of discomfort.
• Consent, control and proportionality matter beyond legal compliance.
• Trust can be lost even when a campaign performs well.