From hotels to digital: why personalisation is the Achilles heel of personalised experiences

Imagine you have booked the holiday of a lifetime as a surprise for somebody you love. First class, beautiful hotel, all the trimmings.

At the airport, an assistant learns that you are celebrating and wishes you a wonderful trip. At the hotel, the guard at the gate has your name. The bag attendant greets you before you explain who you are. The concierge takes you to an upgraded room with a handwritten note and a bottle of wine waiting.

This feels different. It feels personal.

Almost none of it depended on a recommendation algorithm. A team communicated, remembered why the visit mattered and acted on that context.

Personalised and personal are different promises

Mark Schaefer draws a sharp distinction between the two. In his formulation, personalised is cut and paste; personal is unique and reflects an understanding of a human being.

That contrast is deliberately provocative, and the categories overlap in practice. A well-designed personalised service can use data to remove real effort. A personal interaction can still be awkward or intrusive. The useful challenge is to ask whether the experience demonstrates care or merely repeats information about somebody back to them.

Go back to the hotel. The guest's name was not the magic. The information travelled ahead of them, so staff could recognise the occasion and avoid making them start again. The name was a symptom of coordinated care.

Most weak digital personalisation gives you the name without the care.

How personalisation became a strategy

Holding a five-star hotel up as the standard for every digital interaction would be unfair. A hotel has time and staff; a service website may have minutes to help a visitor who has several other demands on their attention.

So firms use affordable, scalable tactics. A first name appears in an email. A site promotes a topic related to an earlier visit. A portal remembers a preference. These choices can be useful when they help somebody continue a task or find relevant information.

The trouble begins when personalisation is promoted from tactic to strategy. The logic becomes: the more data we hold, the more relevant we can appear; therefore collecting and displaying more of it must improve the relationship.

Knowing a person's name and using it is different from understanding what they need and designing the service around that need. One is a data capability. The other requires judgement across the journey.

For a professional-services firm, the difference is easy to see. An email addressed to "James" may still recommend a service the client already buys. A client portal may welcome somebody by name and then ask for information their adviser collected last week. A relationship manager may receive a beautifully scored lead with no visibility of the complaint that person raised yesterday.

The surface is personalised. The service is fragmented.

When personalisation backfires

Personalisation can fail in several ways.

It performs intimacy. Repeating a first name or mentioning a recent action too explicitly can feel as though the organisation is demonstrating what it knows. The customer receives no additional value from the display.

It uses the wrong inference. Behaviour is ambiguous. Reading an article about redundancy, litigation or financial difficulty does not establish somebody's situation or permission to address them as though it does. In high-trust services, a confident wrong assumption can be especially damaging.

It exposes broken memory. The message claims recognition while the next channel loses the context. Nothing makes a "we know you" proposition collapse faster than asking the person to explain everything again.

It optimises a local metric. A targeted message may lift a click measure while increasing unsubscribes, anxiety, complaints or low-quality enquiries. Measure the wider journey and longer relationship, rather than celebrating the easiest response signal.

It consumes attention that belongs on the service. Adding a merge field is easier than redesigning a handoff. One creates a visible marketing feature; the other may remove hours of client effort. Investment follows the visible feature unless the team measures the journey.

It crosses a privacy or trust boundary. Legal compliance is the minimum consideration. Ask whether the use is expected, proportionate and explainable. Sensitive data, inferred characteristics and opaque profiling need particular care and appropriate privacy, compliance and security review.

Design operational personalisation first

Ask a different question: what does this person need at this moment that the service can do for them?

The hotel's effect was operational. Information reached the next colleague with enough context to act. A service firm can translate that principle without pretending every client wants a bespoke digital world.

It might mean:

• a client does not re-explain their situation to the third person they speak to;
• an onboarding form carries forward information the client has already confirmed;
• a relationship owner sees an unresolved issue before making a new offer;
• a portal opens at the task the client needs to complete;
• content reflects the person's role or service relationship without implying a sensitive fact;
• preferences are remembered and easy to change;
• the next colleague knows what was promised and when.

None requires the interface to announce the data being used. The best personalisation may be an absence: no repeated question, irrelevant recommendation or avoidable delay.

Start with a high-friction journey and map where context is created, stored, lost and needed. Include people, policies and systems. Then decide the minimum information that should travel to the next step, who may see it and how long it should remain useful.

This is harder than adding a campaign rule. It is also more likely to improve the experience people remember.

Use a ladder, rather than leaping to prediction

Firms can build personalisation capability in stages.

1. Consistency

Make the same accurate information and service standard available across channels. Fix contradictory content, dead ends and missing ownership. A coherent baseline beats sophisticated targeting on top of disorder.

2. Continuity

Help users resume a task and prevent repeated work. Preserve context across authorised handoffs. Give people control over saved information and clear recovery when it is wrong.

3. Declared preferences

Let clients choose topics, channels, frequency and accessibility needs where appropriate. Declared preference is often easier to explain and govern than inference. Do not ask for information the service will ignore.

4. Contextual relevance

Use known service, role or journey stage to prioritise useful content or actions. Test for wrong assumptions and provide a neutral route when the system cannot be confident.

5. Prediction

Predict a need or next action only when the value justifies the data, model and governance. Define human oversight, monitoring and a route to challenge or correct the result. In regulated services, involve the relevant legal, compliance and risk owners before launch.

Each stage should earn the next. If a firm cannot carry an agreed preference from one channel to another, predictive personalisation is unlikely to solve the underlying service problem.

Measure whether it helped

Click-through rate is insufficient. Define the customer job the personalisation should improve.

Measure completion, time, repeated information, errors, support demand and relevant service outcomes. Review negative signals such as correction, opt-out, complaint and abandonment. Use qualitative research to learn whether the experience felt useful, confusing or intrusive.

Test a non-personalised comparison where that is proportionate. Sometimes clearer information for everybody performs as well as a complex rule. That is a valuable result, because it avoids additional data and operational cost.

Keep a register of active rules, data sources, owners and review dates. Personalisation logic becomes stale as services, roles and client circumstances change. Somebody must be able to explain why an experience appeared and switch it off when it causes harm.

The question for your team

If you switched off every first-name merge field tomorrow, would customers notice a meaningful loss?

If the answer is no, look at what you have been calling personal. Does context travel ahead of the customer? Can the next interaction begin from what the firm already knows and is entitled to use? Does the design remove effort or merely display data?

Choose one interaction where remembered context would remove genuine effort. Test that improvement against a clear, non-personalised version. The better service should win, regardless of which version uses more data.

Worth a conversation. Book a short discovery call with the team at Distinction - no pitch, just an honest look at where your experience feels human and where it doesn't.