What real-world service can teach us about digital personalisation

7 minute read ● Summarise this post with ChatGPT

When you experience good service in a restaurant, a shop... anywhere, certain things tend to be consistent. Staff quickly assess what you need, adapt their approach, and know when to step in and when to step back. No interrupting, no overreaching. Just reading the situation and responding well.

Personalisation on websites and apps works in much the same way. It's about creating an experience that feels relevant and reduces effort, anticipating what someone might need before they have to ask.

The technology to do this is more available now than ever. But digital personalisation can still feel either too generic or oddly intrusive. That's rarely a technology problem. It's about instinct - knowing when to act on what you know about someone, how much to infer from limited signals, and when doing less is actually the better move.

Real-world service has been developing this instinct for a long time, and the same principles translate well to digital.

Contextual personalisation

Contextual personalisation means adapting the experience to a person's immediate situation.

In a restaurant, a server might notice a couple celebrating a birthday and offer a complimentary dessert. It's a small act, but it's personalised to the moment, with no prior knowledge of the customer required.

Digital personalisation can do the same. An online clothing store might detect that a visitor is browsing from a particular region on a cold winter evening. Even without knowing who that person is, the site could surface next-day delivery options for their area and prioritise warm seasonal items on the landing page.

None of this requires knowing who someone is. It just requires responding to the context they arrive in, without overreaching on limited information.

Implicit personalisation

Implicit personalisation learns from behaviour over time, without relying on anything the person has explicitly told you.

There's a record shop I often go to. Over time, the owner has noticed the bands and genres I tend to look for. Now when I walk in, he often pulls out a few new arrivals he thinks I'll like. The recommendations aren't based on anything I told him that day. They're based on what he's picked up from my past visits.

The same principle applies online. A visitor to an online record store might browse several pages of indie rock records without buying anything or stating a preference. The site can infer that interest from their behaviour and start surfacing similar artists or relevant new releases the next time they visit.

The risk with implicit personalisation is getting it wrong. There's a real difference between an experience that feels thoughtfully tailored and one that feels like you're being watched. Good implicit personalisation sits firmly in the first camp, like a helpful shop owner making a well-timed suggestion, not an experience that's trying a little too hard.

Explicit personalisation

Explicit personalisation is more direct. Instead of inferring preferences from behaviour, you simply ask and the experience adapts around the answer.

Think of a specialist bike shop. When you're looking for a new bike, the salesperson might ask about your budget and whether you ride on or off road. Based on your answers, they point you toward options that actually fit what you need. The personalisation comes directly from what you've said, not from assumption or observation.

Websites and apps do the same. A streaming service might ask what genres you enjoy when you sign up. A news site might ask which topics interest you. That information then shapes what you see from that point on.

It's a more transparent form of personalisation. You know you're sharing information, and you can usually see how it's being used. When done well, it's a clear value exchange - you share a little, and in return the experience becomes more relevant.

Connecting information when logged in

Personalisation becomes even more powerful when someone is known, when they log in and can be recognised across visits. At that point, contextual, implicit, and explicit signals can all work together.

In the real-world, my financial adviser is a good example. Because they know me and have a full picture of my finances (income, spending habits, savings goals, mortgage commitments) their advice takes all of it into account. They're not just reacting to a question in isolation. They're responding to the bigger picture.

A banking app works the same way. When you log in, it can combine account balances, recent transactions, upcoming payments, and spending patterns to do something genuinely useful. It might remind you that a mortgage payment is due, or highlight a product that fits your current situation. Because it knows who you are, it can connect the dots.

That's the real value of logged-in data. All the different types of personalisation work in harmony, creating an experience that feels designed for that specific person.

When personalisation becomes predictive

Once identity and history are connected, personalisation can go a step further, anticipating needs before they're clearly expressed.

Predictive systems analyse patterns in past behaviour to estimate what someone is likely to want or need next. An online retailer might predict when a customer is due to reorder a product. A banking app might flag a cash-flow issue before it actually occurs.

It's the difference between a good shop assistant and an exceptional one. Both respond well to what's in front of them. The exceptional one has also quietly noticed that you tend to come in around the same time each month, and has already set something aside.

But judgement still matters. Predictions should support the experience, not dominate it. Not every prediction is worth making, and the ones that are should sit alongside context, behaviour, and stated preferences rather than override them.

Knowing when not to personalise

Greater capability doesn't always mean greater personalisation. Knowing when to hold back is just as important.

A bookshop assistant who follows a customer around suggesting titles based on the first book they picked up is trying to be helpful, but it wouldn't feel that way.

Personalisation earns its place when it reduces effort for someone who already knows what they want. It causes problems when it narrows options for someone who's still exploring, or introduces unpredictability where someone actually wants consistency.

A first-time visitor is usually still working out what a product offers. Acting heavily on early signals (hiding content, reordering navigation, making firm assumptions about intent) risks serving someone a version of the experience that reflects their initial exploration rather than any settled preference. The signals are too fresh to trust that much.

The same applies after a gap in usage. Someone returning to an app after several months may not want to pick up exactly where they left off. Treating outdated behavioural data as current produces an experience that feels off. Restraint isn't a failure of personalisation, it's part of what makes it trustworthy.

The instinct behind good service

Personalisation isn't about doing everything you can with every signal. It's about reading the situation and responding appropriately, respecting the person's context, history, and preferences.

In the real world, good service relies on instinct. Staff learn to recognise the signals in front of them and use judgement to decide how to respond. Digital personalisation works best when it follows the same principle.

The brands that get this right don't treat personalisation as a technical capability. They treat it as an expression of good service. And the experiences that feel truly personal are rarely the ones trying the hardest. They're the ones paying the closest attention.