Bram De VosA personalized customer experience adapts what a customer sees, receives or has to do, using relevant information about that customer, so that the next interaction takes less effort or lands with more relevance. The richest information for it is declared: what customers told you in surveys, complaints and conversations. Most companies chase personalisation with data they bought and skip what their own customers volunteered. A familiar café does the opposite. It remembers what you order and stops offering what you already declined, using nothing but what it learned in the relationship. This article makes the case for doing it the café's way: suppression rules first, referenced follow-up second, visible change third, and the recommendation engine only once those three run reliably.
Key takeaways:
- The perception gap is wide and current: 52% of consumers say brand experiences still feel impersonal, and only 45% feel understood by the brands they deal with.
- Personalisation operates at three levels: segment, context and individual. Segment and context rules create value sooner than individual treatment.
- The richest personalisation data is declared: what customers told you in surveys, complaints and conversations. Start with suppression, referenced follow-up and visible change.
- The explainability test decides trust: would you be comfortable telling the customer how you knew?
- A first name above generic content proves automation, nothing more.
The gap between what brands spend and what customers feel
The personalisation gap is the distance between what companies believe they deliver and what customers say they receive, and it has not closed despite years of investment. SAP Emarsys found 52% of consumers still describe brand experiences as impersonal, even though 68% say they are more likely to stay loyal to brands that tailor experiences to their needs. Only 32% of US marketers describe their own campaigns as "truly personalized".
Twilio's 2025 research across 7,640 consumers in 18 countries reaches the same place from the customer's side. Only 45% of consumers feel understood by the brands they interact with, a slight drop on the year before. Meanwhile 96% of the companies surveyed say AI is improving their customer-facing operations.
The gap has a simple explanation. Most of what passes for personalisation is a first name in a subject line, pasted above a message that proves the company remembers nothing.
What a personalized customer experience is
Personalisation means adapting an experience using relevant information about the customer, with less effort or greater relevance as the result. It operates at three levels:
- Segment level. Different journeys for different groups: the promises per segment your customer experience strategy defined, delivered differently for the new customer, the loyal one, the high-value one.
- Context level. The experience adapting to the moment: what the customer is doing right now, and what just happened to them.
- Individual level. The café version: this customer, their history, their stated preferences.
One concrete example per level makes the bar visible. Segment: a bank's onboarding flow for a first-time account holder explains the app; the same flow for a customer switching from a competitor imports standing orders instead. Context: the airline app opening on the boarding pass on travel day, and on the booking engine every other day. Individual: the support agent who opens with "I see the replacement arrived Tuesday, is it holding up?" because the history was on screen before the call connected. None of these mention the customer's name; all of them feel personal, which is rather the point.
Start by acting on what customers told you
Declared feedback is the personalisation data most companies already own and least often use. Surveys and support conversations contain information customers volunteered about their needs and frustrations. That information is routinely ignored in the next interaction: a customer with an unresolved complaint still receives the routine sales campaign. There are three straightforward ways to use declared feedback:
- Suppression. The simplest rule with the largest effect: no promotional contact to a customer with an unresolved complaint or a fresh detractor score. Route those signals from your voice of customer programme into the campaign tooling.
- Referenced follow-up. When a customer told you what went wrong, the next contact should show you know. Closing the loop personally beats any recommendation engine for felt personalisation, and automated close-the-loop routing makes sure the right person gets the chance.
- Visible change. "You said, we changed" is personalisation at segment scale: customers recognising their own feedback in what the company did.
None of the three requires new data. All three require the feedback to be classified and connected to the customer record. That is what ISAAC, the AI engine inside the Hello Customer voice of customer platform, does with every open answer as it arrives. The 2026 CX Maturity Study, run by Loyalty Group with Hello Customer among 203 CX managers, shows where that connection usually breaks. 37% of the CX managers surveyed name access to structured, relevant data as one of their biggest obstacles, and 37% point at inadequate IT systems. The feedback exists; the join to the customer record is what is missing.
Segment and context, before individual
Segment and context rules create value sooner than individual treatment, even though individual treatment attracts the attention. Use the metadata from your customer feedback analysis, such as country, tenure, value tier and product line, to understand where needs differ. Then adapt the journey where those differences are meaningful. Context prevents avoidable mistakes: a renewal message can acknowledge a recent outage, while onboarding can skip steps the customer has already completed.
In B2B, the same logic works at account level, and the stakes per account are higher. Personalisation there is the review meeting prepared from the account's own feedback and usage rather than a generic deck. The renewal proposal addresses the two complaints the account logged this year, unprompted. The alert tells the account manager that the tone in support tickets hardened last quarter. All of it adds up to the account feeling known, which is what enterprise buyers mean when they score suppliers on "partnership".
Where generative AI fits
Generative AI has changed what is feasible at each level of personalisation. McKinsey describes it as the ability to create and scale relevant messages and experiences, with tailored tone and imagery, at high volume and speed. The capability is real, and it sharpens the underlying question: relevance at volume still depends entirely on knowing something true about the customer.
The CX Maturity Study shows how early that work is. 15% of companies do not use AI in customer experience at all. 47% experiment in isolated cases, 20% use it to drive efficiency, and 10% have made it part of how they design and deliver experiences.
The report's authors, Mikkel Korntved and Sara Landin Riis of Loyalty Group, put the warning in one line that applies directly to personalisation: "AI can scale a good customer experience, but just as easily a bad one." Their example is the chatbot on top of a poorly understood customer journey, which only frustrates the customer faster. Gen AI scales the message; declared feedback supplies the truth.
Keep personalisation explainable
Personalisation built on data the customer knowingly gave you reads as service; personalisation built on data they did not know you had reads as surveillance. The café owner knows things because you told him, in his café, knowingly. That is the whole test.
Customers are explicit about the trade. In Twilio's 2025 research, 61% of consumers do not believe brands use their data in their best interest, and 84% want control over their personalisation settings. 54% want to know when they are talking to AI rather than a human. Transparency is part of the product.
A practical version for every campaign and journey rule: would you be comfortable telling the customer how you knew? If the answer needs a lawyer, redesign the rule. Declared data is a strong starting point, and it still needs appropriate governance, retention rules and access controls.
Five rules to implement this quarter
Personalisation programmes stall on ambition. These five rules are small, declared-data-only, and each one is worth shipping on its own.
| Rule | What it does | Data needed | How to measure it |
|---|---|---|---|
| 1. Open-complaint suppression | No promotional sends while a complaint or detractor follow-up is open | Complaint status and survey scores in the campaign tool | Unsubscribe and escalation rates in the suppressed group |
| 2. Detractor quarantine | After a low score, the next contact is about the problem, for 30 days or until the loop closes | Detractor flags from the survey tool | Retention of contacted detractors versus uncontacted |
| 3. Outage acknowledgement | Any service message within days of a known incident opens by acknowledging it | Incident log matched to affected customers | Reply sentiment on those sends |
| 4. Repeat-purchase respect | Never promote to a customer the product they bought last week at a lower price | Order history joined to campaign audiences | Share of campaigns checked against the rule |
| 5. "You said, we changed" note | Once a quarter, tell the segment whose feedback drove a fix what changed | The topic-to-fix log from the feedback programme | Response rate of that segment in the next survey wave |
Every one of these runs on information customers knowingly gave you, passes the explainability test without a lawyer, and produces its own before-and-after. Ship the first two this quarter and you will have done more felt personalisation than another year of subject-line tokens.
Four common personalisation mistakes
These four personalisation mistakes come up in most programmes, and none of them needs a bigger budget to fix.
- Cosmetic personalisation. A name above generic content signals automation rather than relevance.
- Personalising the message but not the experience. Tailored copy does little if it leads into the same inflexible process; that process is what customer experience design changes.
- Buying inferred data while ignoring declared data. Start with what customers have directly told you and what you can responsibly observe in the existing relationship.
- Missing suppression rules. Sometimes the most relevant action is to send nothing, especially during an unresolved complaint.
Frequently asked questions
What is a personalized customer experience?
A personalized customer experience is one that adapts to relevant information about a customer's segment, current context or history. It should reduce effort or make the interaction more relevant, rather than merely changing the wording.
What are examples of a personalized customer experience?
Examples of a personalized customer experience include suppressing promotions to customers with open complaints and follow-up that references what the customer reported. Others are onboarding that adapts to progress, offers matched to actual usage, and service agents who see the history before the customer repeats it.
Do you need AI for personalisation?
No. AI helps classify feedback, detect patterns and operate rules across large customer bases. Simple improvements such as suppressing a campaign during an open complaint can be implemented without sophisticated models. In the 2026 CX Maturity Study, 62% of companies either do not use AI in customer experience at all or only experiment with it in isolated cases.
How do you measure whether personalisation works?
Measure personalisation the way you measure every other experience investment: by cohort. Compare satisfaction, retention and spend between customers inside and outside the personalised journey, and per rule. The method is the same as in the guide to customer experience ROI.
How does personalisation relate to privacy?
Personalisation should use data for a clear purpose and in a way customers can reasonably understand. Declared and first-party data are sensible places to start, and transparency does not replace a proper privacy and legal review.
Start with the data customers gave you
Remember what the customer has already told you, respond to the situation they are in now, and know when not to contact them. Those three usually feel more personal than an elaborate recommendation that misses the situation. Sequence the work accordingly: suppression rules this quarter, referenced follow-up next, segment journeys after that. Leave the individual-level ambitions until the first three run reliably, because a recommendation engine on top of an ignored complaint is personalisation's most expensive way to look foolish. More than half of the consumers SAP Emarsys surveyed still describe their brand experiences as impersonal. The fastest way out of that half is to finally use the data your customers handed you themselves.