Personalisation can make a digital service more relevant. It can also make a confusing experience more complicated, use weak data with unwarranted confidence and create an expensive stream of content variations that few people see.
The decision should begin with a client problem. Are people struggling because the experience lacks adaptation, or because they cannot find, understand or complete the basics?
If the service page is unclear, the navigation is difficult, the portal task fails or the enquiry route asks for unnecessary effort, dynamic content is unlikely to repair the experience. The source article compared this to fitting heated seats in a car that does not start reliably. The image survives because the sequencing problem is real.
Define personalisation precisely
The term covers very different interventions:
- remembering a user’s stated display preference;
- recommending related content from a known topic;
- adapting a journey to a client’s service or role;
- prompting a relationship manager after a relevant event;
- using behaviour to infer an individual’s interest; and
- generating an experience for one person in real time.
These uses differ in value, data, complexity and risk. A portal remembering a chosen sort order is not equivalent to combining identifiable browsing behaviour with CRM records to trigger sales contact.
For any proposal, state:
- the user need;
- the data and inference involved;
- the experience that changes;
- the simpler alternative;
- the measure of usefulness; and
- the legal, privacy, security and professional review required.
“Personalised experience” is too broad to approve as a capability.
Test the foundation first
Relevance and clarity
Can the intended buyer recognise their problem and understand where the firm can help? Does the content use necessary technical language well? Is evidence connected to the claims?
Serving sector-specific versions of a vague page creates more vague pages. Improve the core proposition before multiplying it.
Findability and task structure
Can people locate priority services, people, proof and contact routes? Can clients complete the important portal tasks without knowing the firm’s internal structure?
Navigation and search should reflect actual user goals. A personalisation rule that hides or reorders content can make an already weak information architecture harder to diagnose.
Performance, accessibility and reliability
Does the experience work across representative devices, connections and assistive technologies? Are important interactions stable and responsive? Does a form deliver its data and confirm the next step?
Personalisation often adds scripts, integrations and variation. Establish a performance and accessibility baseline first, then make the new layer prove that it does not cause material regression.
Content and operational capacity
Who creates, approves, maintains and retires each variation? How will the firm prevent conflicting advice or outdated sector messages?
A platform can make dynamic content technically possible while leaving the marketing team with an unsustainable editorial model. Price the continuing work, not only implementation.
Measurement and data quality
Can the firm identify whether the change helped a meaningful task? Is the underlying data accurate, current, sufficiently complete and legitimately available for this purpose?
Small B2B audiences can make segment results volatile. Qualitative research, repeated task evidence and operational measures may be more useful than a conversion percentage based on a handful of visitors.
Where bounded personalisation may help
Remember an explicit preference
A portal can retain a user’s chosen view, format or notification setting. This reduces repeated effort and gives the person control.
Provide a clear way to change or reset the preference. Ensure it persists only for an appropriate period and does not expose sensitive information to another user of the device or account.
Recommend related information
Good taxonomy can surface material related to the page or service a person is using. This can begin as contextual recommendation without identifying the visitor.
Evaluate whether the recommended item supports the task, adds a useful perspective or merely increases page views. Keep important information accessible outside the recommendation logic.
Adapt by a known role or service
An authenticated client may need a view suited to their role and permissions. A procurement contact, project owner and finance contact can require different actions.
Role-based design should follow service and access needs, not stereotypes. Test exceptions and people holding more than one role.
Prompt a human at a meaningful service moment
An upcoming review, incomplete onboarding step or underused contracted capability may justify a prompt to the relationship team. Present it as context, not a conclusion about intent or churn.
Set frequency, suppression and review rules. Ask whether the contact served the client. A trigger that mainly creates opportunistic sales calls will feel different from one that resolves a known need.
Data can turn relevance into intrusion
Personalisation frequently relies on personal information or inferences. The fact that systems can connect browsing, CRM and service data does not establish that they should.
Define the purpose and minimum data. Establish lawful basis, transparency, retention, access and security with qualified specialists. Consider client confidentiality, information barriers and sector obligations. Do not assume a cookie banner or general privacy notice authorises every reuse.
Inference deserves particular care. Visiting an employment-law page may reflect a client matter, academic research, competitor analysis or chance. Treating that signal as knowledge can create an uncomfortable interaction and a poor decision.
Give people meaningful control where appropriate. Make the unpersonalised experience good enough that withholding optional data does not produce a degraded or punitive service.
Run a fundamentals gate
Before investing in a personalisation platform, bring content, service, analytics, technology, privacy and client perspectives together. Review one priority journey and answer:
- Which observed friction remains unresolved?
- Would one good default serve most users?
- Which variations have a clear need?
- Is the content ready for each variation?
- Is the data fit and appropriate?
- Can the organisation operate the rules over time?
- What would cause the team to stop?
Score evidence quality separately from expected value. A high-value idea with weak evidence may justify a small test, not an enterprise implementation.
The source article described a firm with a sophisticated personalisation tool placed over jargon-heavy pages, difficult navigation, slow mobile performance and a demanding contact form. The delivery mechanism was more advanced than the material and journey it delivered.
Distinction has made a version of this mistake too. We once concentrated on segmentation logic before challenging the underlying content strongly enough. The segments worked; the content remained confusing. That source observation is worth preserving because it demonstrates that sophistication can distract advisers as well as clients.
Invest in the smallest useful layer
Once the foundation is sound, test one adaptation with a specific audience and outcome. Use existing CMS tags, authenticated roles or CRM workflows where they are sufficient. Avoid buying a platform merely to run a use case the current stack can support.
Set a baseline and assess usefulness, operational burden, unequal effects, privacy objections, performance and content maintenance. Compare the personalised version with a strong default. If the default performs as well, keep the simpler service.
Your digital experience is probably worse than you think provides a related fundamentals review.
We also offer a roughly 30-minute assessment of whether a firm’s foundations justify personalisation investment. Distinction can benefit commercially from later platform or experience work, so the assessment should be capable of recommending no personalisation at all.
The useful sequence is not “basic forever, sophisticated later”. It is problem, foundation, bounded adaptation and evidence. Personalisation earns its complexity only when it improves an already coherent service for the people it is designed to help.
