Hyper-personalisation shares its core purposes and strategies with the traditional landscape of personalisation. The differences are efficiency, scale and ambition. Insofar as there’s agreement on a shared definition, hyper-personalisation is characterised by:
- Machine learning and AI-driven semantic understanding of content and user intent
- Massive data pools to mine for intent and interest signals
- Real-time data processing at scale to generate personalised experiences on demand
In short, it’s about being able to address single-person audiences, to segment to the level of the individual. The worry is that users will resist this level of individualisation, perceiving the intrusive eye of the panopticon. But if you build a user-centred personalisation strategy, giving voice to users’ needs and intents, you can reach what Gebremeskel and de Vries call the “precarious balance where users’ differences in interests and tastes are satisfied, where they are neither “overloaded” with content they are not interested in, nor are they served with content more differentiated than necessary”.
We want to show you that hyper-personalisation will work for both you and your users, and we’ve pulled together three articles to help explain how:
- In the previous article, we discussed the landscape of personalisation
- In this article, we’ll define the destination, a manifesto for what we believe hyper-personalisation should deliver both for users and organisations
- Finally, we’ll give directions for how you might navigate your team towards a hyper-personalisation solution
So far, we’ve only interpreted personalisation, in various ways. The point, however, is to change it. Here’s our manifesto.
User-centric hyper-personalisation manifesto
“Manifesto” is a pretty bombastic word to bring to an article about digital marketing, but we’re already stuck with calling this thing “hyper-personalisation” so that ship has clearly already sailed. At least we’re sticking short of calling it a megamanifesto.
Different people have different needs and objectives here, so we’ve organised the principles by user groups.
For content creators
- Use natural content structures and creation processes
Content creators should be able to write normally, to focus on originality and insight rather than discoverability and machine understanding.
- Automate repetitive generation of content variants
Content creators’ time is valuable, they should not be wasting it generating minor variations of genuine insight to appeal to different audiences.
- Avoid manual content classification in large libraries
Categorising content and identifying content relationships is time-consuming, and the effort grows exponentially as the content library grows, leading to failures in consistency across teams and time that destroy the value of the taxonomy.
For publishers and marketers
- Maximise our return on our investment in content
Both creating content and finding an audience for it are expensive – if the content doesn’t find the right audience, that’s doubly expensive
- Don’t discard traditional marketing personalisation
Traditional personalisation capabilities, including push marketing audience building and pull marketing content variation remain essential techniques to connect users to content, services and products. Marketers are human too, and need human scales and mental models to address their customers.
- Treat strangers as people too
No user arrives with a fully formed user profile. Personalisation needs to be able to greet strangers warmly, apply learnings from users who started from a similar place, and offer unintrusive routes for incorrect assumptions to be identified and corrected.
For experience designers
- Avoid low-level manual rule building
Experience creators should not have to define interpretations for hundreds of signals for each message or journey they want to create, but should be able to define interpretation strategies for systems to apply.
- Avoid black boxes
Don’t introduce systems whose workings cannot be influenced, even if this seems like the simplest route to avoid manual specification.
- Always provide a next best action or next best content for this user at this time
Every time the user’s next step isn’t obvious because the obvious next step doesn’t match their own needs, there is a danger that they disengage. Allow them to navigate back a step or up a level, but don’t make it the only relevant option.
For data teams
- Provide intelligent reporting and feedback loops
Provide tracking data out of all systems to show the effectiveness of both content and the personalised delivery of that content, and allow for reporting back to teams to improve delivery.
- Provide back-end data out of algorithmic and AI systems
Provide content analysis and user analysis data for use in contexts outside personalisation, or to support new personalisation channels. Data analysis and content understanding have value outside of the algorithms creating and using them.
For technical implementers
- Provide graph-based interfaces to data and content
REST APIs are not an effective way to query knowledge, content or product graphs for individual users' needs. Platforms contributing to personalisation should provide GraphQL interfaces.
- Composability is king
Being able to connect personalisation systems to content delivery systems is the key to connecting users with the right content. This is particularly important to orchestrating each user’s personal experience across different experience channels.
- NLP and LLM AIs are the connective tissue for user intentions and content
We will use machine understanding of language to understand the intentions behind user interactions and requests, and to understand our content libraries, as these provide for semantic understanding and emergent delivery properties in a way that rule-based mechanisms do not.
- Contextual multi-channel delivery by default
As content variations and access pathways multiply to meet individual requirements, the need to avoid replicating content repositories and associations across every channel intensifies, to ensure that each user’s individual experience remains consistent at every touchpoint.
- Don’t over simplify your content classifications and relationships
Content has complexity and nuances which simple taxonomies designed for human classification fail to capture. Whilst the value in individual detailed content item relationships may only matter to a few users, to those users they may be all-important.
- Act upon all relevant signals
Every time it’s obvious that a personalised delivery has overlooked something important about a user, it resets the sense that they’re in a relationship to a sense that they’re a target market.
- But don’t expose sensitive information in unsafe contexts
Personalisation may amount to an information leak for vulnerable users: signals and deliveries need to be risk-assessed so they don’t inadvertently expose user secrets to potential abusers.
- Act on intent rather than expressions
Understand what users mean to say, and don’t force them to express themselves in your terms.
- Anticipate what users want before they have to express it
Every time a user has to think about what they need from you next, they have to think about interfaces instead of experience.
- Understand and act upon users’ progression
Users' needs change. Sometimes this will be because of their direct content or product experience, sometimes because they have learnt or developed new needs, and sometimes because the world around them has changed. Don’t allow historical signals to overpower future needs.
- Enhance discovery processes
Help users to discover what’s available to them, but don’t put me on tracks. Feeling like their choices are being removed rather than expanded will cause users to resent and try to escape personalisation.
- Social context is important
Users will test if they’ve been set on tracks by comparing their experience to other peoples’. And they will use other people’s experience to test the value of what they’ve been offered. Make it fair.
- Don’t pretend that machines are humans
Users understand that they don’t really have a one-to-one relationship with a brand or service. It’s enough that their experiences are better. Pretending that personalisation is something it’s not pushing it into a relationship uncanny valley and makes it creepy.
- Keep users in control of their experience
Ultimately, users want personalisation when it works for them. And they are the best judges of what works for them. If they don’t feel in control of their experience, their only remaining option may be to disengage.
- Delivering on promises
It’s one thing to say what hyper-personalisation should (and shouldn’t) promise to everyone involved in the process. It’s another thing to deliver on those promises. And we promised that this would be attainable.
The five key pillars
There are five key pillars you need in place to successfully create and deliver a hyper-personalisation strategy.
- Technical capability
- Signals understanding
- Delivery design
- Iteration and refinement
In our next article, we’ll show how to set up those pillars.
In the meantime, if you think you and your organisation need a helping hand in defining a personalisation strategy and then delivering on it, get in touch.