Hyper-personalisation first started to appear as a concept a few years ago, largely in articles about Netflix, Spotify and TikTok. It tends to be synonymous with the use of data-driven machine-learning recommender systems to create individualised collections of contents, adverts or products. This is contrasted with personalisation of the non-hyper kind – typically characterised as segmentation-based content variations.
This is both great and problematic simultaneously. It’s great because there is genuinely a set of technologies and techniques that can create individual relationships between users and brands. But it’s problematic because the focus of the conversation is on engines rather than on users, and on technology rather than on experience.
And with those brands as key examples, it’s easy to infer that hyper-personalisation is beyond the reach of most organisations – you need either dedicated development and a bespoke algorithm, or to buy in a black box recommendations engine. Which matches many digital teams’ experience of ordinary non-hyper-personalisation of the ordinary kind, who have often not had the resources to personalise experiences beyond minor copy variations to cater for different segments in a shared conversion path.
We want to show you that hyper-personalisation is within your reach, and we’ve pulled together three articles to help explain how:
- In this article we’ll discuss the landscape of personalisation
- Next, we’ll define the destination, a declaration of 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
The landscape of personalisation can seem fuzzy; let’s try to sharpen the details.
Digital personalisation falls into three main categories. You may have seen these plotted out as tiers on a maturity ladder, as a ladder to climb. In our view, they’re complementary tools, and each has a role to play in a user-centred hyper-personalised experience.
- Trigger-based personalisation
This normally means action-based functionality triggered directly by a user or as part of a process flow initiated by a user, typically tied to a user account.
Managing account details, viewing transaction histories, receiving event notifications, these are all typical functionalities that generate experiences unique to a particular user, but generally as linear responses to triggers.
This typically will require specific implementation work, or purchase in of pre-existing functionality from an experience platform (e.g. an ecommerce management system).
- Rules-based personalisation
This is generally based on manually created classifications of users to create broad segments (or audiences), with corresponding business rules to choose a variant of a content item (or journey, or recommendation set) to display.
A typical use would include segmented email campaigns with variant landing pages designed to appeal to different audiences, or substitution of content variants within a website based on visit characteristics (time of day, location, pages viewed). But it also extends out to ad delivery systems using audience builders, and product recommendation systems built on manually defined SKU-relationships like bundles.
Each variant is normally manually crafted to match the segment, and consequently digital teams will only create a few different segments.
This normally means a recommender engine built on a knowledge graph automatically generated by a machine learning system.
Personalised content viewing recommendations and product recommendations on large ecommerce sites are typically built using this approach.
These are often either bespoke, or black box (you can see the results but can’t influence how they’re constructed).
Most personalisation patterns rely on interest signals to match users to content. Historically, we’ve tended to classify these into three types of signal:
- Personal information
Who the user is, including identifiable information that requires the highest level of privacy protection.
Counter-intuitively this has been the least useful information for most rules-based personalisation patterns because it’s too individual to use for pattern matching.
- Disclosed information
Non-identifying information that either the user has explicitly told us, or that is drawn from other authoritative sources such as business systems, including preferences, campaign responses, purchases and other deep engagement.
This typically requires login or other authentication to match to the specific user, but this also, and requires you to already have a deep engagement with the user.
- Deduced information
Intelligence we gather about the user based on how they’re behaving on your digital channels – also known as behavioural signals.
These signals are available from anonymous users, so do not need an existing deep engagement, and can be built up iteratively over multiple visits (assuming the user provides tracking consent).
However, it’s difficult to build a co-ordinated personalised experience purely through deduced information – cross-device, cross-channel and lost tracking factors create a strong recency bias to the user’s current interaction, which doesn’t reflect their long-term relationship or progression with your brand, and often results in seemingly lazy calls to simply repeat the most recent behaviour, which may even be a one-time behaviour like a large purchase.
There are some typical components and techniques for delivering personalised experiences:
- Personalised messaging – effectively the mail merge of personalisation, this is interpolating personal or disclosed information into messaging. This could be as simple as, “Welcome back, [insert name here]”
- Variant messaging – altering the copy, imagery or other presentation behind an otherwise shared landing experience or call to action to maximise the appeal to certain audiences – like Netflix changing the imagery in its content menus based on what they think is most likely to make them watch each show
- Variant calls to action – customising the action different audiences are encouraged to take based on which is most likely to appeal to them – so some audiences might be encouraged to explore thought leadership content while others might be encouraged to book a demonstration
- Variant journeys – often the domain of marketing automation, with branching journeys navigating the user through an engagement journey mostly dependent on their previous actions in the mapped journey
- Content or product recommendations – predicting what the user is most likely to be interested in next, or as an alternative to their current selection, and showing the options that are the best match in a “Next up for you” or similar feature
Personalisation capabilities aren’t a function of just one type of digital platform. Depending on your solutions architecture, you’re likely to have several platforms delivering some form of personalisation, either as subcomponents working together, or independent platforms working separately. These might include:
- Marketing automation platforms (e.g. Marketo, Eloqua, Pardot)
- Web experimentation and personalisation platforms (e.g. Optimizely Experimentation, AB Tasty, Webtrends Optimize)
- Digital Experience Platforms (e.g. Optimizely One, Acquia, Sitecore)
- Recommendations platforms (e.g. Optimizely Content/Product Recommendations, Clerk.io, Nosto)
- Customer Data Platforms (e.g. Blueconic, Dynamics 365 Customer Insights)
- Conversational interfaces AKA AI chatbots (e.g. Salesforce Einstein, Watson Assistant, OpenAI/ChatGPT)
- Cross-channel / dedicated personalisation platforms (e.g. Acoustic Personalisation, Salesforce Marketing Cloud Personalisation, Dynamic Yield)
Customer warning: other categories may exist, platform providers may urgently protest that they fit into more than one of those categories although not necessarily to the same standard as the first category, and just because you have one product with Salesforce or Dynamics in the name doesn’t necessarily mean you have the very specific and confusingly similarly named product that you need to have for the personalisation capability you have in mind.
Putting the hype in hyper-personalisation
McKinsey and Salesforce have both conducted research showing that users want the benefits of personalisation. This is backed up by what comes out of our own workshops and research, too – users are resistant to having control of their experiences taken away or their options circumscribed, but they want progressive journeys and relevant content, and targeted products.
In short, users might resist being bracketed as a persona, but they will embrace being treated as a person, even if it’s coming from a brand rather than a human.
How therefore, do we make personalisation genuinely individual, genuinely useful to users, and genuinely effective in improving business objectives, whilst decreasing the workload it creates for content and experience teams?
We think the answer lies in hyper-personalisation, done right. Find out what we think that looks like in our second article of this personalisation series, here.
If you'd like to discuss how we can put these ideas into action for your brand, get in touch with our team today.
Tap into our latest thinking to discover the newest trends, innovations, and opinions direct from our team.