Customisation and Personalisation Part I
Uniqueness at the Core
We want to treat each and every customer as an individual; create a sense of uniqueness by making him or her feel that we only communicate with that particular person at a given moment - we want to adapt to that person, we want to personalise and customise.
But before rushing into details on how to achieve that, let us go through some basic terminology and explain what lies behind these buzzwords.
One of the key terms is personalisation - it is nothing but a simple recognition of a particular user, their classification into a given segment and subsequent response to those facts. We are talking classic marketing personas that represent diverse lifestyles and enable us to better grasp our customers’ motivation and train of thought. Once our user has been identified and categorised (in an ideal world, we also know that the person visiting our website is our loyal customer Jacob Penfield, born in Milton Keynes, father of two, married for nine years, who commutes to work in his new Škoda Kodiaq) we can further handle the data, and there are two ways to do so.
The first one is general customisation, that is adapting content to the user in question. In Jacob’s case, our website would be recommending an Easter discount on winter tyres for his Škoda because he really was not expecting the third wave of cold weather. The second way is so-called behavioural (re)marketing which evaluates complex behaviour across the internet (as far as the tracking tools can reach) and allows you to customise the displayed content not only based on the user data that we have collected ourselves but also based on a user’s other internet activity. If we go back to the example of Jacob, let’s say he was browsing B&Q’s website (DIY store) a week ago, looking for inspiration before starting his garden house alteration, and few days later a sponsored offer on discounted shutters pops up on his Facebook …and they are spot on!
Initial Steps to Personalisation
In principle, there are two different types of personalisation (actually three, but the third type is a hybrid, a combination of the previous two): explicit or implicit. It is quite helpful to realise the distinction and to choose the one to follow right at the start, as it has a significant impact on the implementation costs. You will soon understand why.
Explicit personalisation primarily uses data on topical interests and needs which are provided directly (explicitly) by the user. Implicit personalisation, on the other hand, intends to imply what might be interesting and relevant to the user based on the data we have at our disposal.
For the basic concept of explicit personalisation, we could consider setting a website navigation in accordance with user’s preferences - we don’t necessarily need to know the customer at all in order to adjust their customer journey based on their needs, or a segment of their own choice. A good example of explicit personalisation based on customer’s needs is AIG insurance website, with a navigational menu right on the homepage, under the main slogan, that clearly lists the most common enquiries and effectively navigates its users to resolving a problem that they came to address. You can naturally save such information to cookies but it is necessary to comply with GDPR (get inspired by one of our previous articles) and try skipping this first step the next time. You can also choose segmentation based on user persona, as seen on Concordia insurance website (yes, insurance companies are a handy source of examples) – in this case, the web user does not state their enquiry in the first step but defines their current life situation instead; the subsequent content will adapt accordingly. Just as in the previous example, we ideally store this information in cookies and work with it during further visits.
Budget Is Key
Processing information about past behaviour takes us to the ground of implicit personalisation. There is a myriad of choices on how to implement this approach, the only limiting factor being the size of budget that you can assign to personalisation - more precisely, how you’ve calculated your return, what shall personalisation bring you and your customers. The more customer data you have collected, the more powerful the implicit personalisation will become. To begin with, we are talking basic recommendations such as "You might also like" when choosing products in an e-shop, selecting a language mutation based on user’s browser settings or web content adaptation based on so-called origin - whether the user accesses the website by typing the address into the search bar or whether they are prompted by a paid advertising campaign.
The More You Know the Better
Implicit personalisation is also "simple" when we work with a logged-in user and, ideally, we know more about them than just a name and email address - such as purchase history, contract portfolio, solvency, payment morale, etc.
However, it gets interesting as the complexity increases - for instance when we can adopt the offer in accordance with previous behaviour on the portal, dynamic product portfolio (tailor-made offers) or cases where we can try to identify even an unregistered user thanks to information collected by advertising tools or based on their previous behaviour on a given gadget. It is a public secret that the last domain is ruled by airline companies - try to initiate booking of the cheapest possible flight on offer to a destination you would potentially like to visit, then abandon your search and check the ticket prices on the same portal fifteen minutes later.
What Do You Need to Measure, How to Manage Settings and How to Evaluate
When preparing your personalisation, it is always crucial to define the metrics (for instance number of orders) the given activity should influence and improve. We should ideally work with pre-defined personalisation scenarios already in the preparation phase. Those scenarios should tell us in a nutshell based on what are we going to personalise, what are we going to personalise and with what purpose. To illustrate on a specific example, the output could be something like this - personalisation will be based on targeting campaign segments and we are going to personalise the banner areas on dedicated landing pages in order to increase the number of orders from the campaign. In the context of an example customer journey, we need to measure and compare evaluation metrics for quality and relevance of traffic (website exit rate, page views, etc.), click-through rate (CTR) of the banner areas on landing page, and especially the conversion rate for the final conversion (order). The true purpose of personalisation in such cases is not to reduce the site's exit rate (the rate of immediate site exit without further interaction), or to improve the click-through rate to the site’s internal pages (both only work as partial metrics that we are trying to influence by personalisation), but to increase the returns on campaigns. The personalisation scenario itself (as its name suggests) is ideally based on definition of the persona who visits the website while it takes into account their (already described) needs.
A relatively common mistake that occurs when evaluating multiple concurrent personalisation activities is to compare personalised variants only against each other, i.e. an evaluation of which personalisation scenario leads to the greatest increase in the observed metric compared to others, regardless of the performance of the original (non-personalised) version. The principle is the same as in A/B testing when we want to introduce a new element to our website in order to improve its performance and we test it in two variants without comparing it with the original website without the new element. That means we find the winning variant of the element (for example, a green button works better than a red one), but chances are the web page actually worked even better without any buttons whatsoever.
Where´s the Hitch?
The procedure above describes a trial-and-error method where we try new approaches and test what impact they will really have and whether the impact is actually desired. With the right approach and based on our experience, we can most likely minimise the risk of failure, for example thanks to appropriate prior testing of a narrower sample of users (see our User Research and Testing). But thoughtless attempts can cause absolutely useless financial waste without any results, which can be a source of great dissatisfaction during your first flirt with personalisation that can result in reluctance to personalise further and more intensely.
Similarly, we need to realistically set expectations based on the information we have about clients who we will be working with. It is quite obvious that if we know only the email address, the cost of extracting it in order to get a full user image will be higher than if the client provides us with their complete personal information, possibly including even their transaction history (yes, this time we "recycle" PSD2). We deliberately don’t say that such information mining is not impossible, however, it is a considerably greater investment into implementation and logics of how to obtain information and how to correctly pair it.