Customisation and Personalisation Part II
Personalisation and customisation hide some other pitfalls than those already mentioned in the first part. We also need to set realistic expectations in accordance with client information we will be actually 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" our previous article on 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.
There Are No Limits to Personalisation
However, it is important to realise that you can personalise in effect anything, not just the static content of your website or some particular information you provide to visitors, but also the products themselves! And here often lies the stumbling block, as product portfolios tend to be rather rigid. We are obviously not talking about products with clearly defined and fixed inputs and quite unified output (although even this could be fairly easily contradicted thanks to, for example, the automotive industry where it is possible to choose from a quite wide portfolio of extra equipment and the combination of these extras reach hundreds or even thousands of variants). However, in the world of services, the providers are limited primarily by their imagination and the ability of their systems to process on the basis of different inputs as well as high granularity of outputs.
And What About Results?
Understandably, we cannot disclose the clients in the following examples, however, we would like to give you at least a general idea of the possible impact of content personalisation in its very simple form so that those of you who have no experience can imagine relations, in which these operations can be used and their impacts can be seen.
The first example is the personalisation of the main carousel banner (automatically rotating promotional content) on the main web page of a major FMCG distributor. The starting point for personalisation was prior persona definition, or website visitor segmentation, based on customer needs information. The purpose of personalisation was to improve the carousel click-through rate and subsequent move to the corresponding section or seasonal offer thanks to a more relevant offer. The result was an increase in click-through rate (CTR) for all groups of personalised banners by 5-10%. With the same client, another intervention has been designed to make it easier to navigate through the website (and thus easier conversion). The starting point here was visitor behaviour on the web page and their passage through the registration funnel, where the first step of the form was removed for returning visitors. A user who had already visited the registration funnel without actually completing the registration had a better starting position during their subsequent visit - they did not have to go through a summary of the benefits (which they have already seen) and were directed straight to the final conversion. The result was an increase in the conversion rate for a personalised variant by several percentage units, which is a significant improvement in the view of customer lifetime value.
Another easy option is personalised e-mailing, which is nowadays probably the most popular implementation due to frequent use of own client database (with more detailed information that can be effectively segmented). In this case, a client with tourism background used the main division of the database according to client's solvency and their previous interests in particular destinations in order to establish a truly personalised e-mail communication with the purpose of increasing the value and frequency of orders coming from regular newsletters. Compared to the initial distribution variant, which was based on seasonality and trends, the customised versions took into account also previous orders and resulted primarily in a much more targeted proposal that corresponded to the customer's history, and secondly, a much better control over the investment/profit ratio (i.e. it is better to sell one exotic destination rather than two European ones). The final impact was not only an improvement in campaign results by tens of per cent, but also a significant increase in campaign returns.
The example above was a case of a more vigorous intervention beyond the displayed content and achieved through personalised pricing - this is one of the areas where personalisation has been recently most frequently used (as it should be). As it seems worthwhile in the long run to sometimes offer a dumping (lower than normal value) price to certain (pre-selected) customer segments. In this case, thanks to campaign information (the user came from a competing campaign whose aim was to steal customers) and repeated web access, we offered a significantly lower price than our competitors in order to attract customers and increase our market share (penetration). It is extremely important to work with the real campaign objective (in this case, to increase the number of orders in the segment demanded by competing products). If the campaign objective were the highest profitability, it would have been necessary to closely monitor the impact of the discount on the final cost (cost per action) and even leave some orders alone.
Technology Is Your Friend
Virtually none of these scenarios can be put into practice without a sufficient technological "chassis" on which you could move your personalisation "vehicle" into action. With the exception of the last example, however, the above were all offline operations that can be performed over a familiar set of user information, and it may be the case that most of it did not even surprise you. The truly interesting scenarios happen when we try to process online user data, as is the case of price-making. But let's take a look at some intermediate steps first before exploring some really ambitious projects.
One of them is a state of the art form on the domestic market, used by T-Mobile website for contact with its customer centre. As soon as you enter it (the link is here), the form begins immediately, in real time, analysing inputs in all fields and evaluating the possible causes for the problem the user trying to resolve. The aim is to help the customer before the form is submitted, thus decreasing the number of requests and consequently reducing the strain on the client centre.
There Is More Than Enough Public Information
If we go even further, we can collect customer information in the style of digital onboarding using publicly available registers, such as trade register, invalid documents register of MVČR, or apps ARES. All of them naturally require some sort of user access in the form of personal identification, social security number or ID card number. However, inserting such data enables us to work with a much wider volume of information for creating a bespoke offer for our client. But what if we want to be even more effective and faster?
From Size to Speed
Today’s big buzzword is not big data any more as those are associated with mostly offline operations, but rather so-called fast data that are processed online in real-time (the proper term is actually near real-time, as there is always a minimal delay). Thanks to possibilities enabled not only by paid solutions but also by open-source technologies such as Apache Kafka, or Apache Spark it is possible to process not only large volumes of offline data but also a large number of online inputs where primarily human imagination sets the limits (excluding the budget, that is). There are many potential scenarios that people can come up with, but let us mention a recent one which happened when my colleagues and I set out to the mountains and most of us forgot to arrange a travel insurance in advance - why should this factor not affect the price? Aren’t so-called early birds always rewarded for being so organised? Shouldn’t the client's current situation affect pricing? Isn’t a user who is systematic and hands-on way before departure less risky than someone who just about recalls insurance at the moment of buying a lift pass?
Let's not be afraid to give free rein to one’s imagination. Technology always serves some purpose or idea - if it’s not needed, there is no reason to spread yourself thin.
Purchase Ahead Of Demand?
With reference to already mentioned Article on PSD2) we discussed the possibility of accessing transaction history of a given user for merchants. This access will become a very valuable asset as it uncovers further patterns in buying behaviour for already extensive machine learning interfaces. Those are currently capable of estimating what you need and when and in the near future they will be able to carry out fully automated purchases for at least a narrow circle of aficionados. For a long time, Amazon's algorithms have been able to predict what you'd like to buy even before you submit your order (see link here).
Logically, the more inputs, the more accurate the models, and the wider the portfolio of services affected.
Smartness Is Strength
Imagine an application connected to the online electricity exchange capable of optimising energy purchases for different days of the week based on online information from your smart household - it will know when you usually do your laundry, when you have the most time to watch TV, or when not to bother during the weekends, because you normally spend three out of four weekends elsewhere. Imagine as well that by providing just a few basic personal details, you’ll be able to arrange a mortgage online within 15 minutes. Does that sound absurd to you? Wouldn’t it be great if you simply gave permission to a negotiating application that would check your account balance for the last three years, screen the registry of debtors, and criminal record and, at the same time, it would find out and calculate your mandatory expenses and carried out complete scoring while you are at home, preparing a cup of coffee? And before you take the first sip, you would know that you could get an interest of 2,643% for the upcoming 10 years.
While some cases may sound a bit extreme "Big Brother" style, we need to bear in mind the main objective of personalisation - the greater benefit to the customer. Personalisation should always help the user to get exactly what they want. Ultimately, it should be able to guess what the customer wants before it’s even verbalised, because that is what he or she appreciates and what creates a significant competitive advantage for you. You should also bear in mind that if you are concerned about your privacy from a user’s perspective, there is a good system of counterweight (or checks and balances if we refer to the US political system) that prevents misuse of personal data and reinforces user rights in this dynamic time (yes, it is GDPR). The final recommendation we are going to give you is the following article about the future of predictive artificial intelligence, which is a logical step or the outcome of a complete adoption of customisation - https://hbr.org/2016/11/how-predictive-ai-will-change-shopping.
The Customer Is... Well, Simply a Customer - Human
We are basically returning to the beginning and to the core of our series, which is the customer. A customer is not an abstract entity; it is a human being or countless individuals. It is therefore essential to approach them as individuals, the times when they should be forced to consume are long gone, customers choose. Internalise this premise, focus directly on your customer and build your business around it. The next episode will reveal one of the methods how to achieve this, namely Human Centred Design.