Boosting Website Personalization Through Data Science
Almost 20 years ago Jeff Bezos predicted the rise of digital marketing and introduced concepts like “just in time inventory” that was revolutionary at the moment but have since become a common practice. In the same way, Amazon is continuing to revolutionize e-commerce by being as customer-centric as possible. This is another of Bezos’ ideas, voiced back in 1998 when he described the concept of a personalized store, tailor-made to the client. At this time, Bezos imagined a shopping referral service for Amazon’s 4.5 million customers, essentially 4.5 million stores catering to just as many people. It was only a matter of time and finding the right technical solutions to achieve this vision.
It became possible by looking closely at data and connecting it with the robot-portrait of the person who created it, then generalizing the result. Right now, data is more prevalent than ever, opening limitless personalization opportunities and raising computational difficulties at the same time.
Creating a Personalized Experience
A buyer persona is a marketing concept that helps to segment the audience. It offers an easy, homogeneous approach to market segments and is described in terms so simple that you can almost feel that you know someone like that. It goes well beyond demographics and behaviors, it creates a character and encourages storytelling by the brand.
- Market Segmentation
Personalization begins with segmentation. A good market share is technically identifiable (by IP for example), homogeneous, and relevant from a business perspective. Examples of excellent segments include activity sectors in a B2B setting or a segmentation of customers by marital status or country of origin. It is also common to segment clients by their point along the sales funnel. When defining a segment, it should be narrow enough to require a different experience, but large enough to justify the trouble. Clustering is the technique used on data to identify segments for which further experiences will be designed. Resist the temptation to put in many traits and filters that you create such a fragmented landscape that is hard to keep track of.
- Is there a pattern here?
Each segment will have their own set of values which define behaviors and ways to interact with the website. The good news is that personalization is mainly based on patterns and repetitive actions, therefore once you know a group, it is easy to anticipate their needs, desires, and likes. Always test assumptions and make changes to the website accordingly, never use intuition as your primary means of decision. This is because, in the end, the goal is to automate the process completely.
Creating a customizable website
Once you have the segments, you need to bring them together in a unitary framework that will have different flavors. For this to work, the design needs to be laid down as containers to be filled with appropriate content for each segment. The way the placeholders will be filled can be rule-based like in traditional programming (if-then-else cycles), it could be powered by machine learning or most likely, represent a combination of the two.
While rule-based personalization is ideal for the home page or initial interaction with the website, as the user diversifies their actions on the site it becomes too complicated to keep track of their movements, and deep learning must replace decision rules.
Using machine learning offers another advantage: as the knowledge base or the offer grows, you only need to have your category tagging in place and done correctly, and you can rely on that for a good recommendation.
Apply this logic to any part of the website: display FAQ relevant to each segment, pull those articles out of your blog’s archive that could make the client remain a little longer on your site or make them move towards the end of the sales funnel.
Making sense of data
What about the actual implementation and challenges? Gathering, storing, cleaning and securing the data pose enough problems to require adequate treatment.
- Collecting data
Not all data is valuable or offer answers to your company’s problems. Be sure to create a strategy in gathering your data. Keeping in mind that audiences can include millions of visitors, every piece of collected info should serve a clear purpose, or it is only wasting storage space and computational power. Data science consultants advise preprocessing the received data, especially when dealing with information retrieved from social media. This can help speed up the analysis process and decrease the storage necessity. Most free form data contains significant amounts of noise which needs to be filtered out before using it.
- First-time visitors
To personalize the experience of first-time visitors, the best way is to retrieve third party data and create a profile that dynamically changes the appearance and behavior of the website to match the associated segment. This is usually done through IP and the cookies already stored in the visitor’s browser. Always have in place a generic, best-performing variant of the website in case the guest chooses not to disclose any information. Alert the user of the benefits they can collect if they decide to be less secretive or even create an account.
- Returning Visitors
Once the initial relationship is established, use to the maximum all the previously collected info to customize the experience. Incorporate that in every element, without becoming creepy, just pleasant and welcoming.
- Security issues
Security, data privacy, and protection of personal data remain important problems to be addressed. The solution proposed by some, to encrypt all communication end-to-end is costly, hardly feasible and potentially more dangerous as it can help conceal terrorist acts. A compromise solution is most reasonable.
Erasing the borders in marketing?
Increased personalization leads to an unexpected phenomenon of border dissolution. The rules of B2C are entering the B2B environment and helping companies create a more relevant experience, since, even if you are addressing the needs of a company, the employees are still users that react on an emotional level.
Also, data science is helping companies erase borders through an omnichannel approach that creates a seamless transfer from one device to the other and even extends to the offline approach. The most important lesson to be learned here is that the era of one size fits all is long gone.