Labs | We believe in continuously improving our solutions with the latest advancements in Data Science & Machine Learning to enhance the added value for our clients. With the series 'Labs', we aim at keeping you up to date about our latest innovations and translate these techniques into the business impact it will make on your organization.
Consumers nowadays are flooded with advertisements in their everyday lives. There is almost no escape from it, in both the online and offline environment. From TV commercials to banners on the web, and from billboards to mailings, brands are heavily competing to grab the attention of the customer. With this overload of ads questions arises. Do consumers even pay attention to all these ads? Do they process and evaluate all the information they come across? The simple answer to these questions is ‘no’ when following straightforward human psychology. Consumers, and we as human beings, have only limited resources to process all the information they come across. To deal with limited resources, consumers only pay attention to information, and in this case advertisements, that could be relevant to themselves.
So, advertisements need to be relevant to grasp the attention of the customer. Being relevant to the customer in the e-commerce environment is based upon two important aspects. Firstly, personalization: does the offer speak to the personality, identity, preferences and tastes of the customer? Secondly, timing: does the offer apply to the near rather than the distant future? Consumers consider advertisements more carefully when they believe they soon are going to make a purchase decision. So, timing is everything when competing for the attention of your customers.
In our previous publication in the series ‘labs’ we elaborated on personalization, the first aspect of relevance. We showed how the optimal subset of product recommendations can be selected based on individual buying intentions for all the available products in the assortment. In this blog we tackle the second aspect of relevance, timing. Timing enables e-commerce players to not only recommend relevant products, but also communicate these recommendations at a time that matches the individual customer journey.
Every customer is unique and has its own buying patterns. Some customers are treating themselves every month with new fashion items, while other customers are only shopping twice a year. If we are able to decipher the buying pattern of the individual customer, recommendations can be communicated to the customer at moments when he or she is paying most attention to advertisements. This will increase the likelihood you will grab the attention of the customer and eventually increase conversions.
This all sounds great, but how are we able to predict the likelihood that someone is going to buy at a specific moment in time. The developments within the Data Science & Machine Learning community never stands still. New techniques and models to fix current restraints or solve mathematical problems are continuously evolving. In this case our Data Scientists had to translate the abstract work of Data Science Research into added value for business. By having explored and tested different novel models we can now predict purchasing time, while also including time-varying variables; standard models only work with variables which do not vary over time such as gender. From experiment to real-life applications.
To simplify things, with these predictive models we are able to predict the likelihood an event will happen over time. In an e-commerce setting this means, we can, based on clicking behavior, purchase history and customer characteristics predict the probability that a customer is going to buy/convert on a given day. So now we are able to optimize the second aspect of relevance, timing.
These techniques are quite different from behavioral marketing automation in which marketing actions are orchestrated when a customer shows some specific behaviors. At first glance they look similar as both applications are based and triggered on the behavior of the customer. In contrast to behavioral marketing automation the Data Science approach is individualized and predictive instead of standardized and reactive. This provides the advantage to incorporate personalization and to plan marketing activities in advance.
Now we have a basic understanding of what is under the hood, we can look into the applications of purchasing time estimations and discover how these techniques are optimizing marketing activities. Forget about generic mailings at set times, but instead sent personalized mailings at times your customers have the highest likelihood of leaning towards a buying decision. Spend your marketing efforts effectively by setting business rules on which customers should receive your mailings based on effectiveness and/or return on investment. Decrease the quantity of mails sent and the risk of being perceived as ‘spammy’, but instead be of real relevance to your customers.
Furthermore, insights about which customers are most likely going to buy in a certain time-frame is providing marketeers with the ability to focus on these customers. By for example combining mailings with personalized price discounts the customer can be motivated to buy at your shop instead of a competitors' one. If we look at the other side of the spectrum, these insights can also be beneficial for detecting customers that are slowly leaving your web shop or brand behind, which can be spotted when customers are deviating from their individual buying patterns. Personalized price promotions can be targeted at those who need some extra attention restoring the relationship with your customer. In the end this will contribute to an enhanced customer life-time value and as we all know, keeping your customers is easier then attracting new one's.
In conclusion, when deploying the two aspects of relevance, personalization and timing, in an intelligent way a competitive edge can be gained over competitors while making your customers happy. Feels like a win-win.