Predictive targeting or predictive intelligence is the process of collecting website visitor data based on their online behaviour. This data is interpreted and helps companies apply their behaviour patterns to future website visitors. 94% of companies agree that personalisation is critical to their success. However, less than half are personalising their website experience (Econsultancy).
Predictive Targeting detects similarities between the behaviour of current visitors on the website and a previous group of visitors that is not necessarily on the website right now. This way, iAdvize can set up algorithms that predict how future visitors will behave on a website.
Predictive Targeting – how is it done?
Data from past visitors is collected: we are aware of the behaviour they had on the website according to different variables, we know if they bought something, used the chat solution or if they added an item to their basket. Behavioural groups are created based on past visitors’ behaviour. The algorithm analyses visitors that are currently on the website, it studies their active behaviour and compares it to the behaviour of former groups of visitors.
The iAdvize algorithm checks the specificities of this group (purchasing behaviour – did they talk to an agent?) and in this way, we know if the visitor needs help on the website or if he is likely to make a purchase: they predict his/her next move.
We collect the first range of data that evolves in time. We use an average of 15 data variables we collect from visitors.The data is related to their behaviour and is collected and processed lawfully.
We collect this data in real-time based on visitors’ actions. When a visitor goes to a website page, we can track all his/her actions: maybe he added an item to the basket, he looked at a product page or the checkout funnel. We record this data in our system.
How does predictive targeting deliver ROI?
Predictive targeting has a double advantage:
#1 It’s based on data collected from real human behaviour and not only fixed rules, for example having a basket value of more than £30 on a product page. Over time, a rule like this is not optimal, as the visitor behaviour can change. The rule won’t change it.
#2 With Predictive Targeting, the online shopper behaviour changes, together with the mathematical model, so we don’t necessarily target the same people. This mathematical model helps us target visitors that have a high probability to leave the basket or quit the funnel.
We can use these predictive criteria to define more detailed targeting rules, which are more ROI-oriented and help create value. We are already observing a potential of delivering ROI while simplifying the targeting strategy.
Best practice for retailers
To implement self-learning mathematical models, we use the data collected from visitors’ behaviour on the website, but we also work with retailers to integrate their customers’ data that is not freely available to us. By giving us access to this data, we can refine the behavioural targeting model and improve the performance of the mathematical model.
Having more comprehensive data helps retailers who don’t have a lot of traffic on their website to create simple targeting models with few rules.
For retailers with a lot of traffic on their website, it can be interesting because there are a lot of visitors on the website, and predictive targeting will help them identify the visitors who have the most added value for the retailer.
The future of predictive targeting
“You shouldn’t limit yourself to the information you can have about visitors on the website. Ideally, you should have a cross-channel vision: behaviour on the website, social networks, other websites. By gathering all this data and understand the how visitors use the Internet to inform themselves and purchase, we will be able to build web users profiles,” says Marion Guillard, Data Scientist @iAdvize.
By standardising the knowledge we acquire from visitors on the websites and social networks, we will be able to offer products proactively and efficiently, having prior knowledge of their purchasing habits. Before we think further about other possibilities, we should be able to focus on improving current predictive targeting rules.