How does Lead Ratings work?
When we tell a customer or a friend that our system is able to predict the lead conversion probability rate, they generally don’t believe us. How can you know it with such little information? What’s the trick?
The trick – or better, the non-trick – has a name: Data Science. No, we don’t have the magic ball that let us read the future, but we use historical sales data from our customers in order to get information about channels, products, gender, location, etc. that best work in their campaigns. This analysis is much more sophisticated than simply observing the variables’ results one by one (channel, product, etc). Our models take all variables into account at the same time in order to draw the different buyer profiles and classify them according to their conversion probability.
The process works as follows.
- First, our customer provides us with sales historical data – generally from the last twelve months. This historical data contains customers personal data (location, name, age, etc) as well as sales data (entry channel, product that called his/her interest, etc).
- Our team of Data Scientists analyze all the incoming information, completes and enriches it with external data sources. In this analysis, our team siphons and sorts all data and determines which are the most relevant variables to predict the conversion.
- Once the data is prepared, we create a statistical model that will determine the sales conversion probability for each lead. In order to create this model, we use Machine Learning techniques on the available data, and we check that the predictions match the reality, that is, that the lead patterns identified by the model are stable.
- We also work to get a more powerful model that maximizes the benefits for our customers: we aim to isolate a small group of leads that generates most of the sales. In this way, the customer can concentrate their marketing campaigns on these leads, with the highest conversion rate, increasing the profit margin.
In many cases, we face a compromise between power and stability of our models. Models that are very powerful risk finding patterns too specific, that are not applicable to the whole sample (know as overfitting), whereas others can be very stable but unable to discern high from low conversion probability leads.
- Finally, once we have selected the model that has proved to be the most appropriate and we have the customer’s ok, we deploy the model in our SaaS platform in order to be used through a simple API on the Internet.
As you can see, there’s no magic in our process. Just science!
Does this service sound attractive to you? Try our plug and play solution for free and see how Lead Ratings can boost your business by improving your marketing performance.
Request your free trial by filling in the form on our homepage.