Customer Analytics

At Advantage Data we specialise in using machine learning customer analytics techniques to analyse everything about your customers, from initial acquisition, retention, cross marketing/up sell next best action etc.

Our services are described below. If you have a need for one or more of these services, contact us and ask for a free assessment of your requirements and we can quote you on your needs.

Digital marketing and customer acquisition

There are a plethora of digital marketing firms that use search engine marketing (SEM) or social marketing techniques to drive traffic and many of them do a good job, but few use data science and machine learning to optimise campaigns.

Most do it in an ad hoc fashion and use a set and forget strategy once a campaign is up and running. We use the most advanced machine learning techniques, so campaigns can be optimised as new learnings come to hand. This works across traditional search engine marketing and social platforms as well. The other failing of many traditional marketers is that the quality of customers being acquired is often a secondary consideration. If customer quality is important to you, we incorporate this element into the learning process.

The net result is that we aim to drive high quality customers at the lowest acquisition cost possible for your product.


We have been active in marketing in competitive spaces for a number of years, so we have developed a tool kit across social media, paid search, organic traffic and eDM marketing to deliver you a great customer acquisition solution across a number of possible campaigns. We will also analyse each campaign to determine which ones are providing the best value for money so you can optimise your spend.

A/B testing

Part of our customer analytics strategy involves what is known as A/B testing where we test what is suggested by our model against what is currently in place and only when there is statistical evidence that the new approach is performing better than the old approach will we swap them over and the process continues when a new test is ready to be run.

Most firms do a form of A/B testing, however, our method is completely scientific and backed up 100% by what the data is telling us.

Churn modelling

We have created propensity customer analytics models for many clients to model customer churn. It the simplest form, we produce an estimated probability of a customer cancelling a service within a pre-defined time frame (e.g within the next year). Some of the models that we have built are reasonably simple and some use 100s of features/variables as inputs. Churn modelling is the first step in the process which generally leads to next best offer/next best action campaigns which is a separate area of data science. Churn modelling in conjunction with next best action models are best practice for customer retention. We have worked extensively in the energy, insurance and finance space previously in this area.

Cross sell/up sell models

Cross/sell or up sell models work well when you have a large customer database and you are looking to monetize that database more efficiently by targeting which clients are most likely to respond to a cross sell or an upsell and at which part of their customer life cycle to offer them a deal.

Building a cross sell/up sell model is more time consuming than most machine learning applications because you often need to test different offers to certain segments of the database to measure response rates. That’s why this approach usually requires a reasonable number of customer records. It’s also often industry specific and requires a lot of custom work to achieve results. Done right, this can add a huge upside to your business for no additional customer acquisition costs.

Next best offer/next best action models

Next best action is a customer-centric marketing technique that considers the alternative actions during a customer interaction and recommends the best one. They are notoriously hard to implement but if done correctly can be cash cows.

They require a lot of testing and experimenting presenting various options to subsets of a client base and therefore typically require lots of data and patients. We wouldn’t recommend rolling out these sorts of models unless you have a large database where the benefits of next best action models can justify the effort. However, we are not put off by the challenge and enjoy working with next best action models. If we don’t think it’s an appropriate step for your business we will tell you!