A team of PhD students has undertaken a project that could transform the marketing industry, working with global one-to-one marketing specialist, Clicksco Group, attempting to pin down how to accurately predict return website visitors.
Clicksco has been nurturing the talent of the four-strong team as part of a five-week scheme called S2DS (Science to Data Science), founded and managed by data hub Pivigo. One of Europe’s largest data science training programmes, it turns exceptional analytical PhDs and MScs into data scientists who can produce real and valuable commercial outcomes.
The recruits were tasked with developing a model for determining if, and to what level of confidence, a user will return to a website within a specific timeframe. Two months of audience intelligence was taken from Clicksco’s Carbon, platform – a world-first, cloud-based audience management platform (AMP) that uses machine learning (ML) to better understand the needs, behaviours and intent of online consumers. The team delved into around 500 million data points housed on Carbon, to carry out their profiling.
An ability to predict which customers will return to a website could be a commercial ‘gold mine’, because it would give marketers valuable insights into those customers’ behaviours and allow the industry to adapt messages to the attributes of this loyal audience.
Head of data science and machine learning at Clicksco Dr Al Mclean explains: “This kind of work has huge potential to transform the industry – the monetary value of getting it right is significant.
“Our Carbon platform enables anyone with a website to analyse, create and monetise their audiences. Crucially, Carbon provided the students access to an incredible amount of data to help them better understand consumer behaviour and develop new techniques to profile them. Knowing the attributes of the people most likely to return to a website would fuel the capabilities of Carbon and be vital to both publishers and advertisers.”
The PhDs – Aishah Selamat, Cagatay Capar, Nicolas Contreras and Slobodan Radosavljevic – looked at key insights such as page visits by device type, time of day, distribution of page visit duration and more, to identify patterns in behaviour for those who were return customers. From this they matched variables, such as the lapse between visits and visit duration times, into predictive models mapping the probability of return visits. They found that just 20-25% of consumers return to the same site within the 61-day period the team analysed.
Dr Mclean added: “For advertisers, being able to confidently predict that a visitor will return can help them reduce costs and maximise return. For instance, an advertiser may spend more on a user categorised as ‘less likely to return’ to encourage them to do so. Or, they might want the most likely returning visitors to be delivered a specific page to encourage a conversion. Furthermore, advertisers may want to extend their reach by building ‘lookalike’ audiences based on similar attributes to the visitors most likely to return. On the other hand, a publisher may want audience insights into consumers more likely to return, so they can offer advertisers a way to reach those people more effectively. This project opens up a wide range of possibilities that are currently untapped.”
S2DS has completed over 130 projects with clients such as Royal Mail, KPMG, British Gas, Barclays, M&S, and now Clicksco. Candidates on the programme are among the brightest and best analytical PhD or MSc students, who are keen to kickstart their career and would benefit from a placement working on a live project.
Cagatay Capar, who worked with Clicksco on the project from its London HQ and reported its findings with the team last month, said: “What a potential employer would question [in an interview] is the lack of industry experience you can talk about…Now, I can talk about a project I worked on for a real company on a real data science problem.”
Dr Al Mclean concluded: “Clicksco is proud to partner with the S2DS scheme and help usher in the next generations of data scientists, to work on real world issues that are at the forefront of the industry.”