Arrow up SEE MORE Latest

Thoughts

So you think you're a data scientist?

By Claire Tusler

Having a PhD doesn't make you a data scientist. Turning customer data into immediate actionable insight does.

 

The world of marketing has changed irrevocably over the past five years, and the skills, personnel and make-up of marketing teams have changed with it. With all industries trying to ensure they aren’t left behind in the world of data science, this has led to an influx of new job titles in the marketplace, from data futurist to Internet of Things architect. 

 

So, how has the world of marketing evolved from having teams of analysts to departments of data scientists and what does it take to be a data scientist in the marketing world?

 

Data science involves different skills, but most importantly a different perspective. For me, a data scientist has the mindset and skills to take customer data in its rawest form and turn it into actionable customer insights that can be applied in real time. If there aren’t actionable insights after a piece of work, what was the point in the first place?

 

Today, too many data scientists focus on the data itself rather than the context. This tells you what is happening, not why people do things. Bottom-up thinking only leads to myopic solutions.

Start with the business issue and customer insight

The more specific the issue is, the more articulate and accurate the solution can be. A company that has been built on this premise is Uber. The pricing strategy and data science that sit behind the algorithm has been built to ensure the business is as efficient as possible. When there is a peak in demand, prices increase. So, when you come out of your concert along with 20,000 other people and all need to get to the station, Uber knows this and adjusts accordingly. 

 

From the customer perspective, they know exactly how much their taxi will be and when it will arrive – two of their biggest gripes with conventional taxis. This is smart business and customer-focused, and drives real-time actions. Let’s hope the brand will still be able to operate in London.

 

Another good example is the #YayDelay project for Gordon’s gin created by MullenLowe Profero and The Romans. #YayDelay used real-time train information and geolocation targeting to identify when trains were delayed (so, pretty regularly). 

 

The campaign then targeted commuters waiting at Waterloo station with special offers for a Gordon’s. This not only promoted Gordon’s to core target audiences who are primed to respond, but also showed empathy with those whose biggest grumble is that no-one cares that their trains have been delayed yet again. Genius. 

 

An example of where customer behaviour is linked really simply to a business issue is Burger King’s "Whopper detour" campaign by FCB New York. Burger King app users who came within 600 feet of a McDonald’s were targeted and offered a Whopper for a penny. 

 

Being bold enough to be completely transparent in what they were doing tapped into the behavioural lever of being part of something so obviously slightly naughty. Driving sales, increasing downloads of the app and, most importantly, creating a huge amount of awareness all prove the impact of using readily available data to create a solution. 

 

Four ingredients for success

At Proximity London, our data science department focuses on both business issues and customer insight and determines answers to both. 

 

We have developed a proof of concept for Specsavers, which has an issue with people not turning up for appointments, resulting in lost revenue. Using historical data, we isolated a range of key identifiers on those most likely to be no-shows, from profile info to the type of product bought. In addition, we linked external factors such as traffic information on the day, bank holidays and other seasonal events such as Black Friday. This allowed us to predict who would be most likely to not turn up and pre-empt this with timely reminders, tailored to each individual. For example, mums would get a reminder 24 hours in advance on text, whereas the over-75s would receive a letter a week before. 

 

In conclusion, there are four key elements required to be a truly effective data scientist. They are:

  • Answer a clear business problem

  • Think about the customer insight

  • Drive real-time immediate actions

  • Use a variety of data sources 

It is only by combining the above factors that we can reach actionable insights and apply them to client challenges.