Five Things We Learned at Big Data LDN
Some of our Data & Analytics team went down to Big Data LDN to network and learn about the big talking points in Data today, here are the key insights that they came away with:
Insight: Know when to compromise between complexity and time
Talk: Your next basket – A Bayesian approach to relevancy modelling using WPS analytics (Jeff Ahrnsen, 8451/ World Programming)
With different models, platforms and vast parameters to consider when analysing data we are constantly weighing up the pros and cons of each to get the best output in a timely manner. In the talk on a Bayesian approach to relevancy modelling using WPS analytics, we discovered there is a constant balancing act when juggling different aspects of a project.
Insight: Keep innovating and creating POCs
Talk: Enabling data-driven decisions with automated insights (Charlotte Emms, Seenit)
Sharing ideas and putting them into practice is the start of formulating a POC, and it is important to do this as soon as possible in the thought process. Hearing about one analyst’s journey from dealing with the lack of engagement of her dashboards, leading her to think of an initial idea of insightful email type reports (POC), to learning Python to create a more robust e-mail template (MVP), to finally putting this to practice. This has been developed further since to a slackbot to return key metrics and graphs at your fingertips.
Insight: People are ready for “new nudges” and are more and more are comfortable with AI
Talk: New nudges – the next revolution in customer influence (Alastair Cole, Partners Andrews Aldridge)
With the new era of AI looming round the corner, we are slowly accepting it and moving along with it. Alistair Cole believes we’re ready for the next leap in creative thinking. Adding technology and data is the natural evolution and AI can design emotionally engaging experiences. From this we can create intelligent tools that will generate unique experiences.
Insight: Sometimes it’s best to let the robots do the work for us
Talk: Using AI and time series modelling to improve demand forecasting (Lukas Innig, Datarobot)
With the platform DataRobot, there are already pre-built models created by leading data scientists to reduce the laborious work we would have to do.We were exposed to a working example of how its time series model can help with forecasting with minimal error and allow us to minimise the compromise between accuracy, time and size of the dataset.
Insight: Integration matters, nobody has all the pieces
Talk: Automatic machine learning with guided analytics (Christian Dietz, KNIME)
As analysts we are constantly dealing with multiple data streams in order to get more insight, automating as many of the processes as possible – from pulling the data to producing actionable results. Automation can take out the drivers but it can then also take away their expertise. However, guided automation allows us to automate the tediously long pieces but keep the expert in the loop.
Find out more about our Analytics & Attribution Services here.