Sydney MeasureCamp 2019

A short summary - 26th October 2019

As part of Sydney Technology Week, I attended Sydney MeasureCamp at Google’s headquarters. This “unconference” is an inclusive transfer of knowledge and entertainment where attendees are invited to present on a topic of their choice for 30 minute sessions. Here is a quick summary of three presentations from the morning sessions.

 Why marketers are Flat Earthers!

Reze A (sorry I was so busy thinking, I didn’t get the last name) argued that agencies, like Flat Earthers, start with a misguided theory and then disregard all evidence that contradicts it. He reminded us that Null Hypothesis Testing assumes that scientists are looking for evidence that their theory is wrong. Reze went on to explain the importance of carefully considering key metrics, sample bias, statistical power, and normalising data.  Data scientists have a responsibility to work with marketers, explain complex statistics, contradict concepts, and ensure data hygiene, not just support creative campaigns.

 It’s not all about automation

Anthony Contoleon of Budget Direct shared the lessons he learnt when considering the usability of their website. Data can easily go missing in lower traffic areas and lost page views slip through the cracks because user testing doesn’t always consider every device. In Contoleon’s experience there is no way to automate A/B testing. Instead his team physically check overall trends, week on week comparison and week over moving average comparisons to keep their finger on the pulse of how users are interacting with interfaces. In the Budget Direct business, manual checks are mandatory so they can pick up errors as early as possible.

 Data in the New Oil

In his talk Data is the New Oil, Paul Tuls suggested that data is simply the by-product of measurement. Measurement is designed for a specific task, but data is only retrospective.

“When it comes to integrating data, we are all biased”.

While training and being statistically rigorous can reduce bias, it cannot alleviate it all together. While machine learning is expanding our ability to model data every day, it seems that there will always be inequality in data processing which limits what can be pulled from data. This is where human thinking will always come in.

Thanks to all of the sponsors who made the event possible and everyone who participated.