How I got here as SCMP's lead data scientist
If you leave it there, data is just a number, but it can be an asset if you can use it to create value for the business, according to SCMP’s Lead Data Scientist Emily Xuan.
Takeaways
No one knows everything on the job, one needs to keep learning and growing.
To grow the audience, media publications need to define the value of the audience, analyze their behaviors, find like-minded users, and at the same time, improve content to make the audience return.
If you leave it there, data is just a number, but it can be an asset if you can use it to create value for the business.
Context
Big data is applied in every aspect of life, from putting products on the market shelf and friends’ suggestions on Facebook and claims prediction in the insurance industry, but how is it applied in the newsroom?
Like many others, the 117-year-old Hong Kong-based South China Morning Post is going through a rapid transition. It is leaning heavily on machine learning for projects like content recommendation, customer journey optimization, and image recognition.
Emily Xuan, the lead data scientist at SCMP, talked about that at Splice Beta Online. A former data scientist in the insurance industry, she holds a rare role as one of the few female data experts in media. She talked about how she found her passion in turning data into business insights.
How does SCMP use machine learning?
Machine learning drives SCMP forward on three fronts: advertising revenue, audience engagement, and loyalty. It can be applied in three ways:
Predictive modeling to grow audience
Persona analysis to profile the existing audience and gauge their engagement level
Sentiment analysis (via natural language processing), article readability, and contextual tagging
What are some examples?
The audience growth project called Bluefin aims to identify potential loyal readers, create personalized reader experiences and maximize marketing budgets.
The business question: how to increase retention rate within the fixed campaign cost?
The tech question: how to find a valuable audience with cookies? How to predict which visitors would come back?
In A/B campaign tests in the U.S. and Asia, the predictive algorithms increased engagement by 58% to 78% and cost-effectiveness by 36% to 52%.
In another project with the mission to lead global conversations about China. The team scraped over 60 China social media platforms such as Weibo and Douyin in the entertainment sector, and Snowball for finance.