Today’s organizations are swimming in data. But few know how to turn business information into actionable insights. Machine learning can help to identify course of actions and predict outcomes – which is why imagine Brunei turned to Lynx Analytics.
Many years ago, a young video game developer was invited to the office of a giant publisher in the US to pitch his product. He got into the elevator and was surprised to see that the CEO of the US company already in it.
“What brings you here?”
“I've come to see if your company will fund my game idea.”
"OK, you have the length of this elevator ride to pitch it to me now.”
“What floor are you getting off?”
Decision makers in big companies are busy people. They want to receive information that's short, digestible and – most important – actionable.
And yet, in so many companies, this is not the case. Many data specialists simply lack the skills to turn information into practical advice.
This is certainly not for lack of data. The last 20 years have seen an explosion in the quantity of data generated by the typical company. Spreadsheets, reports, and dashboards abound with endless charts.
What is lacking is the ability to isolate what is important in these docs, set it into context and then suggest a next step.
More likely, a company’s data science or business analyst team will alert the relevant decision makers to an important data point – a fall in sales or a sudden rise in customer support calls — without necessarily providing any insight into the reasons for it.
Management is simply left to ponder what actions to take, relying on experience and gut instinct, or simply requesting further analysis.
Often, the problem simply comes down to education. J.T. Wolohan, author of “Mastering Large Datasets with Python,” says: “Data scientists typically have point-and-shoot skills, but they can’t explain why they are doing what they’re doing.
“They have a hard time working backwards from questions into practical business solutions. That’s really the missing skill set.”
This is a harsh assessment, but one that resonates with many organisations.
The remedy, of course, is data storytelling.
With good data storytelling, an organisation can transform data into narratives. It can replace a report full of numbers into meaningful stories that show why something happened, and what should happen next.
The big question is how to do it. And for many, the secret to data storytelling lies with machine learning (ML).
According to Gartner, data storytelling will be the dominant way to consume analytics by 2025. However, Gartner also believes forms of AI will generate 75% of these narratives.
“It is an inevitability that we move to far higher levels of automation in analytics and that we move away from the current dominant self-service model,” says James Richardson, a research director at Gartner. “We make mistakes. Data stories are powerful, but there are some challenges. How do we resolve that? What we can do is apply compute power to this problem.”
At Lynx Analytics, we understand the power of data storytelling – and the ability of ML to uncover its most powerful narratives.
Indeed, we recently worked with imagine, a telecoms services provider in Brunei, on a pioneering project of this kind.
imagine is a fast-growing business that historically offered fixed broadband to its customers. To accelerate its growth – and to help expand into new areas such as mobile – the company's board set itself a goal to become a data-driven organisation.
Obviously, imagine already employs a number of data engineers and business analysts. However, in the past, these specialists would typically present key metrics to management without predicting future outcomes, such as individual churn risk, and proposing appropriate actions.
This is why imagine turned to Lynx Analytics for help.
It asked us to develop two prediction models. The first would help imagine to identify churning customers before they made the decision to leave the network. Previously, the company would contact departing customers only after they had churned.
The second was to identify which broadband customers would be most receptive to imagine's new mobile offering.
Predictive modelling is, of course, what Lynx Analytics does. Our data scientists use machine learning techniques to look for patterns across a wide variety of data sources. We then analyse these patterns to reveal likely future outcomes.
Obviously, imagine didn't just want Lynx Analytics to complete the models and then walk away. It wanted to upskill its data teams so that it could continue the work in-house.
Which brings us to the second part of the project: training imagine's data engineers and business analysts to become efficient data storytellers.
To do this, we devised an eight week program comprising different types of sessions: classroom and on-the-job.
In addition to specific teaching of programming languages such as Python, we showed the teams how to see narrative cause and effect in the data.
We encouraged them to quiz colleagues across the organisation so they could put every new data event – a fall in sales, a rise in ARPU, a spike in churn – into context.
Most important, we asked the employees to help us create the predictive models. They were learning on the job, which set them up to continue the modelling tasks after the eight weeks were over.
For imagine, the project was a key part of the company’s vision to be data driven. By understanding the use of raw data, it gained better insights into customer behaviour and is now able to develop better products and services.