Whatever your business, it’s now clear you need to be considering how Machine Learning can help. Already, leading companies worldwide are reengineering their value chains around Machine Learning, and they aim to significantly outperform their more traditional competitors as a result. This is because Machine Learning capabilities enable companies to better meet consumers’ appetite for real-time, context-aware services, and allow them to respond to the threat offered by new digitally disruptive start-ups.
With Machine Learning set to become one of the core determinants of business success, enterprises need to understand how they can introduce and scale the technologies across their operations. When doing so, there are three criteria to bear in mind: people, data processes and technology. Here I want to look at the first of these.
Transforming your data science team
Where Machine Learning projects fail to achieve scale, or deliver the expected results, it’s usually due to incorrect problem formulation with the business or stemming from an inability to implement the resulting models into a live environment. A lack of clarity and unity of purpose is the death knell for Machine Learning in businesses. It drains enthusiasm for the project, which quickly leads to frustration and a loss of confidence. If any progress is eventually achieved it is often too late: The business has moved on.
What can you do to avoid this happening? The answer is to ensure you create the right team and working practices from the get go; and this means moving beyond the regimented management structures that sufficed for MI/BI deployments. Today, data science teams must focus on speed and agility, so that data can be funnelled into value-adding smart algorithms as quickly as possible. To ensure you have a team capable of doing this, there are three things to consider:
Ensure you have the right people, with the right skills, doing the right jobs. Machine Learning applications requires small teams of highly-skilled coders and data scientists, able to apply the right technologies and techniques to turn data into actionable insights and fuel autonomous decision making. Other key roles include data engineers, who create the right data ingestions, and visualisation engineers, capable of capturing the "hearts and minds" of non-technical decision makers—important when you are implementing what are to humans "black boxes."
Reduce time-to-value. Agile working practices mean that your analytics team can deliver Machine Learning applications rapidly, ensuring you realise value in a short timeframe. By starting small and working towards two-to-four-week delivery windows, your Machine Learning projects will be able to achieve your business’ immediate objectives.
Embed correct working practices. Co-location is crucial: the data scientists, visualisation engineers and data engineer must work closely together throughout the project so that no time is wasted and no misunderstandings occur. Transparent working practices, such as agile Kanban boards and burndown charts, are also recommended to ensure the team and all interested stakeholders are continually up-to-date with the progress of the project.
With the foundation of a transformed analytics team, with the right skills and working practices, firms will be in a good position to build Machine Learning applications that deliver significant and immediate business value. However, people are only the first leg of the tripod supporting successful Machine Learning. In my next blog, I will look at the second leg: good data processes.