Don’t touch it twice: making the case for automation

As with many other industries, automation has the potential to deliver a radical shift in the way that data scientists approach their work – and a huge boost to the value that they can deliver. In this article, Vijay Balaji Madheswaran – Director of Applied Data Science for dunnhumby APAC – looks at automation’s potential contribution to the future of data science, and the obstacles in the way of wider adoption.

“What do you do for a living?” It’s one of those questions that we all get asked, and the answer to which can change wildly depending on who happens to be asking and the level of insight we think they might have into our particular career path.

For data scientists in particular, what we “do for a living” can be particularly hard to explain. Data science is a broad label that disguises a multitude of tasks and skills, many of which require deep specialisms. That said, I think that most of the answers you’d get when asking a data scientist what their job entails would fall into one of the following five categories:

• Acquisition: the collection, maintenance, and storage of information.
• Engineering: the cleansing, extraction, transformation, and loading of data, as well as anomaly detection.
• Reporting: processing data in a way that can be used for business intelligence purposes, such as analytics, metrics, and segmentation.
• Learning: advanced analysis, including simple machine learning techniques.
• AI: the application of advanced machine learning.

While I think those five areas are a fair encapsulation of what data scientists do, I also think that something is missing: automation. Most data scientists probably wouldn’t include automation as a core part of their job, but it’s increasingly vital – particularly in terms of measuring ROI.

Much like data literacy – something that has the potential to fundamentally redefine what it means to be a data scientist the more ubiquitous it becomes – automation looks set to have a transformative impact on data science. Crucially, the potential benefits go beyond just making processes faster and more efficient; when it’s employed effectively, automation can aid in everything from employee morale through to the creation of competitive advantage. Here’s how.

Automation is the enemy of boredom
Doing the same thing over and over again is inherently boring. Moreover, it’s a waste of time, effort, and – most dangerously – the intellectual capital that you’ve invested in. Automation prevents talented data science professionals from having to repeatedly solve the same problems or carry out the same tasks. As a result, it frees them up to deliver greater value and do more engaging work.

Computers can help people make better decisions faster
If you’re a Retailer running 15,000 promotions every week, putting the analysis of those offers into human hands is likely to be overwhelming no matter how talented your team. By automating that analysis, and asking a machine to provide a recommended course of action accompanied by projected results, a human team can make much smarter and clearer decisions.

Automation can help you replicate success and expand
Say you have a team of 15 people working on the data science that powers 1,000 campaigns a year. By automating some of the processes behind their work, you can bring that team down to just two. That frees up a huge amount of talented resource to find a new problem to solve, expand into new territories, or progress new ideas – all at an identical cost.

Speed is competitive advantage
Decisions delayed are decisions unmade. The faster that a business makes decisions, the faster it can outmaneuver its competitors. Automation gives organizations the opportunity to drive competitive advantage, a factor that will only increase in importance as the application of AI becomes more common.

As with AI, the sheer volume of discussion around automation means that it can be easy to assume that just about every business everywhere has already charged ahead and automated the vast majority of their processes. That isn’t the case, unfortunately, and nor will it be unless we can overcome some of the fundamental barriers standing in the way of wider adoption.

The first of those relates to sheer awareness. Many organizations – and people – just aren’t cognizant of the benefits that automation can bring. And even those who are can still be uncertain about how best to proceed; successfully, at least. Uncertainty, of course, also makes it difficult to justify the expenditure required to drive automation forward, particularly when most organizations already have large technical debts that only increase their resistance to change.

We also need to take into account the human factor here. As referenced above, automation inherently reduces the amount of human resource required to solve problems. That’s less of an issue in data science, where expertise can be diverted elsewhere. Nonetheless, the “humans vs. machines” association is undoubtedly hard to shake, something that can also manifest in the belief that human analysis will somehow end up being more “creative” than that delivered by a machine.

Finally, for some, automation simply won’t be a priority. They may believe that their needs change too fast to benefit from it, or struggle to find the incentive required to take it into serious consideration. That, of course, is true of anything that alters the status quo.

For those organizations that do want to embrace the potential of automation, focusing on a few priorities can help to make that a reality:

• Build awareness: the biggest obstacle to automation is usually internal. Awareness, understanding, and buy-in are all vital.
• Focus on value: have a clear goal and constantly question whether the process you want to automate will deliver more value as a result.
• Make automation a priority: clear existing barriers and try and be on the lookout for anything that could be a problem in the future. Address fears and concerns, and help people grasp the potential benefits of automation.

Vijay Balaji Madheswaran is Director of Applied Data Science for dunnhumby APAC. Focused on investment, partnership, access, and customization, Vijay is passionate about helping Retailers and Brands realize the full value of their data.