Data science needs data scientists
It’s well-documented that the online shopping space has become even more competitive over the course of the pandemic. In fact, figures from Salesforce’s latest ‘Small and Medium Business Trends’ report reveals 35 percent of SMEs added eCommerce functionality over the past 12 months. In the face of market saturation, longer-term growth will be predicated on understanding how to collect and use data to ensure you are reaching the right customers in the right ways.
Data science allows a business to go beyond reacting to the customer behaviours that are revealed in the daily reporting. It makes it possible to predict behaviours by modelling responses to particular scenarios based on that historic data. This then allows you to develop future-focused strategies to identify where exactly growth opportunities lie.
This probably sounds too good to be true, after all why hasn’t everyone jumped on the predictive analytics gravy train? The honest answer is, many have tried, but most have failed. Those businesses that don’t make it work often end up disillusioned and turn their backs on data science.
Putting the foundations in place
There are lots of variables for why data science goes wrong. This starts with trying to run before you can walk. Any data scientist worth his or her salt lives by the mantra ‘rubbish in, rubbish out’. All the customer data you hold needs to be stored in a single and accessible cloud-based database. All the data this holds needs to have undergone quality control, so there is no duplication and no inaccuracies.
Moreover, there needs to be sufficient data in place for it to be meaningful. This translates to 12 months or more of historic customer data in normal trading conditions. Unless there is enough representative customer data to build on, the results of any predictive analysis can veer wildly into the dreaded data drift and present false positives and worse.
Science, not magic
The second issue lies in understanding – and being realistic about – what customer data can actually tell you. It’s doubtful quantum computing’s future lies in eCommerce, we aren’t looking to understand what hunter-gatherer quirks in an individual’s genetic code is going to make them most likely to buy a particular coat on Black Friday.
What data science can reveal is grounded in sales and behavioural trends across broader customer cohorts. We can use this information to inform business decisions – for example where demand lies for a new product range or what geographies are most ripe for expansion.
One step at a time
Think of an eCommerce business as being much like an onion – there are many layers and you’ll end up with tears in your eyes if you approach it wrongly. You can’t do everything at once when it comes to data, it’s all too easy to be overwhelmed.
Unpeeling each (departmental) layer offers fresh insights. The picture will get bigger with each passing quarter and the data should be reviewed on that basis to identify where next to focus to gain the next set of data that will unlock additional growth opportunities. Bear in mind, these might come from fulfilment or customer service, as much as from acquisition.
The aspiration should thus be to gain a helicopter view of customer journey across the entirety of the sales funnel.
Who owns data science… And should they really?
This is all well and good, but the big question is who’s going to be doing the unpeeling?
That depends on the business, but given data science fits firmly into the realm of the ‘techie’, it is often allocated to the IT department. This is invariably a mistake. Technologists have a particular worldview, which is that any project should be approached as an engineering problem. Data science can’t be ‘solved’, it is an evolving tool that needs ongoing attention to keep the data usable.
In other businesses, customer data is viewed as the property of the sales and marketing teams, and they can be very proprietorial about it. In fact, this is exactly the sort of attitude that will doom data science to failure before it’s had a chance to bed in.
As we’ve seen, the real value in data lies – over time - in being able to join the dots between departments. This cannot happen if that data remains siloed. From this perspective, no single team can lay claim to data. Digital transformation can only work with horizontal ownership of data and with all departments working towards shared goals.
This is the essence of a data culture, making it happen needs strong leadership to navigate the internal politics as data is wrestled out of the hands of individual teams.
However, this is still just the starting point. Once data has been unshackled, the value needs to be extracted – and that takes experts.
Data! In an Adventure with Scientists!
So, asking your IT or marketing team to pick up data science on top of their day job will only be an exercise in frustration for all concerned. It may not involve rockets, but as the name suggests, data science is still a science. And it needs scientists to make it work.
Data scientists operate outside the confines of any single department and their skillset lies in seeing the big picture through the minutiae. It’s a highly specialist role that requires engineering and coding skills, alongside attention to detail and a commitment to continued monitoring, tinkering – and occasional retraining – of data to prevent it from degrading.
Making predictive analytics work isn’t alchemy, but it isn’t easy either – and it means investing in or hiring in the technologies and the skills to make it work. It’s not for everyone, data scientists are truly a rare breed. Not appreciating this fact is most typically where most businesses get it wrong.
The eCommerce sector has always been Darwinian, there are only so many customers to go around to support the continued growth that investors expect. We can expect a mass extinction event at some point soon for those unable to keep up. As such, predictive analytics will become a business imperative for those that want to be category leaders – and that’s the only way to ensure growth.
Most digitally savvy businesses will have a data strategy in place by this point, this puts them at an advantage over all the new entrants to the market over the past 18 months – whether those are pureplay eCommerce businesses or traditional retailers that have been forced to branch out. However, these new competitors will be putting down their digital roots and pulling together enough customer data to start their own data science adventure.
Market conditions are still anything but normal, but they will settle. Consequently, there is a receding window of opportunity for the incumbents to put their own data-led growth strategies in place and sew up their particular market segment before an existing or emerging competitor gets there first.