The pandemic has accelerated changes in how people shop: consumers want more choice, faster and with less effort for them. They want the organisations they do business with to understand and know them and predict their needs while ultimately staying in control. Creating a personalised experience that is trusted has become more important than ever. And this means that the pandemic has brought data and ML even further to the forefront.
1. ML success will correlate heavily with teamwork
In 2012, HBR declared data science as being the sexiest job of the 21st century. Since then, in a 2019 study, 7 out of 10 companies surveyed reported minimal or no impact from the contributions of AI. So why did AI fail to deliver for those companies?
One key reason could be the focus on data scientists rather than on data products. Great data products bring together superior knowledge of the business challenge with a very clear understanding of what is technically possible and a clear sense of how to lead the user through the solution. This requires a mixed team with a common goal. Data scientists may be only able to have limited impact if they aren’t truly embedded in the development of data products.
Engineering, Product, Data Science, UX and others all have the same job – they need to create the best product possible. But their different crafts mean that they bring a different perspective and different experiences. Creating the right operating model where these differences create something greater rather than the sum of their parts is key to success with ML. Do this by embedding data scientists in the product engineering teams and by investing in data literacy across the organisation.
2. Quality data will become the cornerstone of MLOps
ML success also depends on making the development, deployment and management of ML solutions at scale as simple as possible. Developer time should be spent on developing and trialling new approaches, not on deploying and monitoring services. And in the case of ML, this doesn’t just apply to code but also to data. Andrew Ng has been speaking about data-centric AI in this context, about how improving the quality of your data can often lead to better outcomes than improving your algorithms (at least for the same amount of effort). So how do you do this in practice? How do you make sure that you manage the quality of data at least as carefully as the quantity of data you collect?
There are two things that will make a big difference: 1) making sure that data consumers are always at the heart of your data thinking and 2) ensuring that data governance is a function that enables you to unlock the value in your data, safely, rather than one that focuses on locking down data.
3. ML thinking will become more holistic, less siloed
We sometimes need to change the conversation about what a product is there to do. Organisations can improve customers' experience by making the full interaction feel like a continuous conversation that flows naturally from end-to-end. For Ocado Technology, from a data point of view, this means thinking about the customer's journey through the shop from the selection of a delivery slot all the way to the checkout, rather than as a series of discrete interactions. Where in the customer flow would it make sense to add a bit of friction now, that will allow us to remove more friction later in the journey? Where can we explicitly ask a user about an assumption that we might have about them to ensure that we really understand their needs and preferences? And how do we make sure that the way we think about ML in a product covers the whole journey rather than just a single interaction or click?
Nowhere is this more important than in the world of grocery retail where an average basket size is significantly larger and shopping frequency much higher than in most other retail segments. Combining algorithms to create a more consistent user experience across the user journey will be a key focus.
Data Science Beyond 2022
Data Science is all about turning data into value. To build the best data products, organisations must draw on multidisciplinary expertise and build robust MLOps centred around data collection and data quality holistically into their business. With the ever accelerating eCommerce landscape, the race to unlock the power of data and meet consumer demands for hyper-personalisation is on.
Thanks very much to Chief Data Officer Gabriel Straub.