In light of your experience, what are the trends and challenges you’ve witnessed happening in the Predictive Analytics space?
An important trend and challenge in our field is building reliable real-time data streams. This technology enables our platform to interact with our customers as they traverse our site rather than passively reacting to stale data. For example, if a user interacts with a recommendation’s carousel or an ad, real-time data streams empower our platform to immediately update their recommendations based on this action. Older batch approaches can only update the user's recommendations the following day, which may be too late to engage with the customer as she may have lost interest by that time. These systems require a lot of infrastructure support and maintenance to function properly. This trend has been an exciting challenge for our team as they build these systems to craft an interactive customer experience.
Could you talk about your approach to identifying the right partnership providers from the lot?
I believe that the best teams are built on diversity, and every partnership should add unique value. Consider where your gaps are in terms of the skill sets of your teams and the capabilities of your systems.
There are numerous specializations in data science, and more are emerging. For example, if your company does not have effective recommender systems in place, you might want to consider partnering with a provider who can augment your efforts.
What are some of the points of discussion that go on in your leadership panel? What are the strategic points that you go by to steer the company forward?
As a leader, I want to ensure that our strategy aligns with our efforts to drive an experiment-driven culture. I reinforce the importance of establishing and standardizing best practices for experimentation across the company, including proper A/B testing and validation metrics. One way to accomplish this is to integrate data science throughout the business. Specifically, you can integrate data scientists into each functional area of your organization. In effect, you can ensure that all your strategic initiatives have a data science resource dedicated to them. Another point I like to emphasize in my leadership discussions is that predictive analytics is experimental in nature, and part of healthy experimentation is allowing for failures. Often, we are working off of probabilities and correlations, which should not be treated as proxies for causal relationships.
How do you see the evolution of the Predictive Analytics arena a few years from now with regard to some of its potential disruptions and transformations?
Predictive Analytics has the power to not only engage customers but also to help drive and inform internal processes within companies. At Overstock, we strive to leverage the same predictive models we use to engage customers to drive decisions within marketing, sourcing, and other teams. These technologies enable automation of data analytics over billions of data points to draw out salient conclusions. Data that would normally takes months to distill can be far more efficiently processed to identify key initiatives. As the field develops, adoption of this technology across an organization will help create a data-driven culture. This doesn’t just help with externally-facing products, but also with internal road-mapping, identification and alerting around failure points, analysis of results on A/B tests, and more.
What would be the single piece of advice that you could impart to a fellow or aspiring professional in your field, looking to embark on a similar venture or professional journey along the lines of your service and area of expertise?
Don’t be afraid to take on new tasks and challenge yourself. There are opportunities everywhere to learn and grow. There are plenty of open-source tools and libraries available on public repositories that you can leverage. Lastly, be confident. If something has never been done before, it just means that you can be the first to do it.