How artificial intelligence and machine learning can revolutionize ecommerce
If we learned anything from 2017 it’s that no industry is immune to a downturn. However, “downturn” might be putting what happened in the retail market in 2017 lightly. According to a Credit Suisse report, when 2017 came to a close, approximately 6,800 stores will have shut permanently, setting a new, ominous record.
While there may be a strong correlation between the rise of ecommerce giants such as Amazon and Walmart and this downturn, the reality is a little more complicated. Brands are now having to compete on multiple fronts, including:
- With the ecommerce giants.
- With smaller retailers in an era where loyalty is declining.
- With consumer expectations.
- With attention sapping social media apps, games and messaging platforms.
- And competing for increasingly scarce engineering and research talent as more retailers evolve into technical ecommerce companies.
All of the above are compounded by the fact that ecommerce teams have to deliver an omnichannel experience without negatively impacting sales margins. With the latest advancements in AI and machine learning, retailers who have an ecommerce presence are beginning to really hone in on creating a custom, brand-focused experience, for each visitor.
Relevant experiences increase the likelihood of customer engagement and the potential for conversion, or purchase. Relevant experiences also enable brands to differentiate themselves and engender customer loyalty and lifetime value.
While some may have dubbed AI and machine learning as the ultimate solution to everything, others dismiss it as hype. The benefits of using AI and machine learning to process data to optimize the customer experience in a tangible way are very real; however, it is also critical to understand what is actually required for AI and machine learning to have the kind of impact that is expected.
To start with, it is a high-technical debt undertaking and requires high quality data, timely decisioning at scale, and a hypothesis-based approach to marketing. Before any machine learning can be successfully applied, it needs data. And not just any type of data.
Firstly, the data must be useful, in a way that reflects first-party behavior on digital channels, such as web pages and single-page applications.
Secondly, the data must be processed quickly. Thanks to the latest cloud technology, data can be processed in an almost real-time fashion, with latency of a couple of minutes, rather than hours. Finally, and by definition, personalization is personal, that means that it’s unlikely that a specific set of machine learning approaches for one brand will translate to another.
In order to achieve success with this, marketing teams need to employ a hypothesis-based approach to marketing, where they use the inferred signals from machine learning in conjunction with creative brand experiences.
AI and machine learning can effectively change what is delivered to each customer based on data derived from a variety of sources such as a visit, a click and historical data, combining to provide experiences that are more effective at helping each customer to find exactly what they need at the right time.
Once the above requirements are satisfied and the machine learning practitioners have the opportunity to mine the rich set of first party behavioral data, the models can lead to multiple benefits. Marketers can use the deep customer insights that AI and machine learning provides through slicing and dicing the data to posit predictive insight about the entire customer base.
This powers customer segmentation in new ways which allow marketers to deliver experiences to more specific segments and launch relevant experiences to these segments while measuring the performance of each campaign and making changes at a much more rapid pace than was previously possible. Furthermore, the predictive model itself can be applied, sometimes in real-time, to enable context-aware experiences for customers.
Before leaping fully on the AI bandwagon for its transformative potential, it is also worth mentioning that deploying AI and machine learning requires engineering expertise beyond just data science and research, like expertise to first determine whether client-side or server-side AI is the best fit, and of course, to employ it correctly. Client-side AI and machine learning requires web front-end expertise.
Server-side machine learning requires cloud platform expertise and in this context, has the distinct advantage because it doesn’t directly impact site performance. However, it does depend on the speed of the internet so there are some variables out of your control.
Latency is a major issue that can make a tremendous difference in conversion and bounce rates. An ExP Platform (a Microsoft company) benchmark study on response time and site loading times reported that for every 100 millisecond of extra site loading time, there is a 7 percent decrease in conversion rate. And, a two second delay in page load time can increase bounce rates by up to 103 percent. Or, to put it another way: today’s consumers are not only savvy and less tied to specific brands, they also expect to find what they need quickly and poor site performance is a deal breaker.
As the right data and expertise fall into place, AI and machine learning will be a real game changer for ecommerce brands. It certainly should not be relegated to the pile reserved for hype.