Consumers increasingly use e-commerce platforms to buy groceries, home decor, electronics, and more. As a result, mature platforms collect data to understand customer behavior and leverage it to draw insights that help enhance customer experience, leading to higher purchase rates, long-term customer loyalty, and ultimately higher market share. In the course of a purchase cycle, the millions of active and engaged customers looking at hundreds of products generate petabytes of data that captures what the customers are looking for and how they are finding it, and more particularly, what’s preventing them from making the purchase at that particular moment.
If envisioned and executed correctly, data and insights-based decisions could be a strategic advantage. While most e-commerce companies are still in their early stages, they tend to have hybrid teams with business intelligence experts who do a bit of everything, such as managing infrastructure/ ETLs and data models, developing dashboards, setting up/analyzing experiments, as well as generating insights to drive business strategy But as they grow, the cost of suboptimal execution increases as constant tool migrations, unstable ETLs, and poorly setup experiments could seriously stymie a rapidly expanding business. Since the end of last year, most online retailers are planning on expanding their product analytics to include more defined verticals within Product Analytics.
This team is responsible for defining what customer behavior to capture in the collected data, how to model it for efficient storage as well as delivering insights at the speed of thought in a reliable and stable environment. A careful selection of tag management tools, cloud-based storage platforms, and highly-skilled engineers who can balance the speed of agile environments with sound long-term decisions is required to work effectively.
An analytics team comprised both technical and business analysts, capable of taking on three significant responsibilities.
It is necessary to define Relevant-Actionable- Measurable (RAM) KPIs to assess business performance and create an intuitive self-service reporting suite. In order to investigate declines or upticks efficiently, you need to understand the relationship between metrics in a Mutually-Exclusive-Collectively-Exhaustive (MECE) tree format.
To better understand how customer behavior changes with different product designs and improved journeys, create an experimentation strategy. Here, hypotheses are defined with a focus on learning, as well as the right metrics to measure impact and experimentation methodology. For example, if traffic is low, try Bayesian instead of frequentist; move quickly by rolling out variations that are good long-term solutions and use Synthetic Controls to measure impact post-hoc, and use Multi-Arm Bandits to run multiple variations without affecting customers significantly.
Customers can be segmented using data science methods; levers to optimize can be identified using feature models, and better solutions can be identified by topic modeling customer feedback.
Any retail tech company would benefit greatly from having a data science team that is comprised primarily of data scientists and machine learning engineers. For example, data scientists should take on some of the most complex business problems to develop the most sophisticated and real-time solutions that will have a huge impact on customers and businesses. Examples include machine learning algorithms that optimize customer journeys in real-time, operations research projects that optimize driver routes to reduce delivery times while controlling costs, and AI chatbots.
As a fourth option, the analytics products team can assist the data science and analytics team by procuring the right tools and by establishing best practices across the organization.
A flexible environment is therefore essential as we move towards specialised role-based org-design, where each business has a squad developed to focus on specific customer problems, with representatives from each vertical embedded in the business unit to establish a balance between quick execution and technical knowledge
