Your team has just rolled out a new feature, and the dashboard is responsive, showcasing a spike in user retention.
The product manager feels fulfilled, but the team, in their curiosity, ask this pertinent question: was it really the new feature that led to this success, or were there other hidden factors at play?
This is the central challenge in product development, navigating the possible deceptions of correlation to find the true currents of causation.
The field of causal inference provides the analytical rigor needed to move beyond these misleading observations and truly understand why a user’s behaviour changes.
This is the specialty of Harrison Enofe Obamwonyi, a senior software data scientist whose expertise is a crucial asset for any organization seeking to make smarter, more confident decisions. Beyond reporting data trends and insights, Harrison acts as a strategic partner, leveraging data to obtain accurate scientific inquiry for teams.
Harrison helps product managers and guides them in knowing the standard approach in designing and executing the right experiments, including standard A/B tests, the gold standard for causal inference.
Also, when the possibility of obtaining a true random assignment is slim, he employs complex quasi-experiments that are necessary, such as a phased rollout of a new feature to different geographic regions over time.
These sophisticated methods are essential for isolating the true effect of a new feature, a pricing adjustment, or a UX tweak, filtering out other confounding factors like seasonal trends or external market shifts.
By carefully controlling for these variables, he makes product decision making easier. Rather than assuming and speculating, teams can confidently state that a specific change caused a particular outcome. To tackle the complexities of these high-stakes decisions, Harrison integrates a powerful toolkit of advanced modeling techniques.
He uses econometrics principles to view user data through an economic lens. This helps him build complex models to forecast user reactions to price or product changes. Teams can grasp ideas like price elasticity with these models.
His method also uses Bayesian modeling. This proves useful for small datasets or uncommon events such as launching a niche product feature. This approach allows for ongoing learning and belief updates as new data arrives. It yields more reliable and precise results than standard statistical methods might.
His use of uplift modeling has the biggest impact. This method does more than measure the average effect of a change. It spots specific user groups most likely to respond well to an action. For instance, a company might use an uplift model to discover that a discount offer works well to convert new users. But it might have no effect on long-term subscribers who would buy anyway. This insight enables a focused and effective strategy.
Through his expertise in causal inference, Harrison provides key tools. These tools help shift from a reactive, correlation-based approach to a proactive, evidence-based one.
His work empowers product teams to not only see what happened but to truly understand why it happened, allowing them to make confident, strategic decisions that drive real, measurable growth.