DATA AUDIT AND STRATEGY
Case: Effectiveness of interventions
A non-governmental organization in the youth development space approached us with program intervention data, which had been collected over time and participant journeys; before, during, and after the program. Through data audit, we discovered a pattern on dates and events, hence formulation of a strategy which resulted in a significant improvement in program outcome.
Evidence based decisions
A large corporation in the finance sector wanted to understand the best ways to plan and store their data to support decision-making correctly and efficiently. A key finding was that most of their data were continuous, and they had been trying to build causal relationships. We set up appropriate logistic regression models, which saved the management a lot of time and discussions, transforming “in my opinion” and “my gut feeling” to “what does the data say”
How well can you segment your customers/population into groups that differ from each other, based on some quantity of interest?
A client wanted to predict the correct level of outcome metric based on some historical data, we used the "small area estimation" method to combine multiple data sources to capitalize on each source’s strengths. Logit models were for prediction. The new level of metric had a higher accuracy when compared to the real results than the previously reported numbers. It was used for planning going forward.
Identifying the factors which affect your metrics of interest
A client in the financial sector wanted to understand the differentiation in user experience, we used a traditional methodology (Mystery Shopping) but on a modern scale (Video analytics). While this innovation enabled quick scaling at an affordable cost, the institution was able to identify areas for “cementing long-lasting client relationships”
Discover patterns in complex data
One of our clients had a massive amount of service subscription data, which had been stored in 2 legacy systems. One of the biggest challenges was that the systems were not speaking to each other and had different types of data. Using machine learning, we were able to integrate the datasets into a modern business intelligence system and run a couple of analyses including prediction, anomaly detection, diagnostics, automating insight, reasoning, and time series prediction. The result was very handy in decision-making under uncertainty.
Slicing the market into smaller portions for optimum understanding and management
A large bank wanted to understand the impact of digital banking on their customer base, and average value per customer, and long-term impact on the business, while this looked like a typical research question, we applied latent class segmentation using the concept of core members to analyze not only the clients base but the market in general.