Using MarketDial’s software, companies can design and conduct statistical tests that evaluate potential initiatives. Using optimal sampling and robust control-store matching, an estimated lift (and associated confidence) for the initiative is calculated and reported.
However, at the conclusion of a test, companies need the ability to probe deeper. In order to make an informed decision, companies must answer several questions such as:
- What store attributes and market indicators drove the highest lift observed in the tested treatment stores?
- What is the predicted change in revenue/profit/traffic when the initiative is carried out on a particular subset of stores in the future?
- Can a reliable confidence interval be given for these predictions?
MarketDial’s Impact Analyzer leverages sophisticated predictive modeling to estimate the impact of rolling out the test on a broader set of stores. With Impact Analyzer you can answer the questions above and more.
To identify the optimal rollout group and the likely impact of an initiative, we employ several statistical models such as:
- K-nearest Neighbors and Principal Component Analysis to remove redundant information from the data and compare stores in the fleet across influential variables (median age, store footage, number of competitors in store vicinity etc.)
- Gradient Boosted Regression to simultaneously perform variable selection and create predictions for the expected lift on future roll-out stores.
- Leave-one-out Cross Validation to ensure that our statistical estimates of noise are robust and unbiased.
Getting It Right
request a demo