Continuous Optimization

Continuous optimization emphasises to never stop it’s optimization campaigns. Being an ongoing activity, it involves the ideation of new tests based on available data, prioritizing them along with a testing roadmap, running personalization campaigns, analyzing the results, optimizing for efficiency and then reverting back to the phase of ideas.

The repeatable process contributes to more growth, and as more results = more learnings = data activation = maximum returns.

But extracting useful meaning from optimizations requires persistence and patience, and marketers must not jump to conclusions, but instead, understand that a good experiment should look beyond short-term results and include long-term KPIs.

This term also refers to the experiments that run on bandit algorithms that are not limited to determining a single definitive winner. Instead, with the help of contextual and multi-armed bandit algorithms, continuous optimization aims to guarantee the best variation is served to each individual visitor based on the current trends and data.

The problem with traditional winners takes all approaches to A/B testing is that permanent implementation of a single winning variation provides a homogeneous experience for all visitors, making it a fundamentally flawed optimization strategy. Marketers need to be patient and understand that a good experiment should look beyond short-term results and include long-term KPIs such as predicted Average Lifetime Value and impact on revenue — which is where dynamic experimentation solutions come in handy. 

When testing different sample spaces of consumers on-site, some test results end up being inconclusive and retailers feel as though they may lose potential revenue. Inconclusive test results may occur due to different reasons, the most common one being that the test is not run long enough to reach an error-free conclusion, But the need for human interference to end a test is eliminated with continuous optimization.

Contextual Bandit algorithms and predictive analytics help effective optimization solutions focus on dynamically maximizing the collateral impact of continuous optimizations in the long-run, instead of statically optimizing single A/B testing initiatives for the short-term.