As straightforward as it sounds, continuous optimization is an emphasis on the part of a company to never stop running it’s optimization campaigns. An ongoing process, continuous optimization involves the ideation of new tests based on available data, prioritizing them along a testing roadmap, running personalization campaigns, analyzing the results, optimizing for efficiency and then cycling back to the ideas phase.
This repeatable process not only allows for more tests to be run within a given timeframe but also contributes to more growth, seeing 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.
Continuous optimization also refers to experiments that are running 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 winner 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. Simply put, one variation cannot be suitable for all visitors. 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.
With the help of Contextual Bandit algorithms and predictive analytics, effective optimization solutions can 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.
