Features

Understanding Genetic Optimization

14 min read

Eevy AI does not run simple A/B tests. It uses a genetic algorithm — an AI optimization technique inspired by natural evolution — to continuously test and evolve your review section layouts. Instead of testing A against B, the genetic algorithm creates entire populations of layout variations, measures their performance, and breeds the best performers to create even better ones.

This guide explains the core concepts, how Eevy AI applies them to your store, and how to interpret the results in your dashboard.

What is a Genetic Algorithm?

A genetic algorithm mimics biological evolution. It starts with a population of candidates (layout configurations), measures their fitness (RPV performance), and uses selection, crossover, and mutation to create the next generation. Selection: best-performing layouts survive. Crossover: high-performing layouts combine — layout A's card style with layout B's sort order. Mutation: small random changes explore new possibilities. Over many generations, the population evolves toward higher performance. For background, see our post on genetic algorithms in e-commerce.

What Eevy AI Tests

The algorithm tests many parameters simultaneously: Layout type — which of the 22+ section types to show; Display configuration — number of reviews, card size, spacing, columns; Sort order — most recent, highest rated, most helpful, or mixed; Visual styling — card styles, arrow types, color variations; Content prioritization — photo reviews, video reviews, long reviews, or verified reviews first. Each unique combination is a "variant." The algorithm can test hundreds per generation.

How Generations Work

A generation is one cycle of testing. Each generation, Eevy AI creates a population of layout variants, distributes traffic, and measures performance. A generation typically runs 3-7 days depending on traffic volume. After completion, the algorithm evaluates all variants, selects top performers, and creates the next generation through crossover and mutation. Early generations show more variation (exploring), while later generations converge on high-performing configurations.

Fitness Function: Why RPV Matters

Eevy AI optimizes for Revenue Per Visitor (RPV) — total revenue divided by total unique visitors. Unlike conversion rate alone, RPV accounts for average order value. A layout might convert slightly less but drive higher-value purchases, producing more revenue overall. RPV captures this. The algorithm assigns higher fitness to variants generating more RPV. This is why Eevy AI finds non-obvious winners that simple A/B tests miss. Read more in our Maximizing RPV guide.

Reading Your Optimization Results

The Dashboard Optimization tab shows: Current generation — which generation the algorithm is on; Top variant — the highest-performing configuration so far; RPV trend — RPV over time across generations; Population diversity — how varied the current population is (high = still exploring, low = converging); Improvement — percentage RPV improvement since the first generation. Do not panic if RPV dips between generations — the algorithm introduces mutations intentionally.

When to Intervene (and When Not To)

The most common mistake is intervening too early. Let the algorithm run for at least 5-10 generations. Do not manually fix the layout to your "preferred" option. Valid reasons to intervene: significant product catalog changes, major seasonal events, or large batches of new reviews. In these cases, reset the algorithm from your dashboard for a fresh start with new data.

Wrapping Up

Genetic optimization is what makes Eevy AI fundamentally different from other review apps. Instead of testing one thing at a time, it explores a vast space of possibilities simultaneously and evolves toward the best configurations for your specific store. Trust the process, give it time, and focus on RPV as your north star metric.

Free up to 25,000 monthly visitors