How Particle Transformed CRM Marketing With GenAI, and Gained a 35% Increase in Revenue

+34.8%

Revenue per recipient

(up to +195% in best campaign)

+23.2%

Click-through rate

(up to +92% in best campaign)

Approach

Particle partnered with Eikona to add an adaptive marketing layer on top of its existing CRM stack. The team stayed on Klaviyo - integration took minutes, with no migration - while the process itself transformed: daily blasts and occasional A/B tests gave way to adaptive campaigns steered by reinforcement learning, with commercial impact visible as early as the proof-of-value period. Production-ready, data-driven creatives: Eikona generates 20 production-ready emails per campaign - every one on-brand and pixel-perfect on the product - with variations probing different intents and messages, informed by everything learned from previous sends. Dynamic campaigns instead of blasts: contextual bandits serve every user creatives based on their first-party data and what they've reacted to - the way targeting already works on acquisition channels - so delivery decisions are completely data-driven, made in real time rather than guessed in advance. Continuous learning: the system learns from every winning variation, so each campaign starts smarter than the last, and winners and the concepts behind them are analyzed and redeployed across paid, social, and other channels.

Business Challenge

Blue checkmark inside a blue circle on a white background.

Scale creative production, segmentation, and testing far beyond what a manual calendar allows.

Blue checkmark inside a circular arrow indicating confirmation or completion.

Unlock the revenue hiding inside an already high-performing retention channel.

Blue circle with a check mark inside signifying confirmation or success.

See defensible, control-measured impact fast — without replatforming away from Klaviyo.

Two large dark quotation marks on a white background.

We've heard every 'AI for email' pitch there is, so we demanded proof - every campaign against a live control, and the uplift held up under every cut we took at it. But the real shift runs deeper: we've finally activated our first-party data - users get the creatives they actually respond to, the way our acquisition channels have worked for years. Every send is a test, every test becomes an insight, and the program keeps getting smarter.

Alon Turchin

VP Retention, Particle

Key Takeaways

Adaptive creative and reinforcement learning made a strong retention program measurably stronger - and every send raises the baseline for the next.

Proof, Not Promises

(up to +195% in best campaign)

A Layer, Not a Migration

Eikona plugged into the existing Klaviyo stack in minutes, upgrading the process without touching trusted infrastructure.

First-Party Data, Finally Activated

Users receive creatives based on what they've reacted to, and every winning concept feeds the next send - and the next channel.

Campaign Case 1:

Campaign Example 1

In this campaign, Eikona discovered that the same product message performed differently across geographic audiences. Rather than assuming that one visual direction would resonate equally nationwide, the engine explored distinct production-grade approaches, from an ingredient-led product composition to a bold, urgency-driven design, while preserving Particle's brand language, product accuracy, and visual standards.

Reinforcement learning measured how customers in each region responded and progressively shifted traffic toward the creative most likely to drive engagement. The results showed that no single creative direction was universally optimal: the ingredient-led concept performed best globally, while the bolder, urgency-driven direction generated stronger CTRs across the Northwest, Central, and Mountain regions.

+18.9%

RPR

+42.2%

CTR

Geographically adaptive Underoutfit campaign example

Campaign Case 2:

Campaign Example 2

In this campaign, Particle's Easter Sale was transformed from a single promotional send into a dynamic campaign testing multiple ways to communicate the same offer. The engine explored distinct creative languages, including a premium Easter egg concept, a clean product-led composition, a seasonal lifestyle scene, and the existing control, while keeping the offer, code, and product range consistent.

Every new variation was system-generated and production-grade, preserving Particle's brand identity, product accuracy, and campaign messaging. With every send, reinforcement learning measured live audience response and progressively shifted delivery toward the creative framing most likely to drive engagement and revenue.

+113.5%

RPR

+48.7%

CTR

Intent-driven Underoutfit activewear campaign example