All platform pillars

Platform · Infrastructure

Personalisation & Experimentation Engine.

The right experience for every visitor, without code changes.

The thesis

Serving the same site to every visitor is leaving money on the table. Running experiments by hand is too slow to catch up. The Personalisation & Experimentation Engine collapses both into one substrate, so agents can target, test, and learn at a cadence no human team can match.

What it does

The capabilities inside Personalisation & Experimentation Engine.

Each capability is wired into the same data model and shared by every agent. Nothing here is a side feature; the platform is the product.

01

Targeting

Segment by source, geo, device, behaviour, lifecycle stage, predicted value. Composable rules, no engineering tickets.

02

A/B and multi-arm

Built-in statistical guardrails, auto traffic allocation, auto winner calls, auto archive. Experiments graduate themselves.

03

Variant generation

Creative Agent generates variants natively. The engine deploys, measures, and rotates them without an operator in the loop.

04

Content personalisation

Headlines, hero imagery, recommendations, pricing, promos. Anything on the page is targetable, no code changes required.

05

Continuous experiments

Experiments never stop. As one ends, the agent spins up the next, building a stack of compounding wins.

What we believe

Why we built it this way.

Every architectural decision here reflects a strong opinion about what makes AI useful inside a real growth team, not what makes a demo land.

Belief 01

The engineer was always the bottleneck.

Most experimentation tools assume an engineer ships the change. We assume an agent does. That single shift is what makes experiment-per-week velocity possible.

Belief 02

Personalisation is a feedback loop, not a feature.

Rule-based personalisation rots the moment the catalogue or audience moves. Agent-driven personalisation re-tunes itself as the data shifts.

Belief 03

Statistical rigor is non-negotiable.

Calling a winner on noise is worse than not testing at all. The engine refuses to graduate variants until the math is honest.

How the agents use it

One pillar. Five agents compounding on top.

Personalisation & Experimentation Engine is shared infrastructure. Every agent reads from it, so the work each one ships is consistent with everything else moving through the store.

In one line

Most brands run two or three meaningful experiments a quarter. With this engine, the floor is two or three a week.

How the platform behaves

Four principles every pillar inherits.

Memory over prompts

Agents read from structured, retrievable context. Prompts can lie about what an AI knows. A graph cannot.

One source of truth

Every agent queries the same data layer. No copies, no divergence, no silent disagreements between tools.

Continuous, not one-shot

The platform tunes itself as the catalogue, audience, and market move. Static defaults rot fast.

Auditable by design

Every action ties back to the context it read and the data it saw. Nothing the agents do is a black box.

See Personalisation & Experimentation Engine on your store.

A 20-minute walk-through. Real catalog, real data, the platform running against a duplicate of your live store.

Join the Waitlist