Retail & E-Commerce

Models in the live behavioural stream.

Pricing, churn, and personalisation models running on live behavioural data, in one engine, with a Python bridge for the models you already maintain.

500+Statistical and data functions, native to the live pipeline
~30μsRust/Python cross-process roundtrip for the ML models you maintain
<10μsEngine overhead per pipeline step
1Engine for the data pipeline and the models
The Problem

Models served through a separate path.

Pricing, churn, and personalisation models are typically built in a data-science environment and served back to the application through a separate path. Keeping them running on live behavioural data, rather than on periodic extracts, usually means another system to integrate and operate.

Traditional streaming tools move the behavioural data, but the models run elsewhere, and the data crosses a boundary to reach them.

[ diagram: behavioural stream → extract → model-serving path ]
One Engine

Run the models on the stream.

Lightning runs the statistical models and the data pipeline in one engine, on live behavioural streams, with a Python bridge for the machine-learning models a retail team already maintains. Scoring and segmentation update as the customer engages, computed where the events arrive.

Sub-millisecond latency is available where it helps. For most retail workloads the advantage is the unified engine and the full statistical and ML support rather than the microseconds.

[ diagram: behavioural events → model → decision, one engine ]
Use Cases

Where it runs.

ICON

Dynamic pricing

Prices computed against live demand and inventory signals as they move.

ICON

Live churn scoring

Churn probability updated on behavioural signals while the customer is reachable.

ICON

Real-time personalisation

Segmentation and recommendations computed as the session unfolds.

ICON

On-site fraud

Account and transaction anomalies scored on the live event stream.

Scope

What it does, and where it stops.

Lightning runs the live decisioning and analytics - the statistical models and the data pipeline in one engine, on your behavioural streams, with a Python bridge for the models you maintain.

It is the decisioning and analytics engine, rather than a customer-data platform or a recommendation service. It computes the scores and decisions and routes them to the systems that act on them.

FAQ

Common questions.

Does it replace our CDP or recommendation engine?

No. Lightning is the live decisioning and analytics engine. It runs the models and routes the scores and decisions, alongside your customer-data platform and recommendation services. Recommendations are not native to Lightning, though it can run them through the Python bridge where needed, and its linear-algebra support is there for teams building their own.

Can our data scientists keep their Python models?

Yes. Your pricing, churn, and personalisation models run inside the live pipeline through the Liquid bridge, unchanged. Your team stays in Python, and the engine runs it live.

How does it fit our existing data and model stack?

Lightning is written in Rust and calls any standard REST API or integration endpoint, so it runs beside your data and model stack and routes results to the systems you already operate.

Do you need our customer data?

No. Your customer data stays inside your own environment. Point a live behavioural feed at Lightning and it processes it in place, scoring and personalising while the customer is still on the page. You keep the data, and the immediacy is the point.

Book a demo.

A confidential technical walkthrough. We respond within one business day.

Book a demo