The questions we get asked first.

Deep learning neural networks are machine-learning systems modelled loosely on the architecture of the human brain: layers of interconnected 'neurons' that learn patterns directly from raw data, without anyone telling them in advance what to look for. They power most of the AI you've used recently: large language models, image generators, recommendation engines, self-driving cars. They're now the new frontier of marketing and customer engagement. They surface customer cohorts and behavioural patterns that rules-based and statistical segmentation simply can't see.

RFM, K-Means, and GMM are all inherently human approaches to segmentation that require defining up front. They're excellent at finding what you're already looking for, but not so good at discovering new things about your customers. Neuralift performs unsupervised learning on your data, surfacing non-intuitive correlations, cross-domain insights, and emergent properties that traditional segmentation simply can't find.

Yes! Not only does Neuralift handle the complexities and cost of running the deep learning so that you don't have to, but we also run the outputs through frontier models to provide plain language definitions, context, and more. This means non-technical users can benefit without the need for a PhD in data science.

No! It's important to only have one user ID per row, and not to bring PII into Neuralift. Other than that, our team and tools will help you to structure your data so that it's ready for the deep learning process.

Breadth is more important than depth, and we've worked with customers with tables as small as 3,000 rows to great effect. Our largest run to date totalled 100 billion data points.

We don't provide third-party data ourselves, but if you've already enriched your warehouse with additional attributes, these can be really valuable ingredients in the deep learning process. Many of our customers licence external data from the likes of Experian or Acxiom, or bring in flags from other systems like DMP segment names, to great effect.

Neuralift provides a number of ways to extract the data and insights from our platform post-discovery. Individual segments can be exported from within the UI, but we also provide warehouse-level table exports of fully enriched data for common destinations like Snowflake, Databricks, Azure, and AWS. Every user ID is assigned a new segment ID and segment name and has all the additional context appended for use in your advertising or agentic activation workflows.

We can be live with discovery in weeks, not months. Different use cases may require different discovery cadences, and many of our customers repeat the deep learning process weekly, monthly, or quarterly once up and running.

We operate a consumption-based pricing model. No strict lock-ins or annual renewal cycles. Customers buy bundles of deep learning credits based on their use case and data volume.

We hold ourselves to enterprise security standards including SOC 2, but also hold ISO 42001 certification alongside this. Security reviews and DPAs handled during onboarding. For our latest certifications and security posture, visit our Trust Centre ↗.