Ask your analytics anything. Trust the answer.
Point an LLM at incomplete, cookie-based analytics and it invents confident, wrong answers. SealMetrics is AI analytics built the other way round — complete cookieless data, a semantic MCP the model can't misread, and private AI hosted in the EU. Ask Claude or ChatGPT your revenue questions and get a number you can defend.
Private AI · EU-hosted · 0 PII
The single source of truth eCommerce signs against
Dreamplace Hotels recovered +30% more traffic vs GA4 and closed a 15–20% gap in sales attribution against their CRM. Palladium Hotel Group recovered 35% of unattributed bookings and improved Display CPS by +165%.
An LLM is the easy part. Trustworthy answers are the hard part.
Six properties separate AI analytics you can act on from a chatbot guessing over a warehouse. We wrote the long version in how SealMetrics ships self-service analytics.
Complete data
AI is only as good as its input. Cookieless, first-party, 100% of traffic, never sampled — so every answer describes the whole of your traffic, not the fraction that accepted a banner.
A semantic MCP
40+ named, read-only tools, each mapping one business concept to one canonical metric. The model calls a documented contract instead of guessing at raw columns. Nothing to misread.
Private AI, in the EU
LENS private AI runs inference on an open-source model (Gemma) hosted by Scaleway in Paris, while your data stays in Dublin. Never leaves the EU, never shared, never trains third-party models.
Choose the algorithm
Use SealMetrics private AI, or bring your own Anthropic, OpenAI or Gemini key, or connect the hosted MCP from your own client. Same complete data underneath — you only pick the model.
Zero PII, honest answers
Aggregate-only measurement: no per-user journeys, no cross-session identifiers, no multi-touch models. The model cannot fabricate a person because none was ever stored.
Answers you can trace
Every answer resolves to a named tool over an explicit period in your timezone, and attribution is last-click at the event level — so you can always see which metric produced the number.
Generic AI on a warehouse vs AI-native analytics.
| Property | LLM + raw warehouse | SealMetrics LENS AI |
|---|---|---|
| Data foundation | Cookie-based, consent-gated (partial) | Cookieless, 100% captured |
| How the AI reads it | Raw SQL against unseen tables | Semantic MCP — named tools |
| Wrong-but-plausible answers | High — a staging column looks real | Structurally constrained |
| Personal data exposure | Possible — raw rows carry PII | None — 0 PII by construction |
| Where inference runs | Warehouse creds to a US model | Private AI in the EU, or your key |
| Data residency | Mostly US | EU — Dublin + Paris |
| Attribution | Whatever the schema encodes | Last-click on 100% of traffic |
| Time to first answer | Build pipelines + dashboards | Connect MCP, ask in words |
Your data answers your questions — inside the EU.
LENS AI is the private-AI layer over your cookieless analytics. Inference runs on an open-source model (Gemma) hosted by Scaleway in Paris; your analytics data stays in Dublin. Both in the EU — never shared, never used to train third-party models. Prefer your own stack? Bring your Anthropic, OpenAI or Gemini key instead.
- Connect Claude, ChatGPT, Cursor or Claude Code via MCP at mcp.sealmetrics.com
- 40+ read-only tools: overview, channels, conversions, funnels, landing pages, properties
- Private AI on Gemma / Scaleway (Paris) or bring your own model key — you choose the algorithm
- Enterprise and Corporate can get a dedicated, non-shared private AI instance
- Data is complete before 6 AM every morning; dashboards update live when it matters, like a promo-day spike
> Sales look soft this week versus last. What happened, and where should I look first?
Traffic held — this is a conversion problem, not acquisition. It's isolated to your /checkout page, where the rate fell mid-week and dragged direct conversions with it, while paid search was unaffected. Revenue loss concentrates in a single category. Start with the checkout page, not the media budget.
One AI layer, every role.
Ask why ROAS moved, get an answer
“Why were sales soft this week?” returns a diagnosis — traffic vs conversion, which channel, which landing page — grounded in complete data, not a modeled estimate.
Revenue by SKU, in plain language
Drill into category, variant and campaign revenue by asking, not by building a report. Every item is queryable — no sampling thresholds hiding the long tail.
An AI layer compliance signs off on
Private AI hosted in the EU, zero PII in the data, no third-party model training, a DPA included. The AI never sees a person because none was ever stored.
Ship a Claude analyst in an afternoon
Connect the SealMetrics MCP to Claude, ChatGPT, Cursor or Claude Code and let it answer revenue questions directly — the same server can even provision a new site from the chat.
The compliance case rests on architecture, not a certificate — GDPR by design, ePrivacy-clean, EU-hosted in Dublin, DPA included. For the full posture, see security & compliance and how it holds up as GDPR-compliant analytics.
The questions teams ask about AI analytics.
Honest answers about AI analytics, private AI, MCP, and what has to be true for an LLM to answer your revenue questions without making them up.
Still have questions? Our team — including the founder — is one message away.
Talk to us →Give your AI complete data. Then trust what it tells you.
Book a demo and ask LENS AI a question about your own traffic. Complete cookieless data, private AI in the EU, zero PII. Thirty minutes, on your numbers.
Built by a founder · supported by a founder · EU-hosted by design