Better decisions begin
when the evidence
survives questions.

Not a vendor list. The 12 technical requirements a web analytics platform must meet — pixel weight, real time, 100% of the data, API, MCP — and how to test each one.

Matomo can run banner-free in France under the CNIL criteria, with cookies still on. The six conditions, what the exempt configuration costs you in attribution, and why it does not travel to Germany.

Adobe Analytics can be deployed compliantly under GDPR — consent, a DPA, a transfer assessment and deliberate configuration. What that costs you is not legal, it is 40–60% of your EU visitors.

Does a Spanish online store need a cookie banner for analytics? Is GA4 legal in Spain? What does the AEPD allow, and what does LSSI non-compliance cost? Direct answers, one per question.

GDPR and ePrivacy are two different laws, and analytics has to clear both to run without a consent banner. The legal test, plus a per-tool verdict for GA4, Matomo, Plausible, Piwik PRO and Sealmetrics.

Point an LLM at incomplete GA4 data and it invents answers. How complete cookieless data plus the Sealmetrics MCP let teams query their own analytics — no analyst in the loop.

How European eCommerce teams measure revenue, attribution and conversion without cookies or consent banners. Shopify, WooCommerce and Magento reconciliation patterns.

How hotel groups measure direct bookings, meta-search revenue and multi-property portfolios without cookies. PMS reconciliation patterns for Mews, Cloudbeds, Opera.

How European SaaS teams measure trial-to-paid, PQLs and self-serve funnels without cookies. Works alongside Mixpanel/Amplitude, exports to BigQuery.

Consentless analytics captures 100% of DTC traffic without a cookie banner. How it works under GDPR/ePrivacy, and what European DTC teams get from it.

Ranked comparison of GA4, GA360, Adobe Analytics, Piwik PRO, Sealmetrics, Amplitude, Mixpanel, and Matomo for enterprise teams.

GA4 loses data at three levels: consent rejection, ad blockers, and browser restrictions. The result is roughly 13% of real EU traffic. Here is the math.

We cataloged every cookie from 9 analytics tools using vendor docs and DevTools. GA4 sets 2 first-party cookies. Adobe sets 6. Sealmetrics sets 0.

Every external domain is a DNS lookup, a privacy risk, and a GDPR liability. We mapped every domain contacted by 9 analytics tools.

GA4 starts sampling at 10M events in Explorations. Adobe varies by contract. Piwik PRO: optional. Sealmetrics: never. We documented every threshold.

Heavy analytics scripts slow your site, consent banners hide visitors, and ad blockers erase data. The compound effect is costing you conversions.

We measured major analytics scripts from production CDNs. GA4 was 132x heavier than Sealmetrics, the lightest of the ten tools tested.

The DUAA 2025 exempts certain analytics from consent requirements in the UK. We published our self-assessment showing how Sealmetrics qualifies.

GA4 applies data sampling when traffic exceeds certain thresholds. Here is how it works, why it matters, and what you can do about it.

The European Commission proposed the biggest change to EU data law since GDPR. Cookie consent moves to GDPR, and first-party analytics may not require consent.

Cookie banners could vanish for 60% of websites. First-party analytics gets explicit legal authorization. A deep dive into COM(2025) 837 and what to do now.

The French CNIL allows certain analytics tools to operate without cookie banners. We documented how Sealmetrics meets all 14 technical criteria.

Cookies are disappearing. Learn how cookieless analytics works, why it captures more data, and what it means for GDPR compliance.

35% of EU visitors reject cookies. That means 35% of your traffic is invisible. Here is the real impact on attribution, conversions, and revenue reporting.

GPT, Claude, Perplexity, and Google AI Overviews are sending traffic to your site. Traditional analytics cannot see it. Here is why it matters.

Your attribution model is only as good as the data feeding it. When 87% of touchpoints are missing, every attribution conclusion is wrong.

Cookieless tracking captures visitor behavior without browser cookies. Learn how it works, why it matters for GDPR compliance, and how it compares to traditional analytics.

Analytics data loss is the gap between real traffic and reported traffic. Learn the four causes, quantify the impact, and understand how to eliminate it.

Compare 7 GA4 alternatives for eCommerce: pricing, data capture rates, EU compliance, and eCommerce features. From GA360 to cookieless platforms.

Consent banners are not always required for analytics. Learn the legal basis, technical requirements, and which analytics tools qualify under GDPR and ePrivacy.

A European datacenter does not make your AI vendor European. The US CLOUD Act follows the company, not the server. How to tell the difference — and why it decides where your data can be reached.

We ran 162 live queries across three open models against real analytics data to pick the AI inside Sealmetrics. What decided it wasn't MMLU — it was tool-calling, grounding, and a security flaw that only showed up in one language.

Ask your analytics questions in plain language and get grounded answers — with inference that runs in the EU only, retains nothing, and trains no one's model. How Seal AI works and why it's private by architecture, not by promise.

Five questions that separate a private AI vendor from one that just has a European datacenter — with what a good answer and an evasive answer look like for each.

Safe Harbor fell. Privacy Shield fell. The Data Privacy Framework is under appeal at the CJEU with a fresh challenge announced. Which analytics setups would need re-papering overnight — and which are structurally immune.

Most privacy work for AI is damage limitation, because the data was personal to begin with. If your analytics never collects IPs, cookies or visitor identifiers, there is nothing personal to send to the model in the first place.

What actually lands on a marketing team using AI analytics: provider vs deployer, the Article 50 transparency duty from 2 August 2026, and why open-weight models are easier to audit than a closed API that can change silently.

Gemma 4 looped. Mistral Small refused things it shouldn't. gpt-oss-120b held. The full lineage of how we picked the model inside Sealmetrics, including the two picks we got wrong.

A methodology you can copy: test on the real stack, compute ground truth from your database, grade deterministically before you let any model judge, and archive the runs you discard. A benchmark that hides those is a demo.

A model asked for a comparison returned a chart with two y-axis keys instead of one. Strict validation rejected it and the whole answer died with it. Three rules for anyone shipping LLM structured output.

The same injected instruction was ignored in every Spanish run and obeyed in English. Safety alignment is trained unevenly across languages — and a monolingual evaluation will certify a vulnerable model as safe.

LLM judges favour their own family. So we gave the scoring pen to our winner's direct competitor, kept it away from anything objective, and published the limitation we couldn't fix.

MMLU measures knowledge in a vacuum. Your product needs tool-calling, grounding and latency under real tool loops. How to read agentic benchmarks' footnotes — and a five-step method to evaluate a model yourself.

A general chatbot answers from memory. An analytics assistant must answer only from data fetched at query time. Why the model we ship scores badly on world knowledge — and why that is exactly right.

With a closed API you rent behaviour you cannot inspect, that can change under you silently, at prices set unilaterally. With open weights you can pin the version, move hosts, or run it yourself.

Seal AI is the default so privacy doesn't depend on a checkbox. But you can bring your own OpenAI, Anthropic, Gemini or DeepSeek key. An honest guide to what each option costs you — including the data-transfer analysis you inherit.

A period comparison, a drill-down with reasoning, and an engagement check — what to ask, what happens under the hood, and how to verify the answer came from your data.