HTML dashboards with Claude: more flexible than a BI tool

An HTML dashboard is a web page with analysis and slides that Claude builds from your marketing data and that you host on Cloudflare. Changing a chart or adding a metric such as CAC, CLV or a cohort analysis takes one prompt and is online seconds later. A classic tool such as Data Studio is the better fit when many people build reports themselves.

As a Google Analytics agency, we build such dashboards with Claude. Below you will read why, how to define the key metrics correctly, when a classic tool is still the better fit, and how to protect the data.

Why we build dashboards in HTML

The biggest gain is flexibility. A BI tool limits you to its charts, connectors and calculated fields. An HTML page can show anything you can describe, in your brand's house style.

What that means in practice:

• Changes in seconds. An extra column, a different period or a new metric is a prompt, not an evening of dragging in a report builder.
• Advanced metrics within reach. For Claude, CAC, CLV and cohorts are arithmetic, not a separate module.
• Cheap hosting. Requests to static files are free and unlimited on Cloudflare ^[1].
• Test first. A new version gets its own URL before it goes live ^[2].

Our route: Coupler.io pulls in the data on a fixed schedule, Markifact opens up the ad platforms, and via MCP Claude reads both. Claude builds the page and publishes it on Cloudflare; we sometimes share an interim version as a Claude artifact.

What else you need: fixed agreements on the definitions and someone who reviews. The dashboard also follows your design system, so report, site and decks form one brand. How that works is covered in a design system for Claude.

Define CAC, CLV and cohorts before you show them

Claude calculates what you ask. If the definition is vague, the figure shifts with every change. So set down the formula for each metric, in the prompt and visibly in the dashboard.

The simplest formula for CLV is average transaction value times average number of transactions times customer lifespan ^[4]. Shopify calls a CLV to CAC ratio between 3:1 and 5:1 efficient ^[3]. What clicks and campaigns cost, and therefore what feeds your CAC, is covered by Benjamin in how much does Google Ads cost.

When Data Studio is the better fit

Google's Data Studio is free for people who create and view reports, and connects directly to more than 1,400 data sources ^[6]. Until recently it was called Looker Studio. The paid Pro version adds team workspaces, and reports owned by the organisation instead of a single user ^[7].

Choose a classic tool if:

• many people want to build reports themselves, without someone managing the code;
• you want live connectors without your own data feed to the dashboard;
• ownership and management of reports in a large team matter more than design freedom.

Choose HTML if you want your own metrics and combined sources, have a fixed group of readers, such as management or a client, and change things often.

Do count on maintenance, though. New data has to come in on a fixed schedule, and someone has to review every change. That is the real price tag of an HTML dashboard, not the hosting.

Protect your data

A static page can be read by anyone who knows the link. Test versions on Cloudflare are public by default too; you protect them with Cloudflare Access ^[2]. Access denies everyone by default: only those who meet an access rule get in, and every request is checked for a valid token ^[8].

Three rules that always apply:

• Only summarised figures, no names, email addresses or customer numbers.
• No keys or passwords in the page. Everything in an HTML file can be read in the source code.
• Accounts in the company's name, with two-step verification on the hosting and the code.

An interim version as a Claude artifact is visible only to you at the start ^[9]. How the links in our stack work together and which permissions each gets is covered in our AI marketing stack.

How to build one

A first version is up faster than the meeting to discuss it. The order:

1. Choose five to eight KPIs and write out the formula for each.
2. Have the data come in summarised, on a fixed schedule.
3. Ask Claude for the page, with your design system and your definitions attached.
4. For each metric, check one figure against the source. Why GA4 and Google Ads can differ there is explained by Cédric in why GA4 and Google Ads show different numbers.
5. Publish first as a test version or as a Claude artifact, switch on Access and then go live.

A dashboard is only as good as the measurement behind it. If you have doubts about that, start with the free tracking audit.