Turn your Microsoft Clarity data into 5 actions, with Claude.
Clarity shows you how people actually use your site: where they scroll, click, rage-click and quietly leave. This prompt turns Claude into your analyst. Connect the Microsoft Clarity MCP, paste it, and get a full read of your data plus five prioritised things to fix this week.
Traffic, engagement time, scroll depth and every friction signal, pulled across device, browser, country, source and page.
Plain-language synthesis
What is actually going on: which sources engage, which just bounce, and where people get stuck, in words, not jargon.
Visuals at a glance
Traffic by source, device and browser splits, and per-page engagement, scroll and friction charts you can read in seconds.
Five ranked actions
Exactly five specific fixes for this week, each tied to a real Clarity signal on a real page, ranked by impact.
How to use it
Three steps, about five minutes.
Step 1
Connect Clarity to Claude
Add the Microsoft Clarity MCP connector in Claude and authorise it with your Clarity project. You generate the API token inside Clarity under Settings.
Step 2
Paste the prompt
Copy the prompt below and paste it into a Claude chat that has the Clarity MCP connected. Nothing to fill in.
Step 3
Read the analysis
Claude pulls your last three days of Clarity data across every dimension, charts the key findings, and hands you five prioritised actions.
The prompt
Copy this, paste it into Claude with Clarity connected.
Clarity analytics strategist prompt
You are my web analytics strategist. I have connected the Microsoft Clarity MCP.
Pull a complete, holistic picture of my Clarity data, visualise it, and give me
5 prioritised actions.
HOW THE DATA SOURCE WORKS
The Clarity export returns at most the last 3 days and allows at most 3
dimensions per request. Make several separate calls and stitch the results
together. Do not stop after one call. Always request the maximum window
(numOfDays = 3). First check which tools and parameters the connected Clarity
MCP actually exposes, then use them.
STEP 1: PULL EVERYTHING
Fetch every available metric: total traffic (sessions, distinct users, page
views, pages per session, bot traffic), active and total engagement time,
average scroll depth, and every behavioural signal Clarity exposes (dead
clicks, rage clicks, quick backs, excessive scrolling, JavaScript errors).
Break these down across every dimension, three at a time per call: Device,
Operating System, Browser, Country/Region, Traffic Source, Medium, Campaign,
Channel, and Page/URL.
Prioritise a per-page (URL) breakdown of engagement, scroll depth, and every
friction signal, because that is where fixes actually happen.
If a metric or dimension is unavailable, say so plainly. Never fabricate
numbers or fill gaps with estimates.
STEP 2: SYNTHESISE HOLISTICALLY
In plain language, tell me:
. The headline numbers and the exact date range they cover.
. Where traffic comes from, and which sources turn into real engagement
versus which just bounce.
. Which devices, browsers, and countries perform best and worst.
. Top pages by traffic, and for each: how deep people scroll and how
engaged they are.
. The biggest friction points: where rage clicks, dead clicks, quick backs,
excessive scrolling, or errors cluster, and on which pages.
. Anything surprising, broken, or anomalous.
STEP 3: VISUALISE
Produce clear visuals of the key findings, not just raw tables. At minimum:
traffic by source, a device and browser split, a per-page engagement and
scroll-depth comparison, and a friction-signals-by-page chart. Make it
readable at a glance.
STEP 4: FIVE ACTIONS
Give exactly 5 prioritised, specific actions I can take this week to lift
engagement or cut friction. For each one, name the page or segment it applies
to, the exact Clarity signal that justifies it, the change to make, and the
outcome to expect. Rank them by impact. No generic advice.
No fields to fill in. It works as-is, as long as your Claude chat has the Microsoft Clarity MCP connected.
Want another one? Our Primary Inbox prompt helps any email land in the inbox instead of spam or Promotions.
Why it works
It reads your data the way an analyst would.
It pulls everything, not one report. Clarity's export caps each call at three days and three dimensions, so the prompt makes many calls and stitches them together instead of stopping at the first one.
It works per page. The synthesis is anchored to real URLs, because that is where a fix actually happens: this page has rage clicks, that page loses people at 40% scroll.
It never fabricates. If a metric or dimension is not available, it says so plainly rather than filling the gap with a made-up number.
It ends in decisions. Not a dashboard to interpret, but five specific, ranked actions with the signal that justifies each one.
Good to know: Clarity only returns roughly the last three days per export, so run this as a weekly habit to catch changes over time. And Clarity tells you how people behave, not why, so pair the biggest findings with a couple of session recordings to see the moment for yourself.
Rather not do this yourself?
We read the data and fix the page for you.
A prompt tells you where the friction is. Acting on it, week after week, is the work: rewriting the page that loses people, fixing the dead click, cutting the step where they quit. That is what we do for clients, across the whole marketing system, not just one page.