# MetricMind

> Privacy-first analytics for websites and apps, with a built-in AI analyst that explains what changed in traffic, contacts and crashes, and what to do next.

MetricMind is in private beta and free during the beta. It collects traffic, contacts, site search, Web Vitals, errors and app crashes without cookies, and a built-in AI analyst explains what changed with the query behind every number. Data is hosted in the European Union.

## Pages
- [Product and waitlist](https://metricmind.si/): what MetricMind does, pricing and FAQ
- [AI analyst for web and app analytics](https://metricmind.si/ai-analyst): An AI analyst built into your analytics: plain-word answers with the query behind every number, morning briefings, and checks that your fixes worked.
- [Mobile app analytics for iOS, Android and Flutter](https://metricmind.si/mobile-app-analytics): Privacy-first app analytics for iOS, Android, React Native and Flutter: screens, events, crashes and release health beside your website. No IDFA or ATT prompt.
- [Track ChatGPT, Claude and Perplexity traffic](https://metricmind.si/ai-traffic): See the visitors ChatGPT, Claude, Perplexity, Gemini and Copilot send you, whether they convert, and which AI crawlers (GPTBot, ClaudeBot…) read your pages.
- [Cookieless web analytics, hosted in the EU](https://metricmind.si/cookieless-analytics): Analytics with no cookies, no fingerprinting and no stored IP addresses, hosted in the EU: traffic, contacts, search, Web Vitals and errors in one 9 KB script.
- [Documentation](https://metricmind.si/docs): install on websites and apps, events, users, feature flags, query API, server-side events, webhooks, MCP
- [Changelog](https://metricmind.si/changelog): what shipped, newest first
- [Live demo](https://app.metricmind.si/share/site_demo): a public dashboard with real analyst answers on sample data
- [Status](https://app.metricmind.si/status/s_b9tRGxQGuyvu)
- [Privacy policy](https://metricmind.si/privacy)

## The analyst

Dashboards wait for you to notice. The analyst notices first, checks the obvious explanations, shows its work and only speaks up when a change is real. Ask it anything in plain words, or let it brief you every morning.

### Ask in plain words

Which campaign brought visitors who stayed? Why did contacts drop on Android? Answers come with the query behind every number, a chart when a trend is clearer as a picture, and follow-up questions in one click.

### A morning briefing

At your local hour it compares yesterday with the same weekday and the week before: traffic, contacts, goals, errors, Web Vitals and frustrated visits. Normal swings and tiny numbers stay quiet; real changes get investigated.

### Investigations that start from facts

Before the model says a word, code works out which source, page, country or device carries a change, the hour it started and which error is behind a rise. Every number in a finding is checked against the data before it's saved.

### It checks that the fix worked

When it recommends a fix, it adds a measurable check, like errors on /sell below 1 a day. A few days later MetricMind measures again and tells you: fix confirmed, or still not fixed.

### Questions on a schedule

Any question can be answered again every morning or every Monday. Quiet mode only writes when a threshold is crossed or something needs action.

### Follows your team's rules

It learns your business from your site, and you tell it what counts as a lead or which traffic to ignore. Every answer lists its assumptions; turn one into a rule with a click.

### Acts only when you click

When the next step is to measure or test something, it drafts the goal, funnel, alert, timeline note or switched-off feature flag as a card. Nothing changes until you click.

### Where your team already is

Findings and alerts go to email, Slack and signed webhooks. Claude, Cursor and other MCP clients can query the same metrics and ask the analyst from your editor.

## What it measures

### Traffic and sources

Pages, referrers, UTM campaigns, ad clicks by network, countries and devices, counted per visit. Page templates turn thousands of listing URLs into one page.

### AI visitors and crawlers

Visits from ChatGPT, Claude, Perplexity and Gemini, even without a referrer, whether they contact you, and which AI crawlers read your pages.

### Web Vitals

LCP, INP, CLS, FCP and TTFB from real visitors, by page type and device, with one experience score next to Google's pass or fail.

### Errors, rage and dead clicks

Exceptions grouped by message and page, the buttons people click again and again, and the ones that do nothing. Each links to the visits that hit it.

### Contacts and searches

Phone, WhatsApp and email links and “Contact the seller” buttons count automatically. Site search shows what people look for, and what found nothing.

### Sessions and journeys

Follow any visit step by step, with time and scroll on each page, searches, frustrated clicks, unsent forms and contacts, plus the paths people most often take. Rebuilt from events: nothing on screen is recorded.

### Goals and funnels

Track key actions with one call or a data-track attribute. Funnels show where people drop, the time between steps and which sources complete them.

### Forms

Who started a form, who sent it, and the field people were on when they gave up. Field names only, never what anyone typed.

### Users and feedback

identify() merges a person across browsers and devices. Feedback from your own form arrives with the page, release and any error in the same visit.

### Your board

Pin any panel or any query behind an answer. The board re-runs live for the range you pick: trends as charts, rankings as tables, headline numbers as tiles.

### Feature flags and experiments

Roll out to a share of visitors or run an A/B test. Results are read on your goals, and no winner is called before each variant has enough visitors.

### Uptime and status pages

Checks every minute, TLS certificates watched, outages marked on your traffic chart, and a public status page with incidents.

### Alerts you write yourself

“Tell me when contacts go below 2 a day.” Visitors, errors, goals, events or frustrated visits, per hour or per day, by email, Slack or webhook.

Also included: Revenue from Stripe and Paddle, Server-side events, Saved segments and filters, Public dashboards and share links, Short links with click analytics, Query API and MCP server, Signed webhooks, Import from Plausible, PostHog, Simple Analytics, CSV export, Two-factor sign-in, roles, audit log, Exclude your team's visits, Light and dark mode.

## AI analyst for web and app analytics

Source: https://metricmind.si/ai-analyst (updated 2026-10-28)

MetricMind's AI analyst is built into the analytics, so it can query your traffic, contacts, searches, Web Vitals, errors and app crashes directly. Ask a question in plain words and you get an answer with the query behind every number. Every morning it checks what changed and investigates the changes that matter. It drafts goals, funnels and alerts, but nothing changes until you click.

### Questions you can ask

Ask the way you would ask a colleague who knows your data. The analyst picks the metrics, runs the queries and explains the result, with a chart when a trend reads better as a picture.

- Why did contacts drop on the Android app this week?
- Which campaign brought visitors who actually contacted us?
- Which page got slower since the last release?
- What do people search for that we don't have?
- Is the one-page form better than the old one, or is it chance?
- What changed since Monday?

### How an answer is built

- It runs read-only queries over your own site's data: pages, sources, campaigns, countries, devices, contacts, searches, sessions, goals, funnels, Web Vitals, errors, app releases and revenue.
- For “why” questions it splits a change across platform, source, page, country, device, browser and app release, and ranks what drove it.
- For “is it better” questions it runs a significance test and says plainly whether the difference could be chance.
- Every answer shows the queries behind it, lists its assumptions (which days “this week” covers, what counted as a contact) and offers follow-up questions.

### The morning briefing

At your site's local hour, statistics (not the model) compare yesterday with the same weekday and with the week before: visitors, pageviews, contacts, goals, errors, Web Vitals and frustrated visits. Normal swings and changes on tiny numbers stay quiet.

When a change is real, the analyst investigates it from facts worked out by code first: which source, page, country or device carries the change, the hour it started, and which error is behind a rise. The finding ends with a next step: act, ask, or keep watching. New sites get a first look on day one and a first-week report after seven days.

### Safeguards that keep it honest

| Safeguard | What it does |
| --- | --- |
| Query receipts | Every number in an answer links to the query that produced it. |
| Assumptions | Each answer and finding lists the judgment calls behind it; any of them can become a team rule in one click. |
| Number check | Every number in a finding is checked against the query results before it's saved; invented numbers go back to the analyst. |
| Fix verification | A recommended fix comes with a measurable check, re-measured a few days later: fix confirmed, or still not fixed. |
| Read-only access | Queries can only read, and only your site's data. |
| Click to apply | Goals, funnels, alerts, timeline notes and feature flags it suggests are created only when you click. |
| Untrusted text stays data | Page titles, search terms, event names and feedback written by visitors are treated as data, never as instructions; the analyst's test suite plants such instructions and checks they are ignored. |

### Where answers reach you

- Chat in the dashboard, with conversation history.
- Findings and alerts by email, Slack and signed webhooks.
- Scheduled questions answered every morning or every Monday, with a quiet mode that only writes when something needs action.
- Claude, Cursor and other MCP clients can query the same metrics and ask the analyst from your editor.

### Questions

**What data does the AI analyst see?**

Your question, your business context and the results of the queries it runs, which are mostly aggregated figures. It never receives identities: emails and traits sent with identify() stay out of its reach. The AI model provider processes these results only to write the answer.

**Can the analyst change my settings?**

No. It can propose a goal, funnel, alert, timeline note or feature flag (created switched off), and nothing is created until you click.

**What happens when the data can't explain a change?**

The analyst says so, names what it ruled out, and suggests what to measure so the next change can be explained.

**Can I teach it how my business works?**

Yes. MetricMind reads your site to learn what you do, and you can add rules such as what counts as a lead or which traffic to ignore. Every answer follows them.

**Can I use it from Claude or Cursor?**

Yes. MetricMind runs an MCP server: connect with your account (OAuth) or an API key, choose which sites the app may read and whether it may ask the analyst.

## Mobile app analytics for iOS, Android and Flutter

Source: https://metricmind.si/mobile-app-analytics (updated 2026-10-28)

MetricMind's SDKs for iOS (Swift), Android (Kotlin), React Native and Flutter measure screens, sessions, events, users, contacts, searches, errors and crashes, and send them to the same site as your website. Each app gets its own dashboard with release health per version, and one overview compares the apps with the website. There's no advertising ID, no fingerprinting and no App Tracking Transparency prompt.

### What each SDK measures

| Measurement | iOS | Android | React Native | Flutter |
| --- | --- | --- | --- | --- |
| Screens, sessions and engaged time | Yes | Yes | Yes | Yes |
| Events, users, contacts, searches, feedback | Yes | Yes | Yes | Yes |
| Crashes and errors | Crashes (MetricKit) | Crashes | JavaScript errors and crashes | Dart and Flutter errors |
| ANRs, native crashes, low-memory kills | No | Yes (Android 11+) | No | No |
| Hangs | Yes (MetricKit) | No | No | No |
| Cold start time | Yes | Yes | No | No |
| Screen load time | With screenReady() | Automatic | With screenReady() | Automatic |
| Slow and frozen frames | Yes | Yes | No | Yes |
| Campaign links (utm_*) that open the app | Yes | Yes | Yes | Yes |
| Consent mode (nothing sent until optIn()) | Yes | Yes | Yes | Yes |

### Release health

- Users, adoption, crash-free users and crash-free sessions for every version, newest first.
- Android ANRs as a share of users, against Google Play's 0.47% bad-behavior threshold, and iOS hangs.
- An alert by email and Slack when the newest release crashes more than the previous one, or crosses the ANR threshold.
- New versions marked on the timeline (“Android app 2.4.1 (41) released”), so a change explains itself.
- A crash sent after an update still counts against the version that crashed.

### Apps and website in one place

Apps send to the same site as your website, so a person who signs in on both is one user, and every report can be scoped to the website, the iOS app or the Android app with one switch.

App vs website compares pages or screens per visit, time spent, and how often a visit ends in a search or a contact, with a one-sentence verdict once both sides have enough visits.

### Privacy and the app stores

- No advertising ID, no device fingerprint and no permissions beyond internet access: under Apple's definition that isn't tracking, so no App Tracking Transparency prompt is needed.
- The iOS SDK ships Apple's required privacy manifest.
- The docs list exactly what to declare in App Store Connect and in Google Play's Data safety form, with a privacy-policy paragraph you can adapt.
- Unsent events stay on the device and are sent when the network returns.

### Install

Each SDK is one or two plain source files you add to your project, with no package registry needed. Initialise it once at launch, then name screens with one call (SwiftUI's trackScreen, UIKit's viewDidAppear) or let it follow your navigation: one screen per Android Activity automatically, React Navigation and Flutter named routes.

### Questions

**Do I need an App Tracking Transparency prompt?**

No. MetricMind uses no advertising identifier, no fingerprint and never combines your data with other companies' data, so under Apple's definition it isn't tracking. You still declare analytics and diagnostics data in App Store Connect.

**Does it work with Expo?**

Yes. The React Native SDK is a single TypeScript file that uses AsyncStorage for its queue, and it works in Expo apps, including updates shipped over the air.

**Can I see the app and the website together?**

Yes. They share one site: an overview shows every platform side by side and in total, and each platform has its own dashboard.

**Is mobile app analytics free?**

Everything is free during the beta, including apps.

## Track ChatGPT, Claude and Perplexity traffic

Source: https://metricmind.si/ai-traffic (updated 2026-10-28)

MetricMind recognises visits from AI assistants (ChatGPT, Claude, Perplexity, Google Gemini, Microsoft Copilot, DeepSeek, Grok, Le Chat and Meta AI) by their referrer or utm_source, and compares them with your other visitors: pages, contacts and conversions. With a few lines on your server it also counts AI crawler reads, such as GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot, by company and purpose, and checks your llms.txt, robots.txt and sitemap.

### AI assistants recognised

| Assistant | Recognised from |
| --- | --- |
| ChatGPT | chatgpt.com, chat.openai.com, or utm_source=chatgpt.com |
| Claude | claude.ai |
| Perplexity | perplexity.ai |
| Google Gemini | gemini.google.com |
| Microsoft Copilot | copilot.microsoft.com |
| DeepSeek | chat.deepseek.com |
| Grok | grok.com |
| Le Chat (Mistral) | chat.mistral.ai |
| Meta AI | meta.ai |

### AI visitors compared with everyone else

The AI agents page shows how many visitors each assistant sent, the pages they land on, and whether they contact you or convert more or less often than other visitors. The analyst can answer questions about them like any other segment, for example “do visitors from ChatGPT contact sellers more often?”.

### AI crawlers

Crawlers don't run JavaScript, so no analytics script sees them. MetricMind counts them on your server instead: a small helper reports each crawler request to MetricMind with your site's server key. Search-engine bots are recognised too, so you can compare.

| Crawler | Company | Purpose |
| --- | --- | --- |
| GPTBot | OpenAI | Model training |
| OAI-SearchBot | OpenAI | ChatGPT search index |
| ChatGPT-User | OpenAI | Fetches a page for a user's question |
| ClaudeBot | Anthropic | Model training |
| Claude-SearchBot | Anthropic | Search index |
| Claude-User | Anthropic | Fetches a page for a user's question |
| PerplexityBot | Perplexity | Search index |
| Perplexity-User | Perplexity | Fetches a page for a user's question |
| Applebot-Extended | Apple | Model training |
| Meta-ExternalAgent | Meta | Model training |
| CCBot | Common Crawl | Open dataset used for training |
| Amazonbot | Amazon | Search |

### AI readiness check

- Whether your site publishes an llms.txt file for AI assistants.
- Which AI crawlers your robots.txt allows or blocks.
- Whether a sitemap is published and reachable.

### Questions

**Why don't AI crawlers show up in normal analytics?**

Analytics scripts run in a browser, and crawlers don't run JavaScript. They have to be counted on the server, which is what MetricMind's crawler helper does.

**Does ChatGPT always send a referrer?**

No. Links from ChatGPT usually carry utm_source=chatgpt.com, which MetricMind reads even when the referrer is missing. Visits from apps that strip both look like direct traffic in any analytics tool.

**Can I see which pages AI crawlers read?**

Yes: reads per crawler, company, purpose and page, over any period.

**Does tracking AI visitors need cookies?**

No. AI visitors are counted the same cookieless way as every other visit.

## Cookieless web analytics, hosted in the EU

Source: https://metricmind.si/cookieless-analytics (updated 2026-10-28)

MetricMind measures visits without cookies, fingerprinting or stored IP addresses. One script of about 9 KB counts pageviews, sources, contacts, site search, scroll depth, Web Vitals and errors. By default, visitor and session IDs are random values in the site's own storage that visitors can clear; in storage-free mode nothing is stored on the device and visitors are counted with a hash that changes every day. Data is hosted in the European Union.

### What is stored, and what never is

| Data | Stored? |
| --- | --- |
| Page path (without the query string) | Yes |
| Referring domain and UTM campaign | Yes |
| Country, derived from the IP address | Yes (the IP address itself is never stored) |
| Browser, operating system and device type | Yes (the raw user agent is never stored) |
| Your own events and their properties | Yes |
| Cookies or device fingerprints | Never |
| IP addresses | Never |
| Form values, emails and phone numbers | Never |
| Ad click IDs (gclid, fbclid…) | Never; only the ad network is kept |
| Screen recordings | Never |

### How visits are counted without cookies

By default the script keeps a random visitor ID and session ID in the site's own browser storage. They identify a browser, not a person, are never shared with other sites, and disappear when the visitor clears site data or opts out with mm.optOut().

In storage-free mode (data-storage="none") nothing is stored on the device. Visitors are counted on the server with a hash of a random daily salt, the site, the IP address and the browser; the salt is deleted after two days, so no one can be followed from one day to the next, and the IP address is never stored. A visitor who returns on another day counts as new.

Optionally, with data-respect-gpc="true", a browser's Global Privacy Control or Do Not Track signal turns measurement off.

### What you still get

- Pages, sources, campaigns, ad clicks by network, countries and devices, counted per visit.
- Contacts (phone, WhatsApp, email and contact buttons) and site search, with no code.
- Sessions step by step, rage and dead clicks, forms people abandon, scroll depth.
- Web Vitals from real visitors and JavaScript errors.
- An AI analyst that explains what changed, and never sees identities.

### Where the data lives

Data is hosted in the European Union (Hetzner, Germany). Deleted websites are erased after 14 days, and you can export your data at any time.

### Questions

**Do I still need a cookie banner?**

MetricMind sets no cookies and builds no profiles. In the EU, storing an ID on the device can need consent even without cookies; storage-free mode stores nothing in the browser and avoids that question. Your obligations depend on your jurisdiction and the rest of your stack, so check with your counsel.

**Does cookieless analytics count returning visitors?**

In the default mode, yes, as long as the visitor keeps the site's storage. In storage-free mode, visitors are counted per day, so someone who returns on another day counts as new.

**Will it slow my site down?**

The script is about 9 KB, loads with defer, batches requests and uses idle time for non-urgent work.

**Can I import my old analytics?**

Yes: imports from Plausible, PostHog and Simple Analytics, and CSV export of your own data.

## Privacy

We count visits. We don't collect people.

- No cookies. By default visits are measured with a random ID in the site's own storage that visitors can clear; in storage-free mode nothing is stored on the device at all.
- No fingerprinting and no cross-site tracking, ever.
- The AI analyst never sees identities: emails and other traits you send with identify() stay out of its reach.
- IP addresses are never stored.
- Data hosted in the European Union (Hetzner, Germany).
- Opting out is one call, mm.optOut(), or clear your site data. Measurement stops and the stored IDs are deleted.

Stored: Page path (no query string); Referring domain and campaign; Country; Browser, OS and device type; Your events and their properties.
Never stored: IP addresses; Cookies or fingerprints; Form values, emails, phone numbers; Ad click IDs; Anything on screen.

## Pricing

Free while we're in beta. Join the waitlist for early access to everything on this page. When paid plans arrive, you'll pay for events, not seats, and beta users keep a generous free tier.

## FAQ

**What is MetricMind?**

MetricMind is privacy-first analytics for websites and mobile apps with a built-in AI analyst. One cookieless script and SDKs for iOS, Android, React Native and Flutter measure traffic, contacts, site search, Web Vitals, errors and crashes; the analyst answers questions in plain words with the query behind every number, briefs you every morning and investigates what changed. Data is hosted in the EU, and it's free during the beta.

**Will the analyst change anything on its own?**

No. When the next step is to measure or test something, it proposes a goal, a funnel, an alert, a timeline note or a feature flag (created switched off), and nothing is created until you click it.

**Can it be wrong?**

Yes, which is why every answer shows its queries and lists its assumptions, and every number in a finding is checked against the data before it's saved. When the data doesn't point to a cause, the analyst says so instead of guessing.

**What data does the AI see?**

Your question, your business context and the aggregated results of the queries it runs. It never receives identities: emails and traits sent with identify() stay out of its reach.

**How is MetricMind different from other analytics tools?**

Most analytics tools give you dashboards and leave the reading to you. MetricMind collects traffic, contacts, searches, Web Vitals, errors, crashes and uptime from one cookieless script and four app SDKs, and a built-in analyst explains what changed, checks the obvious causes first, and shows the query behind every number.

**Does it work for mobile apps?**

Yes, in beta. iOS, Android, React Native and Flutter SDKs measure screens, sessions, events, users, contacts, searches, errors and crashes on the same site as your website, with release health per version. Each app gets its own dashboard next to the website's, under one overview that compares them. The docs list exactly what to declare to Apple and Google.

**Do I still need a cookie banner?**

MetricMind sets no cookies and builds no profiles. In the EU, storing an ID on the device can need consent even without cookies, so for the strictest reading turn on storage-free mode (data-storage="none"), which stores nothing in the browser. Your obligations depend on your jurisdiction and the rest of your stack, so check with your counsel.

**Will it slow my site down?**

The script is about 9 KB, loads with defer, batches network calls and uses idle time for non-urgent work. Your visitors won't feel it.

**Does it work with single-page apps?**

Yes. Client-side navigation in Next.js, React Router, Vue and others is picked up automatically.

## Recent changes

### 2026-10-29: Storage-free mode, and the dashboard as a conversion view

- Smarter alerts: "tell me when contacts are unusually low" needs no number (each day is compared with the previous four weeks), and "visitors drop by more than 30% on the same day last week" catches slides. Before saving, any daily alert shows how many times it would have fired in the last 30 days.
- Full reports: the expand icon on any dashboard panel opens every row of it, with visits, bounce rate, visit duration and contact rate per source, channel, campaign or country, and time on page and scroll depth per page. Press / to search, sort by any column, download as CSV.
- More date ranges and comparisons: today, yesterday, month to date, last month, year to date and all time, compared with the previous period or the same dates last year, with the earlier period drawn on the chart. Every option has a one-key shortcut (D, E, W, T, M, P, Y, L, A, and X to compare).
- Channels: see how much traffic comes from organic search, paid search, organic and paid social, email, AI assistants, referrals and direct visits, worked out from each visit's source, UTM medium and ad click. Filter any report by channel, and see each channel's conversion rate in a goal's view.
- Click a goal on the website dashboard and the whole page becomes its conversion view: unique conversions, total conversions and conversion rate against the previous period, conversions per day, and every table (sources, campaigns, conversion pages, entry pages, countries, devices, browsers) with conversions and the conversion rate of each row. See at a glance which source or page brings people who actually contact you. The analyst can answer the same questions.
- Add data-storage="none" to the script tag and MetricMind stores nothing in your visitors' browsers. Visitors are counted on our server with a hash of a random salt that changes every day, your site, the IP address and the browser; the salt is deleted after two days, so nobody can be followed from one day to the next. Useful when you want to avoid any question about consent for device storage. Find the snippet under Settings → Tracking.

### 2026-10-28: Answers that don't get lost

- If your connection drops or you reload the page while the analyst is working, the answer is no longer lost: it keeps working, and the chat picks the answer back up as soon as you're connected again. Follow-up questions on Insights work the same way.
- The iOS and Android apps show their own logos on the overview and in the platform switcher.

### 2026-10-27: A first-week report

- New websites get a first-week report: after seven days the analyst writes up the trend, where visitors come from, what leads to contacts, what's abandoned and what to set up next, replacing its first look. Analyst notes on the overview now show their date.
- Funnel suggestions for your mobile apps, built from the screens they actually report: from the home screen to a listing to a contact, and through the selling flow, each previewed with this period's numbers.
- A person who signs in keeps the same made-up name on Sessions and Users, on every device.
- Clearer app releases (over-the-air updates read as "1.0.10 (14) · update 01a178f7"), contact channels named the same everywhere, a dated errors-per-day chart, and event charts that show quiet days as zero.
- Searches that found nothing, for websites too, with no code: when the results page says "0 résultats", "41 results" or "No results", the count is sent with the search. A weekly note lists the searches that found nothing, and every search term opens the visits that searched it.
- Real logos for browsers, operating systems, sources and AI assistants (Chrome, iOS, Instagram, ChatGPT…), and other referring sites show their own icon. We fetch those icons from the sites ourselves, so no third party learns where your traffic comes from.

### 2026-10-26: Live dashboards

- A new look: calmer neutrals in light and dark, one accent for what's selected, red only for what's wrong, and timeline notes as small markers instead of colored bands.
- A new sidebar: icons for every section, a setup checklist card at the top, and a button to fold it into a slim icon rail.
- Each website now has an overview across the website and its apps, and every platform has its own dashboard: the website, the iOS app and the Android app. A switcher in the sidebar keeps every report on the platform you picked.
- Every dashboard redrawn: cards topped with area charts, tables with share bars and source, browser and device badges, a dotted globe on the Map, tick gauges for Web Vitals, funnels as step columns, and 90 days of uptime bars.
- Dashboards update themselves every minute while they're open and the period includes today, so new visits, contacts and app sessions appear without reloading.
- Mobile apps lists the most viewed screens even when the app reports no load times (React Native), and counts someone who signs in partway through a session as one user.
- App vs website on the Mobile apps page: pages or screens per visit, time spent, and how often a visit ends in a search or a contact, with a plain verdict once both sides have enough visits. Plus what app users do and search for.
- The mobile SDKs' contact and search calls take properties, such as the category contacted or the number of search results, so you see which categories bring contacts and which searches find nothing.
- Sessions: follow any visit step by step, with the pages or screens, time spent and scroll on each, searches, clicks that did nothing, rage clicks, forms left unsent, contacts and errors, plus a one-line summary of the visit. Filter to frustrated, converted or bounced visits. It's rebuilt from the events you already send, so nothing on screen is recorded.
- Mobile app dashboards show stability day by day: the share of sessions without a crash, against a 99% target.
- Outbound links show the link's text and unique visitors next to clicks, and the Audience tables show pageviews next to visitors.
- Common journeys on the Sessions page: the paths visitors most often take through the site or app, and how many of those visits converted or got frustrated.
- The morning briefing now watches frustrated visits too: a jump in rage clicks, clicks that do nothing or forms left unsent is investigated like a traffic drop. The analyst can also query journeys and how visits went.
- Exclude your team's visits: add your office or home address (or a range) under Settings → Security, or click "Exclude my current address". Checked as events arrive and never stored.
- Event pages show which property values lead to a contact: for listing views, which categories, cities or price bands most often end in a call or WhatsApp message.
- Pin any panel of the website dashboard or Audience to your Board with the pin icon; it re-runs live for the range you pick.
- Alert rules can watch frustrated visits: get an email or Slack message when visits with rage clicks, dead clicks or abandoned forms pass a number per hour or day. Error groups link to the visits that hit them.
- Website searches can report how many results they found: add data-mm-results to your results page and site search shows which searches find nothing, as it already does for apps.

### 2026-10-24: Filters, funnels and charts

- Filters with is, is not and contains, from an Add filter button that suggests your site's real values.
- Funnels show the typical time between steps and how each traffic source completes them; goals and funnels can now be edited, and the analyst can propose fixes to them.
- The analyst draws charts in its answers when a trend is clearer as a picture, and suggests follow-up questions you can ask in one click.
- Insights is now an inbox: what needs you, what's being watched, what's done.
- Forms, recorded automatically: who started, who sent, and the field people were on when they gave up (field names only, never what they typed). Dead clicks show buttons and links that do nothing when clicked.
- Page templates: count /listing/* as one page instead of thousands of listing URLs.
- Share a finding with a read-only link, for a boss or client without an account. Stop sharing at any time and the link stops working.
- Two-factor sign-in with an authenticator app, with recovery codes. Once on, it's asked for every way in: password, email link, Google or GitHub. You can also change your password and sign out other devices.
- Server-side events: send signups, payments and background jobs from your backend with a key, joined to the visitor's visit when you pass the browser's IDs. A small helper for Node, Bun, Deno and edge functions batches and retries.
- More accurate page views: pages the browser prerenders count only when opened, and going back to a cached page counts as a view. Optionally honor Global Privacy Control with data-respect-gpc.
- Ask follow-up questions right under a finding (is it fixed? which browsers?). The analyst continues from its investigation instead of starting over, and the thread stays with the finding.
- Scheduled questions: ask the analyst something once, then have it answered every day or every Monday, with a quiet mode that only tells you when a threshold is crossed or something needs action. Run it now any time.
- A page for every custom event: how often and by how many visitors, where it happens, the latest ones, and each property shown the way its values suggest (a split for yes/no and a few categories, top values with 'other' for the rest).
- Board: pin any query behind an analyst answer or a finding, and it stays on your site's board, re-run live for whatever range you pick. Charts for trends, tables for rankings, tiles for headline numbers.
- Choose how hard the analyst works: Quick, Balanced or Thorough. Each answer shows the tokens it used, and your conversation history can be renamed or cleaned up.
- Uptime monitors now watch your TLS certificate: days left on Pulse, and an alert 14, 7, 3 and 1 days before it expires or as soon as it stops being trusted. Check interval and timeout can be changed any time, and a monitor that hasn't been checked recently shows as unknown instead of up.
- Incidents on your status page: post what's happening, follow up with updates until it's resolved, and visitors see the timeline. An open incident replaces "All systems operational".
- Analyst rules: tell the analyst how your team reads its data (what counts as a lead, which traffic to ignore) and every answer follows them. Analyst review shows what people asked, which answers hit a wall, and what would help.
- Pick the date range your site pages open with (Settings → Account), for example the last 7 days.
- The analyst's working steps show while it works, then fold into one line you can open to see every query. Typing a question you asked before offers the earlier answer first.
- Connect Claude, Cursor and other MCP clients with your MetricMind account instead of an API key: add the URL, sign in, choose which websites the app may read and whether it may ask the analyst. Disconnect any time.
- Every answer and finding now lists the assumptions behind it: which days "this week" covers, a partial day left out, what was counted as a lead. Agree with one? Make it a rule in one click, and the analyst follows it from then on.
- Long answers can notify you when they're ready, so you can switch tabs while the analyst works (opt in under the chat box). Scheduled questions now stop after 90 days unless you extend them.
- Light and dark mode: pick Light, Dark or System in the sidebar or under Settings → Account → Appearance. The overview's number cards show their own trend and switch the chart with a click. Web Vitals has gauges and a single real experience score next to Google's pass/fail test.
- A Mobile apps page for each site: app users, crash-free users, ANR rate against Google Play's bar, releases, start time, slow screens and crashes in one place, with install snippets until your app sends data.
- Webhooks: findings, alerts, uptime and the weekly digest as signed JSON to your own endpoint (Discord or Teams relays, PagerDuty, Zapier, your backend), with a filter per event type.

### 2026-10-24: An analyst you can check

- Every number in an investigation is checked against the data before it's saved; invented numbers go back to the analyst to fix.
- Investigations start from facts computed by code: which source, page, country or device carries a change, the hour it started, and which error is behind a rise.
- When the analyst recommends a fix, it adds a measurable check (for example errors on a page below 1 a day). MetricMind re-measures it a few days later and tells you whether the fix worked.

Full changelog: https://metricmind.si/changelog