RUM metrics and dimensions in CoreDash
What CoreDash measures from real visitors and every way you can split it, with the codes the API filters on.
What can CoreDash split Core Web Vitals by?
In short
- 43 dimensions for LCP, INP, CLS, FCP and TTFB: the page, the visitor, how they arrived, the element or script behind a number, and labels your own site sends. LCP, INP, TTFB and FCP also break down into 12 sub-phases.
- Template: Folder, or a Page label your site sends. Device: Device, mobile or desktop. Geography: Country. Release: Release and A/B test.
- Each value has a code, such as
d=mobile, and the API and MCP filter on that code. Every page view is kept 56 days, and daily totals 13 months for the rows that say so.
Jump to a group: which page, who the visitor was, how they arrived, what was slow, your own labels or the metrics and their lines.
"Add filter" with "All dimensions" open: every dimension in its group, with a search field on top. Device has its own switch in the top bar, so the list leaves it out.
What do CrUX and PageSpeed Insights split by?
CrUX is the Chrome data behind PageSpeed Insights (PSI) and Search Console. It gives one number for your origin, or for a single URL with enough Chrome traffic, and splits it by device and little else. CoreDash keeps each page view with its dimensions, so every key in the CoreDash column is a row further down this page. Why the numbers differ is on CrUX and Search Console next to CoreDash.
| Split | CoreDash | CrUX and PageSpeed Insights |
|---|---|---|
| Section or template | Folder ff, Page label lb, Path path, Hostname host | No. The whole origin, or one URL at a time. |
| URL | Every URL with visits, URL u, plus Query parameter qsf | Only public URLs with enough traffic, with the query string dropped. PSI falls back to the origin. |
| Device | Device d, mobile or desktop, plus Device memory and Operating system | Phone, desktop and tablet. PSI shows a mobile and a desktop tab. |
| Country | Country cc | Not in the CrUX API or PSI. CrUX on BigQuery has it per origin, per month. |
| Browser | Browser browser and Browser version bv, Safari and Firefox included | Chrome users who opted in, on desktop and Android. No Safari, Firefox, Edge or Chrome on iOS. |
| Connection | Connection type et, Network speed dl, Client capability score ccs, Chromium only | No split. The connection type was removed as a dimension in February 2025; round-trip time is now a number of its own. |
| Navigation type and back/forward cache | Navigation type nt and Not restored reason (bfcache) nrr | The share of each navigation type in the CrUX API. The vitals can't be split by it, and PSI doesn't show it. |
| LCP element | LCP element lcpel, its type, the image URL and its priority | The share of image and text LCPs and four image sub-parts at p75 in the CrUX API. No element. PSI names the element only in its Lighthouse lab test. |
| INP script | INP script lurl, plus the element, the input type and the load state | No. |
| CLS element | CLS element clsel and five more CLS dimensions | No. PSI names shifting elements only in its Lighthouse lab test. |
| Release or A/B test | Release rel and A/B test ab, sent by your site | No. |
| Your own labels | Page label lb, plus Logged in, Cart status and Indexable | No. |
| History | Every page view 56 days with all dimensions; daily totals 13 months for the five metrics and the rows that say so | One 28-day window in the CrUX API and PSI. Up to 40 weekly windows in the CrUX History API. Monthly origin data back to 2017 on BigQuery. |
Can I see one section of my site on its own?
Yes, with Folder: the first folder of the path, so /blog/my-first-post counts under /blog/. On most sites that folder is a section or a template, which makes it the quickest way to compare product pages with the blog. Folder and Hostname go back 13 months as daily totals; Path and URL go back 56 days. Product pages straight under the root each count as a folder of their own, so a template like that needs a Page label from your site.
| Dimension | What it answers | Values and filter codes | Metrics and history |
|---|---|---|---|
Folderff | The first folder of the page path, which groups the pages of one section. | The first folder of the path, without slashes, such as blog, productsNot set: /, filter [empty] | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Hostnamehost | The domain or subdomain of the page. | Text as recorded, such as www.example.com, shop.example.com | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Pathpath | The path of the page address, without the domain and the query string. | Text as recorded, such as /, /blog/my-first-post, /products/shoes | All 5 metrics Kept 56 days. |
URLu | The full address of the page, without the query string. | Text as recorded, such as https://www.example.com/, https://www.example.com/blog/my-first-post | All 5 metrics Kept 56 days. |
Mobile or desktop, and which country?
Device is mobile or desktop by the width of the browser window, and in the app it is the switch in the top bar. Country is the geography split, one two-letter code per country, with no region or city. "Connection type", "Network speed", "Device memory" and "Client capability score" read browser APIs that only Chromium has, so Safari and Firefox visits have no value there, or Unknown for the score. The tracker fills in three rows by itself: "Logged in" from WordPress, common login cookies or a user id in the Google Tag Manager data layer; "Cart status" from the cart of Shopify, WooCommerce, Magento or BigCommerce, or an add-to-cart in that data layer; Indexable from the robots meta tag and the canonical. A page variable on your site overrides each of them, and the value Admin only comes from that variable.
| Dimension | What it answers | Values and filter codes | Metrics and history |
|---|---|---|---|
Browser languageul | The language the visitor's browser is set to. | Text as recorded, such as en-US, nl, de-DE | All 5 metrics Kept 56 days. |
Cart statuscart | Whether the visitor had items in the shopping cart, and whether they were just added. | Empty cart (0)Just added ( 1)Has items ( 2) | All 5 metrics Kept 56 days. |
Client capability scoreccs | How fast the visitor's connection answers, by the browser's round-trip time in 50 ms steps, from Excellent to Constrained. Chromium browsers only. | Unknown (0)Excellent ( 1)Good ( 2)Moderate ( 3)Limited ( 4)Constrained ( 5) | All 5 metrics Kept 56 days. |
Countrycc | The country the visitor was in. | A two-letter country code, such as NL, US, DE | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Indexableidx | Whether search engines may index the page, going by its robots meta tag and canonical. | Not indexable (0)Indexable ( 1) | All 5 metrics Kept 56 days. |
Logged inli | Whether the visitor was logged in, logged out or an admin. | Logged out (0)Logged in ( 1)Admin ( 2) | All 5 metrics Kept 56 days. |
New or repeat visitorfv | Whether the page view was the visitor's first ever on your site, in this browser. | Repeat visitor (0)New visitor ( 1)Not measured, off in settings ( 2)Not set: Repeat visitor, filter [empty] | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Browserbrowser | The browser the visitor used, without its version. | Text as recorded, such as Chrome, Safari, Firefox | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Browser versionbv | The major version of the visitor's browser. | Text as recorded, such as 142, 26 | All 5 metrics Kept 56 days. |
Connection typeet | The connection speed class the browser estimates for the visitor, from 4G down to Slow 2G. Chromium browsers only. | 4G (4g)3G ( 3g)2G ( 2g)Slow 2G ( slow-2g) | All 5 metrics Kept 56 days. |
Deviced | Mobile when the browser window is under 724 px wide, desktop otherwise, so a wide tablet counts as desktop. | Mobile (mobile)Desktop ( desktop) | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Device memorym | How much memory the visitor's device has, in GB, as the browser reports it. Chromium browsers only. More: Optimize the Core Web Vitals for low end devices | A number in GB, such as 8, 4, 2 | All 5 metrics Kept 56 days. |
Network speeddl | The download speed the browser estimates for the visitor's connection, in Mbps. Chromium browsers only. | A number in Mbps, such as 10, 5, 1 | All 5 metrics Kept 56 days. |
Operating systemos | The operating system of the visitor's device. | Text as recorded, such as Windows, Android, iOS | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Did they reload, come back or click an ad?
A page that comes from the back/forward cache (bfcache) or was prerendered is on screen almost at once, so "Navigation type" keeps those page views apart from full page loads. When a back or forward visit missed the bfcache, "Not restored reason (bfcache)" gives the first reason the browser named. Only Chromium reports it. Campaign is worked out from the navigation type, where the visitor came from and the query parameters, so it only has a value for a full page load from outside your site. "Query string" keeps the parameter names and shows every value as *, so ?q=shoes and ?q=boots count as one. Redirects and "Same-site redirects" count the redirects the browser reports, and the browser only reports them when every hop stays on the page's own host. A chain through another host, such as an ad or tracking link or example.com to www.example.com, counts as no redirect. Its time is part of Waiting + Redirect, the first sub-phase of TTFB.
| Dimension | What it answers | Values and filter codes | Metrics and history |
|---|---|---|---|
Campaigncamp | Which ad click or campaign link brought the visitor to the page from outside your site. More: Strip Tracking Parameters with Cloudflare Workers and Transform Rules | Google Ads click (gads)Microsoft Ads click ( msads)Facebook or Instagram link ( meta)Other ad click ( ads)Other campaign ( utm)Other query parameter ( other)No query parameter ( none)Not set: Not a full page load from outside, filter [empty] | All 5 metrics Kept 56 days. |
Query parameterqsf | The names of the query parameters in the page address, such as utm_source or gclid. | Text as recorded, such as utm_source, gclid, page | All 5 metrics Kept 56 days. |
Query stringqs | The query string of the page address, with each value shown as *. | Text as recorded, such as utm_source=*&utm_medium=*, page=* | All 5 metrics Kept 56 days. |
Navigation originno | Whether the visitor came from another page of your site or from outside. | Within the site (1)From outside ( 2)Unknown ( 0) | All 5 metrics Kept 56 days. |
Navigation typent | How the page view started; prerenders, reloaded discarded tabs and route changes (when switched on) come from Chromium only. | Page load (navigate)SPA route change ( softnav)Reload ( reload)Back/forward ( back-forward)Back/forward from bfcache ( back-forward-cache)Prerendered ( prerender)Restored ( restore)Not set: Left before load, filter [empty] | All 5 metrics Every page view kept 56 days, daily totals 13 months. |
Not restored reason (bfcache)nrr | The first reason the browser gives for not restoring the page from the back/forward cache. Chromium browsers only. | Fetch cancelled on unload (fetch)Lock held on unload ( lock)Hidden by the browser, often a cross-origin iframe ( masked)Navigation error ( navigation-failure)HTML parsing not finished ( parser-aborted)Open WebSocket ( websocket)Unload event handler ( unload-listener)Cache-Control: no-store ( response-cache-control-no-store)Another page holds a window.opener reference ( related-active-contents)Another page holds a window.opener reference ( non-trivial-browsing-context-group)Network request open on unload ( outstanding-network-request)Blocked by a browser feature ( appfeature)Page still loading on unload ( navigating)BroadcastChannel message while cached ( broadcastchannel-message)HTTP error status, such as 404 ( response-status-not-ok)And other reasons, as the browser names them. | All 5 metrics Kept 56 days. |
Redirectsredir | The exact number of redirects within your own site that came before the page loaded. More: TTFB Waiting Duration: Find and Remove the Redirects | No redirect (0)1 redirect ( 1)2 redirects or more | LCP, FCP, TTFB Kept 56 days. |
Same-site redirectsrdg | How many redirects within your own site came before the page loaded: none, one, or two or more. | No redirect (0)1 redirect ( 1)2 redirects or more ( 2) | LCP, FCP, TTFB Kept 56 days. |
| Folder | Page views | p75 LCP | Distribution | Over 2.50 s |
|---|---|---|---|---|
| /products/ | 18,400 | 3.06 s | 62% good | 6,992 |
| /category/ | 9,800 | 4.21 s | 52% good | 4,704 |
| / | 12,600 | 2.18 s | 81% good | 2,394 |
| /blog/ | 7,200 | 1.92 s | 86% good | 1,008 |
| /checkout/ | 2,100 | 2.41 s | 77% good | 483 |
LCP split by Folder on the LCP page: one row per section of a shop, with its page views, its p75 and how those page views spread over Good, "Needs improvement" and Poor. The rows are sorted by the page views over the Good line.
Which element or script is behind a bad score?
Each page view adds one value per row, from its LCP element, its slowest interaction or its largest layout shift. For LCP, CoreDash records the element as a CSS selector, its type, the image URL and how the image was asked for: preloaded, high fetchpriority, lazy, or none of these. For INP it records the element the visitor used, pointer or keyboard, how far the page had loaded, and the script that ran longest. That script comes from the Long Animation Frames API (LoAF), so only Chromium browsers report it. CLS has six dimensions, from the element that moved to how far the visitor had scrolled, all Chromium only because no other browser reports layout shifts. On the INP and CLS pages in the app the tabs drop the metric from the name ("Input type", "Load state", "Scroll depth"), and "CLS interaction element" is called "Tapped element".
| Dimension | What it answers | Values and filter codes | Metrics and history |
|---|---|---|---|
LCP elementlcpel | Which image or block of text was the LCP element of the page. More: Fix & Identify Largest Contentful Paint (LCP) Issues | The element's CSS selector, such as main>div.hero>img, h1.title, #banner>picture>img | LCP Kept 56 days. |
LCP element typelcpet | What kind of element the LCP was: text, an image, a background image or a video. | Text (text)Image ( image)Background image ( background-image)Video ( video) | LCP Kept 56 days. |
LCP image prioritylcpprio | How the browser was told to load the LCP image: preloaded, high fetchpriority, lazy or none of these. More: Optimize the LCP Resource Load Delay | Not an image (0)Preloaded ( 1)High fetchpriority ( 2)Not preloaded ( 3)Lazy loaded ( 4) | LCP Kept 56 days. |
LCP image URLlcpurl | The address of the LCP image, when the LCP element is an image. More: Optimize the Largest Contentful Paint (LCP) Image | The full URL of the image; tables in the app show its file name, such as https://www.example.com/images/hero.webp, https://cdn.example.net/banner.jpg | LCP Kept 56 days. |
INP elementinpel | The element the visitor clicked, tapped or typed in for the slowest interaction. More: Find and Fix Interaction to Next Paint (INP) Issues | The element's CSS selector, such as button.add-to-cart, #menu-toggle, input#search | INP Kept 56 days. |
INP input typeinpit | Whether the slowest interaction was a click or tap, or a key press. | Pointer (pointer)Keyboard ( keyboard) | INP Kept 56 days. |
INP load stateinpls | How far the page had loaded when the slowest interaction happened. | Loading (loading)DOM interactive ( dom-interactive)DOM content loaded ( dom-content-loaded)Complete ( complete) | INP Kept 56 days. |
INP scriptlurl | The script that ran longest during the slowest interaction. Chromium browsers only. More: Long Animation Frames (LoAF) API: Debug INP in Production | Text as recorded, such as /js/app.js, cdn.example.net/widget.js | INP Kept 56 days. |
CLS elementclsel | The element that moved most during the largest burst of layout shifts. Chromium browsers only. More: How to Find and Fix CLS Issues (Cumulative Layout Shift) | The element's CSS selector, such as main>div.hero>img, #cookie-banner, header>nav | CLS Kept 56 days. |
CLS interaction elementclsiel | The element the visitor last clicked, tapped or typed in before the largest layout shift. Chromium browsers only. | The element's CSS selector, such as button.load-more, #faq>details>summary | CLS Kept 56 days. |
CLS last interactionclsint | What the visitor did last before the largest layout shift, such as a tap or a scroll. Chromium browsers only. | No interaction (none)After tap/click ( pointer)After keypress ( keyboard)After scroll ( scroll) | CLS Kept 56 days. |
CLS load stateclsls | How far the page had loaded when the largest layout shift happened. Chromium browsers only. | Loading (loading)DOM interactive ( dom-interactive)DOM content loaded ( dom-content-loaded)Complete ( complete) | CLS Kept 56 days. |
CLS scroll depthclssd | How far down the page the visitor had scrolled at the largest layout shift, in steps of 5%. Chromium browsers only. | 0 to 100%, in steps of 5 | CLS Kept 56 days. |
When it shiftedclsm | When the largest layout shift happened: during load, after load, after a click or tap, or after a scroll. Chromium browsers only. | During load, before any input (load)After load, without input ( after)After a click, tap or keypress ( input)After a scroll ( scroll)Not set: No shift or not recorded, filter [empty] | CLS Kept 56 days. |
Can I compare releases and A/B tests?
Yes, once your site sends them. "Page label" comes from the page variable __CWVL, "A/B test" from __CWVAB, and Release from __CWVREL or the release API. Advanced setup options shows the page variables and CI/CD the release API. With Release set, the page views of the old and the new version sit side by side after a deploy. All three keep every page view for 56 days, with no daily totals after that.
| Dimension | What it answers | Values and filter codes | Metrics and history |
|---|---|---|---|
Page labellb | The page type your site gives the page, such as product or category. Your site sends this value: advanced setup options. | Text as recorded, such as product, category, checkout | All 5 metrics Kept 56 days. |
A/B testab | Which version of an A/B test the visitor saw, as your site names it. Your site sends this value: advanced setup options. | Text as recorded, such as control, variant-b | All 5 metrics Kept 56 days. |
Releaserel | The version of your site that served the page view. Your site sends this value: CI/CD. | Text as recorded, such as v2.4.1, 2026-11-20 | All 5 metrics Kept 56 days. |
Which metrics, and where are the lines?
LCP, INP and CLS use Google's lines. FCP and TTFB are diagnostic metrics, with the lines web.dev gives for them. Good includes its line, so an LCP of exactly 2.50 s is Good. "Needs improvement" runs from there up to the Poor line. The sub-phase lines are CoreDash's own: the good lines of the three INP sub-phases add up to Google's 200 ms, their poor lines to 500 ms. The p75s themselves don't add up, because each sub-phase is its own p75. "Load Delay" and "Load Time" only exist when the LCP element is an image or a background image; a text LCP goes from the first byte straight to "Render Delay". "First byte to FCP" is FCP minus TTFB on full page loads, and its lines are FCP's minus TTFB's good line. Sub-phases are kept 56 days; the five metrics also as daily totals for 13 months.
| Metric | What it measures | Lines | History |
|---|---|---|---|
Largest Contentful PaintLCP | How long until the largest image or block of text on screen is painted. More: Largest Contentful Paint (LCP): What It Is, How to Measure & Optimize | Good: Up to 2.50 s Needs improvement: 2.50 s to 4.00 s Poor: Over 4.00 s | Every page view kept 56 days, daily totals 13 months. |
Interaction to Next PaintINP | How long the slowest click, tap or key press on the page takes until the next frame is painted. More: Interaction to Next Paint (INP): What It Is and How to Improve It | Good: Up to 200 ms Needs improvement: 200 ms to 500 ms Poor: Over 500 ms | Every page view kept 56 days, daily totals 13 months. |
Cumulative Layout ShiftCLS | How much visible content moves unexpectedly on the page, in its largest burst of layout shifts. More: What is Cumulative Layout Shift (CLS) and how to fix it | Good: Up to 0.10 Needs improvement: 0.10 to 0.25 Poor: Over 0.25 | Every page view kept 56 days, daily totals 13 months. |
First Contentful PaintFCP | How long until the browser paints the first text or image of the page. More: First Contentful Paint (FCP): What It Is and How to Improve It | Good: Up to 1.80 s Needs improvement: 1.80 s to 3.00 s Poor: Over 3.00 s | Every page view kept 56 days, daily totals 13 months. |
Time to First ByteTTFB | From the start of the navigation to the first byte of the HTML. More: Time to First Byte (TTFB): What It Is and How to Reduce It | Good: Up to 0.80 s Needs improvement: 0.80 s to 1.80 s Poor: Over 1.80 s | Every page view kept 56 days, daily totals 13 months. |
| LCP sub-phase | What it measures | Lines | History |
|---|---|---|---|
Time to First ByteTTFB | The same metric as Time to First Byte above. | Good: Up to 0.80 s Needs improvement: 0.80 s to 1.80 s Poor: Over 1.80 s | Every page view kept 56 days, daily totals 13 months. |
Load DelayLOADDELAY | From the first byte until the browser starts loading the LCP image. Only for an LCP element of type Image or Background image. More: Optimize the LCP Resource Load Delay | Good: Up to 400 ms Needs improvement: 400 ms to 800 ms Poor: Over 800 ms | Kept 56 days. |
Load TimeLOADTIME | How long the LCP image takes to download. Only for an LCP element of type Image or Background image. More: Optimize the LCP Resource Load Duration | Good: Up to 400 ms Needs improvement: 400 ms to 800 ms Poor: Over 800 ms | Kept 56 days. |
Render DelayRENDERDELAY | From the end of the download, or the first byte for text, until the element is painted. More: Optimize the LCP Element Render Delay | Good: Up to 400 ms Needs improvement: 400 ms to 800 ms Poor: Over 800 ms | Kept 56 days. |
| INP sub-phase | What it measures | Lines | History |
|---|---|---|---|
Input DelayINPUTDELAY | From the input until its event handlers start. More: INP Input Delay: Causes, Diagnosis, and Fixes | Good: Up to 40 ms Needs improvement: 40 ms to 100 ms Poor: Over 100 ms | Kept 56 days. |
Processing TimePROCESSING | How long the event handlers of the interaction run. More: INP Processing Time: Speed Up Slow Event Handlers | Good: Up to 80 ms Needs improvement: 80 ms to 200 ms Poor: Over 200 ms | Kept 56 days. |
Presentation DelayPRESENTATION | From the end of the event handlers until the next frame is painted. More: INP Presentation Delay: DOM Size, Layout and Rendering | Good: Up to 80 ms Needs improvement: 80 ms to 200 ms Poor: Over 200 ms | Kept 56 days. |
| TTFB sub-phase | What it measures | Lines | History |
|---|---|---|---|
Waiting + RedirectWAITING | Before the request leaves: redirects, including ad and tracking links, and the browser's own start-up. More: TTFB Waiting Duration: Find and Remove the Redirects | Good: Up to 50 ms Needs improvement: 50 ms to 100 ms Poor: Over 100 ms | Kept 56 days. |
Worker & CacheCACHE | Service worker start and the HTTP cache lookup. More: TTFB Cache Duration: Speed Up the Service Worker | Good: Up to 50 ms Needs improvement: 50 ms to 100 ms Poor: Over 100 ms | Kept 56 days. |
DNSDNS | Looking up the IP address of your domain. More: TTFB DNS Duration: Fix a Slow DNS Lookup | Good: Up to 50 ms Needs improvement: 50 ms to 100 ms Poor: Over 100 ms | Kept 56 days. |
ConnectionCONNECTION | TCP and TLS handshakes with your server or CDN. More: Connection Duration in TTFB: Speed Up TCP and TLS | Good: Up to 150 ms Needs improvement: 150 ms to 250 ms Poor: Over 250 ms | Kept 56 days. |
RequestREQUEST | Your server or CDN building the HTML, plus the network time until the first byte. More: TTFB Request Duration: Reduce Server Processing Time | Good: Up to 500 ms Needs improvement: 500 ms to 1,200 ms Poor: Over 1,200 ms | Kept 56 days. |
| FCP sub-phase | What it measures | Lines | History |
|---|---|---|---|
Time to First ByteTTFB | The same metric as Time to First Byte above. | Good: Up to 0.80 s Needs improvement: 0.80 s to 1.80 s Poor: Over 1.80 s | Every page view kept 56 days, daily totals 13 months. |
| First byte to FCP | From the first byte to the first paint, on full page loads: render-blocking CSS, JavaScript and fonts. Shown on the FCP card; the API and MCP have no key for it. More: Render-blocking requests: fixes with or without a developer | Good: Up to 1.00 s Needs improvement: 1.00 s to 2.20 s Poor: Over 2.20 s | Kept 56 days. |
The API and MCP use the same keys
The CoreDash API and the MCP server filter on the keys and codes in these rows, such as d=mobile or ff=[empty] for an empty value.
The GET /api/v1/dimensions endpoint and the MCP tool list_dimensions both return the list of dimensions. Not measuring yet? Install CoreDash first.