CoreDash vs Vercel Speed Insights

If your site runs on Vercel, real-user monitoring is one import away. Add the Speed Insights component, deploy, and field data shows up next to your deployments and logs. Since August 2026 there's a free tier on every Vercel plan.

I built CoreDash, so keep that in mind. Where Vercel does something better I'll say so.

What Vercel Speed Insights does well

Setup. On Next.js, Nuxt, SvelteKit, Remix or Astro you install one package (or paste one script tag on a static site) and deploy. The script and the endpoint that receives the data run as routes on your own domain, so there's no third-party domain to add to your CSP and no new connection. The browser reuses the connection it already has to your site.

vercel.com, October 2026.

It understands your framework. Speed Insights picks up dynamic routes by itself, so a thousand blog posts become one row for /blog/[slug] instead of a thousand rows with a few visits each. Preview and production are measured separately and every data point has the deployment ID, so traffic on a preview URL never ends up in your production numbers.

Privacy is done well. No cookies, nothing tied to an IP address or a visitor, and Vercel says sessions can't be reconstructed from the data. The client package is open source (MIT), and with sampleRate and beforeSend you can control volume or drop events before they leave the browser.

Vercel has also done work for coding agents. The vercel metrics command (June 2026) queries Speed Insights from the terminal: p75 for LCP, INP, CLS, FCP and TTFB, grouped by route, device, country or project, and --all compares every project in a team. Vercel markets it for agents and it works fine for that. On the paid tier Drains stream every raw data point to your own warehouse or to Datadog. And viewer seats on Pro are free, so a client or manager can look at the numbers without costing you a seat.

Free field data for every Vercel team

Until August 2026 the free version only covered one project on the Hobby plan. Now every Vercel team gets field data for free, and that matters more than anything else on this page.

Lighthouse loads your page once on a simulated phone and never clicks anything, so there's no INP at all. CrUX only counts opted-in Chrome users and stays empty until a page has enough of them (see not enough data in Search Console). Speed Insights measures your actual visitors. If you run it and it answers your questions, great, nothing here is a reason to rip it out.

CoreDash and Vercel Speed Insights side by side

Pricing and features checked September 2026. Prices in USD per month.

Why I think CoreDash is the better choice long term

It works on any host

Speed Insights is part of the Vercel platform. Collection runs through routes Vercel adds to each deployment, Vercel advises against putting a reverse proxy in front of it, and the old intake API you could call from anywhere is deprecated. Fine as long as every site you care about is on Vercel. But once the marketing site moves to WordPress, or the shop runs on Shopify, that site has no RUM. And if you leave Vercel completely you lose the tool and its history.

The history is short anyway. Plus keeps 30 days on Pro. Swapped your image CDN in March and want to know in September if it helped? Then you need Drains, your own storage and your own queries. CoreDash keeps 1 year on Standard, 3 years on Studio and 8 years on Enterprise, on any host. The script tag works the same on Next.js, WordPress or a PHP app from 2012.

A CSS selector isn't enough

Speed Insights tells you which element was slow, and according to the docs that's where the attribution stops: a CSS selector like html>body img.header. The docs don't mention LCP phases, INP phases or Long Animation Frames.

So you know the hero image is slow, but not why, and the fix depends on the why. A long load delay means the browser found the image late and you need a preload or fetchpriority. A long load time means the file is too big or the CDN is slow. A long render delay means something is blocking the render.

CoreDash splits LCP into TTFB, load delay, load time and render delay, and records whether the LCP image was preloaded, had fetchpriority=high or was lazy loaded (LCP priority). For INP you get input delay, processing and presentation, plus the input type, the load state and, through Long Animation Frames, the URL of the script that did the work. On a Next.js site I usually check load state first, because a slow tap on mobile while the page is still hydrating is a different bug than a slow tap on a page that's fully loaded.

Speed Insights also puts the Real Experience Score up front, a 0 to 100 number weighted like Lighthouse 10. Google doesn't rank on a composite score though. It looks at p75 LCP, INP and CLS from CrUX, pass or fail per metric. CoreDash uses Google's web-vitals library with the same thresholds and shows CrUX next to your own data, so you can see when a fix starts showing up in Google's numbers.

Soft navigations

Speed Insights only measures hard navigations. On a Next.js app that means the first page view of a session, and every client-side route change after that isn't measured. On an app where people click around a lot, that can be most of your page views. CoreDash measures soft navigations too, so those route changes are in your data.

Field data for your AI agent

The vercel metrics CLI gives an agent numbers: p75 per route, device or country. Vercel's MCP server has no Speed Insights tool, only a generic one that runs the CLI. There's no diagnosis built in, so the agent has to work out the why by itself. A lot of my own performance work starts with an agent reading field data these days, and a p75 alone doesn't get it far. It needs the phases and the script URLs.

The CoreDash MCP server is included in every plan, Starter too, and it's the same endpoint as the API. Three read-only tools return metrics, time series with a trend summary, and histograms, filtered by any of the attribution dimensions above. Claude connects to it as a custom connector with OAuth, and Cursor, Windsurf, VS Code and Gemini CLI work as well.

CWV Superpowers runs on top of that. It's free and MIT licensed. It pulls your field data, finds the phase that takes the biggest share of the time, then records a Chrome trace with mobile emulation, Fast 3G and a 4x CPU slowdown to find what causes it. You get a diff with file and line, or an HTML report for a developer. No Lighthouse score involved. Fixing INP with an AI agent shows a real run.

The Speed Insights docs don't describe any alerts either, so somebody has to remember to look. CoreDash AI Auto Alerts check hundreds of segments every day and only flag a regression that hits at least 2% of your traffic.

What it costs

Speed Insights Plus needs the Pro plan first. On top of that it's $10 per project per month plus $0.65 per 10,000 events, with no events included. Each measured page load sends 3 to 6 events, so 1M events covers roughly 170K to 330K page views and costs $75 a month for one site, plus Pro. CoreDash Standard covers 500K pageviews on up to 2 sites for $39 a month, or $31 billed annually.

Vercel docs, Speed Insights limits and pricing, October 2026.

For agencies it's the project fee that adds up. Ten small client sites cost $100 a month in project fees before a single event is counted, and each one has to be a Vercel project. The CoreDash agency Studio plan covers 10 client sites with 1M pooled pageviews and unlimited users for $89 a month ($71 billed annually), on any host. With the Agency API a master key creates client projects over REST, and the MCP tools read any client project by ID. CoreDash isn't white-label: you get the data and keep your own reporting.

When to pick Vercel Speed Insights instead

• Every site you care about is on Vercel and you want zero setup. It's one component and a deploy.
• You want deploy tracking without any extra work. CoreDash can tag releases from CI, but you need to add that step to your pipeline yourself. Speed Insights does it out of the box and keeps preview and production apart.
• Your site is small. The free tier covers roughly 1,700 to 3,300 page views a month, and that's real field data for $0.
• Your data already goes to a warehouse or Datadog and you build your own views there. Drains fit that.
• It's already on your site and answers your questions. Then just keep it.

Run both for two weeks

You can run CoreDash next to Speed Insights on the same Vercel site. The CoreDash script is under 2 KB and loads async, and the trial is 14 days without a credit card. See which one tells you what to fix first. The Next.js measurement guide covers the setup. If you want to see the dashboards first, open the CoreDash demo, and check pricing for your traffic. Other tools are covered in all comparisons.

Sources: Speed Insights overview, limits and pricing, free tier announcement, metrics, using Speed Insights, privacy policy, vercel metrics CLI, Vercel MCP tools, Speed Insights drains.