CoreDash vs Datadog RUM

What each tool is for

Datadog is an observability platform: infrastructure, logs, APM, security, synthetics and Real User Monitoring. The RUM part is there so an engineer can go from a slow or broken session in the browser to the backend trace behind it. CoreDash only measures Core Web Vitals for your real visitors and helps you figure out why pages are slow.

I built CoreDash, so keep that in mind. I'll try to be fair to Datadog though. It's a good product, and a lot of what it does CoreDash doesn't do at all.

What Datadog RUM does well

The best thing about Datadog RUM is what's behind it. A visitor gets a 4 second LCP, you open the session and follow it into the APM trace, the logs and the profile of the request that built the HTML. If the slow part of your LCP is the server, I don't know a faster way to show it. A browser-only tool tells you TTFB was 1.8 seconds, Datadog can show you the query that took 1.5 of those.

datadoghq.com, October 2026.

The Core Web Vitals data is good too. The SDK reports the element behind LCP, INP and CLS, splits LCP into TTFB, load delay, load time and render delay, and INP into input delay, processing and presentation delay. Since SDK v6 it collects Long Animation Frames instead of long tasks, with script attribution mapped back to your unminified source. With the optional browser profiling on top (it ties a slow interaction to a specific JavaScript function) that's a pretty complete INP setup.

The RUM pricing is clever as well. With what they call RUM without Limits you ingest every session for the standard performance metrics at $0.15 per 1,000 sessions (annual), and only pay the Investigate price for sessions you keep in full, like the ones with a poor LCP or an error. So the old choice between sampling and a huge bill is gone. Session Replay, error tracking and mobile RUM for iOS, Android, React Native and Flutter are all in the same place. Datadog also has an official MCP server with RUM tools, so an AI agent can search RUM events and aggregate percentiles.

Already running Datadog RUM?

If you already run Datadog RUM and you're happy with the Core Web Vitals data in it, fine. Lots of the sites I look at have nothing but Search Console: 28 days of Chrome users, grouped into URL groups, telling you something is slow but not what. Datadog measures all your visitors, Safari and the iPhone included, and tells you which element was the LCP. That's a big step up. I wouldn't rip out a setup that works because of a comparison page, mine included.

CoreDash vs Datadog RUM at a glance

Pricing and features checked September 2026.

Where CoreDash fits better

In a two week trial you won't see much difference. You notice it later, say six months in, when you want to know why something got slower or whether the fix that went live in spring still holds.

Old data keeps its detail

People often miss this. In Datadog the raw RUM events (views, vitals, long tasks, with all their attribution) are kept for 15 to 30 days. What's kept for 15 months are pre-aggregated metrics, split by a fixed set of dimensions like browser, country, app version and service. The LCP element, the phases and the script that blocked the frame are all in the events. You can build custom metrics from RUM events to keep more, but then you have to decide now what you'll want to ask a year from now. Didn't create a metric for the LCP element per page group in March? Then that detail is gone by December.

Datadog docs, data retention periods, October 2026.

Performance work takes months. A fix goes live, traffic changes, marketing adds another tag manager container, Black Friday brings a different device mix. In CoreDash every dimension stays with the data for as long as your plan keeps it: 1 year on Standard, 3 years on Studio and up to 8 years on Enterprise. That includes the URL, the LCP element and the LoAF script. "Was the checkout faster last November" is just a filter.

Built around how Google judges your site

CoreDash reports p75 against Google's thresholds and cross-references your data with CrUX, so you can see where your own numbers and Google's dataset agree and where they don't. Datadog doesn't ingest CrUX. An SRE team probably doesn't care, but if you're the one explaining a red Search Console report to your manager you will.

The dimensions are built around Core Web Vitals questions too. Was the LCP image preloaded, lazy loaded or neither (LCP priority)? Was the slow interaction a click or a keypress (input type), and did it happen while the page was still loading (load state)? Those are filters you just pick, no query needed. And Diagnosis points at the vendor behind a slow LCP, checks for 14 LCP anti-patterns and gives fixes for your stack (WordPress, Shopify, Next.js and others).

What your AI agent can work with

Datadog has an MCP server too, and for general RUM work it's good. Its RUM tools search and aggregate events, so the agent has to write Datadog queries against RUM events, and it can only look back as far as those events are kept.

The CoreDash MCP server is made for Core Web Vitals. get_metrics returns the p75, the rating and the good / needs improvement / poor split for any dimension. get_timeseries tells the agent whether a metric is improving, stable or regressing, and get_histogram gives the full distribution. It's included in every plan, Starter too. Same endpoint as the API, read-only, and it works in Claude, Claude Code, Cursor, Windsurf, VS Code and Gemini CLI.

And there's CWV Superpowers, a free MIT-licensed agent. It starts from your field data, runs a Chrome trace on the slow page with mobile throttling, finds the bottleneck and writes the fix with file, line and diff. Ask "which templates fail INP on mobile and why" and it answers from your real visitors, then profiles the page (here's an example). I haven't seen another tool that goes from field data to a trace to a code change.

Script size and pricing

The Datadog SDK is about 66 KB gzipped, 51 KB for the slim build. Fair enough for everything it does (errors, actions, replay and so on), but if you only want Core Web Vitals it's a lot of JavaScript, and the browser has to parse and run it on the main thread of the page it's measuring. The CoreDash collector is under 2 KB, async and passive. I'm probably more fussy about this than most, but I've seen a RUM tool get in the way of LCP before.

On price, Datadog's Measure tier is cheap. The detail you need to actually fix things is in the Investigate sessions though, at $3.00 per 1,000, and only for the sessions your retention filters keep. So what you pay, and what you can still dig into later, depends on how well someone set up those filters. CoreDash charges per pageview tier: the bill goes up with your traffic and every plan has every dashboard. If you run a consultancy, the agency plans pool pageviews across clients, and with one master key you can create projects and pull data for all of them through the Agency API. Datadog can do multi-client too, with multi-org accounts that their support has to enable.

When to pick Datadog RUM instead

• You already run Datadog APM and your slow LCPs are mostly slow servers. Then front-end to back-end tracing beats everything else on this page.
• You need Session Replay or error tracking next to your web vitals.
• Your company has standardized on Datadog and procurement won't sign for another vendor.

You can also run both. The CoreDash collector is under 2 KB, so adding it next to Datadog for Core Web Vitals and CrUX barely costs your page anything. Datadog keeps doing the traces, CoreDash keeps years of Core Web Vitals history for you and your AI agent.

Try it on your own traffic

The 14 day trial needs no credit card and data shows up from the first pageview. You can look around the demo first, or check pricing. Comparing more tools? See all comparisons.

Sources: Datadog pricing list, Monitoring page performance, Datadog on Long Animation Frames, RUM without Limits, RUM without Limits metrics, Data retention periods, Datadog MCP server tools, Multi-organization accounts. SDK size measured on the Datadog CDN (v7) in September 2026.