Hacker News new | ask | show | jobs
Show HN: See Google's Data on Real User Site Performance for Any Site (crux.run)
50 points by jloveless 2444 days ago
7 comments

PSA: be very careful with what you query in BigQuery on this dataset! You will be billed $5/TB scanned. I was using this dataset a couple weeks ago for fun, and ran up a $60 bill in about 10 mins. It’s important to understand that using WHERE doesn’t shrink the size of your query; instead your FROM clause should only be limited to a single date or month... not ‘*’
This is EXACTLY why we moved this data into our own database for Crux.run. We took the JSON export of the Crux database and cleaned and loaded it into a database on a single server. E.g each of those queries are hitting a single box in us central
There’s some of that information in the docs : https://docs.crux.run/#/introduction?id

But yeah- BQ can get very expensive very quickly if miss step

Thanks for doing so!
This is based on Google's Chrome User Experience database. Some details are available at https://docs.crux.run You could query this data with BigQuery - but this tool does all the aggregations for you (and has a REST API).
The input placeholder is confusing. It prompts the user for "https://" - but it doesn't require this.
That’s helpful- and especially since we then directly strip it :/
None of my sites are in the database.
They need 10k+ views a month for them to be in the CrUX database
Also the database is origin specific - so “foo.com” and “www.foo.com” are different records- same as the Google source in big query
Trailing forward slashes seem to confuse it. I pasted `https://news.ycombinator.com/` and it thinks the website is not in the database.
Our engineering team (at Plumbr.io) built this as a marketing tool: https://ismysiteslow.com/.

All of the reports we show are based on crux data.

Even though it’s a single metric point I like the comparison! It’s always fun to compare two web optimization companies. E.g Fastly and Cloudflare. https://crux.run/report?domain=www.Fastly.com https://crux.run/report?domain=www.Cloudflare.com
Interesting.

It shows bing.com having DOM load time of 4 seconds, and Google.com almost double at 7 seconds. Any ideas why this happens.

That is weird- I checked bigquery and see the same thing. Note that’s the p95 - maybe more mobile users for google.com than bing?
Can you segment by Geo? It may be that users with slower connections or phones use Google more
We can’t do geo segments (yet) on Crux.run but we could test against bigquery directly
I haven’t looked at it, but if the p50s are comparable, I’d say more mobile users on google.com is a good hypothesis.