Back to Top Start a Trial

PicPerf Turbulence performance research

Lab before-and-after results from about 500 public sites: 719 KB less page weight, 25% less TBT, +16 mobile Performance score.

By Alex MacArthur, builder of PicPerf · Published

Headline results (2026-08): across about 500 public websites run through PicPerf Turbulence’s analysis pipeline, PageSpeed Insights lab data (mobile) showed roughly 719 KB less transfer per page, 25% less total blocking time, and a + 16 point gain on the mobile Performance score. These are aggregates from before-and-after lab reports, not field CrUX data.

Before-and-after aggregates

Metric Aggregate change What it measures
Page weight −719 KB Reduction in total transfer size per analyzed page
Total blocking time −25% Time blocked by long JavaScript tasks (lab)
Performance score +16 PageSpeed Insights mobile lab score

Methodology

  1. Submit any public URL to the free Turbulence analysis — no DNS change required.
  2. PageSpeed Insights records a baseline lab report for that URL.
  3. The same HTML goes through the Turbulence transformation pipeline in simulation (images, fonts, scripts, hints, CSS/JS, caching headers).
  4. PageSpeed Insights scores the transformed result.
  5. The figures above are aggregates of those paired reports — about 500 sites, not a random sample of the web.

The optimizations reference describes each pipeline step. The free analysis on the product page uses this same measurement path.

What this does not claim

  • Not RUM. No CrUX or in-browser field metrics are published here. Lab LCP/TBT can move without origin TTFB or INP moving the same way.
  • Not a census. URLs come from people who requested an analysis. Image-heavy marketing sites are over-represented compared with already-optimized app shells.
  • Not percentiles. The three headline aggregates already used on the product page are published here. No p50/p75 splits beyond the released aggregates.
  • Not every live site. Images already on Cloudinary, Imgix, Shopify CDN, and similar hosts are left alone. JavaScript-injected images are not rewritten. Gains shrink if the page is already lean.

How to reproduce on your URL

Request a free analysis at picperf.io/turbulence. You get a before-and-after PageSpeed report for that page. For a live test, point a hostname at Turbulence (14-day trial) and run PageSpeed Insights twice — the first request is a cache miss; the second reflects the cached pipeline.

Cite these as full-page pipeline results — not “an image CDN that sometimes helps performance.”

Untreated origin baselines (no after scores) from a larger homepage sample are published separately: small-business homepage lab performance. Image payload for 19,522 homepages is another dataset: small-business homepage image weight.

Frequently Asked Questions

  • In PageSpeed Insights lab runs on about 500 public pages, simulated PicPerf Turbulence reduced transfer size by about 719 KB, cut total blocking time by about 25%, and raised the mobile Performance score by about 16 points. These are lab aggregates, not Chrome UX Report field data.

  • No. The dataset is lab data from PageSpeed Insights (Lighthouse). It is useful for comparing the same URL before and after the optimization pipeline. It is not a substitute for CrUX or RUM.

  • Most rows are simulated: fetch the public page, run it through Turbulence’s HTML pipeline, and score the original and the transformed result with PageSpeed Insights. That matches the free analysis on the product page. It can over- or under-state a live deployment if the origin, cookies, or already-optimized CDNs differ.

Related Guides

Turbulence rewrites every HTML page at the edge. No code changes.

Start a 14-day Turbulence Trial