Want to beat your competition on page load time? Seeing too many abandoned carts? Slow sites cost you money:
→ 1-second delay = 7% fewer conversions
→ 3 seconds = 40% of visitors gone
→ Slow loading = Google dropping your rankings

I help ecommerce brands and agencies fix these problems without the technical headaches. Whether you need to boost Core Web Vitals, stop layout shifts frustrating your customers, or simply convert more visitors, I've got practical solutions that actually work.

Following (or connecting) me here on LinkedIn will be a first step towards increasing pagespeed, browser and debugging knowledge. 

Need a speed boost or team session today? There's a faster way:

As a Google Developer Expert who's helped brands like Danone and Wehkamp, I speak plain English or tech jargon, tailored to your team.
I'll work directly with your team to make performance improvements (and revenue) stick via interactive sessions. With the ultimate goal to get fast, stay fast and increase your (customers') SEO and revenue month on month.

What started as a developer journey in 2001 (building my own CMS because... why not? 😜) evolved into creating RUMvision—a tool that shows how real people experience your site, not just what Google's lab tests say. 
I'm also an international speaker, having shared insights on how to improve web performance at conferences like Performance.now(), We Love Speed, MageTitans and others. 

I'm platform-agnostic: Magento, Shopware, BigCommerce, Hyvä, Shopify, SPA's like React or Vue or custom builds—if it loads in a browser, I can make it faster.

Follow me for tips that make web performance less boring, or let's chat about how to stop losing sales to slow load times.

Yes, I know it's Friday… but I deployed this anyway! 😎 

We added the No-Vary-Search response header to a bunch of sites to ignore tracking parameters like `utm_*`, `gclid`, and `fbclid`.

They’re consumed by third party tracking scripts, but they shouldn’t create separate cache entries or duplicate URLs. 

By telling browsers to ignore them for caching, the same page can be reused instead of fetched again.

*edit: CDNs don't support it yet, but it would be great if CDNs started honoring it out of the box!

Result:
Cleaner caching, fewer unnecessary URL variants, and a slightly faster site without breaking analytics.

Do note:
You would still want to add them to your local caching mechanisms! 💡 

Let's make cache popular again! 🤓 

The copy and paste version:
params=("utm_campaign" "utm_content" "utm_medium" "utm_source" "utm_id" "utm_term" "brid" "fbclid" "gclid" "wbraid" "gbraid" "gclsrc" "gad_campaign" "gad_campaignid" "gad_source" "msclkid" "epik" "pp")

Erwin Hofman

Core Web Vitals consultant for ecomm & agencies | Google Developer Expert | audits, talks & in-house training 🎓

Nice #caching finding in a webshop #sitespeed audit today!
This merchant did something smart: they moved product data out of the HTML and loaded it via JSON instead.

That way, their HTML page became much smaller!

Result:
→ HTML stayed well below Google's 2MB uncompressed crawl limit
→ Collection page including facet filters/search renders faster

Filtering or pagination then loads the products via a JSON request.
Then, rendering of those products can happen.

And to speed that up, they added a `<link rel="prefetch" ..>` for the same file.
Very smart and nerdy 🤓 

But there's a BUT (otherwise this post wouldn't exist 😅).
→ the Network tab showed something odd
→ the browser never actually used the prefetched file as the browser considers it different request types
→ so users ended up downloading the JSON twice per collection page
Ouch!

Luckily, this can be solved! Either via:
1️⃣ proper (Cache-control) response header + additional #prefetch attributes
2️⃣ or using the `cache: "force-cache"` parameter for your `fetch` call (convenient if you can't easily change response headers or the generated HTML link element) 

Easy fix, big win! 😎

Google quietly lowered the max crawlable size to 2MB (did not go unnnoticed by you / the #SEO community though 😜).

But was this actually affecting sites in reality?

So I decided to look at real user data instead of opinions.
By tracking the #uncompressed #HTML size at RUMvision - Core Web Vitals monitoring.

Resulting in many millions of HTML datapoints, which we bucketed by size.

Here’s what we found:
→ About 7.8% of pages exceed 2MB
→ Over 90% sit safely below the limit

So yes, you could say most sites are fine.
But almost 1 in 13 pages are now partially ignored by #Googlebot.

Like this example of >6MB 🤯

Every #WordPress developer knows #RocketLazyloadScript
A while ago, I discovered it had an interesting bug. Last week, I learned it got fixed by the WP Media / WP Rocket team! 🎉

My followers know I’m not the biggest fan of aggressive #JavaScript lazy loading plugins:
→ Features sometimes don’t work on first interaction
→ It can hurt real #UX in ways that don’t show up in #Lighthouse (CLS, INP)

But that didn’t stop me from reporting this bug.

𝐖𝐡𝐚𝐭 𝐡𝐚𝐩𝐩𝐞𝐧𝐞𝐝?
When scripts were delayed, `onload` listeners were not always executed correctly.

For websites using RocketLazyLoadScript, this could mean third-party scripts not firing on every pageview = missed analytics or website features.

We spotted this in RUMvision - Core Web Vitals monitoring while monitoring #CoreWebVitals with Google’s official web-vitals library:
TTFB was underreported because the library relied on the onload event, which wasn’t consistently firing in this setup. So:

👨‍💻 I went down a rabbit-hole (first thinking it was a bug in web-vitals)
✍️ Then filed a bug + detailed report.
✅ The fix has been shipped in 3.20.4

𝐍𝐨𝐰 𝐰𝐡𝐚𝐭?
Also using the RocketLazyloadScript plugin, just like my client?
Be sure to update the plugin if your website is using third parties! 💡

With #CoreWebVitals data, we usually help agencies and merchants improve their sites and shops. But sometimes the data helps browsers too.

An example: 
We recently spotted very inflated INP numbers across many sites on mobile (P75) specifically for `back-forward-cache` navigations. 

So I segmented it per browser and saw this:
→ Google Chrome: 53ms
→ Safari: 11013ms (see attached, 207x higher than Google Chrome)

And I'll admit: Before calling something a browser bug, I always assume the problem is on my side first 😅
So I double-checked the methodology, the segmentation, and the raw data.

But the pattern was very consistent across sites and datasets. Moreover:
1️⃣ Safari started supporting INP relatively recently
2️⃣ and another edge case was discovered and reported before (by Barry Pollard). 

So I ended up filing a WebKit bug report including:
→ screenshots
→ percentile distributions
→ Chrome comparison data

There's already good news too:
The WebKit team created a pull request (within barely 2 weeks) to address the inflated INP values for bfcache navigations 🎉 

So, a fix should be shipped soon!
(and yes: it's nice and nerdy to know that we contributed to this 🤓)

Reminder for sites that want their website to be faster than their competitors! 🚀 
There's an open Google Chrome origin trial for prerender-until-script.

A bit of background info:
> Prerender until script is a new, middle ground, between prefetch and prerender. It prefetches the HTML document (as prefetch does), and then starts to render the page, including fetching all the subresources (as prerender does).

For example, if you're currently use:
1️⃣ prefetching, but think it could do/prefetch a bit more
2️⃣ or prerendering, but think it's too aggressive or will skew analytics
prerender-until-script might turn out to be a better candidate!

Using RUMvision - Core Web Vitals monitoring? 
Then we've made things even easier:

Check your tracking settings or reach out to us, and we'll make sure the origin-trial metatag will be injected, so it will work for all Google Chrome visitors 💪

Last Friday I implemented `prerender_until_script` on a few sites. But you might know how to get started with origin trials, or are afraid to break your current #SpeculationRules strategy (or both).

So I extracted the most important bits from our #JS implementation and pushed it to github (which in short looks like the attachment) 🤓 
That should make it much easier for others (like you!) to get started with `prerender_until_script` during the trial 💪 

My script:
1️⃣ checks for and injects the origin-trial token if provided
2️⃣ only enables itself for the #Chrome versions where the trial is active (up until v150)
3️⃣ and can be dropped in with minimal setup

Until v150, so there's still some time!
By giving #prerender_until_script a shot (with or without my code) you're also helping to shape the web!

Already using RUMvision - Core Web Vitals monitoring for tracking your shop's #CoreWebVitals? 
Let me know and we can enable the token out of the box + help track effectiveness of this new Speculation Rules type!

It’s already March, which means event season is back in full swing! And the perfect moment to look back at the stages of 2025. Afterall:
→ events are like tango: you also need speakers and visitors
→ and last year, each one "shined" in its own unique way

A recap of the highlights:

🏝️ Hyvä Developers Paradise 
The ultimate spot for Hyvä, Magento, and fast frontend lovers. 

The location and athmosphere was stunning! 
And if I’m to believe Karlijn Löwik, the pre-party was legendary 😜

→ Next up: in Marbella this April. Closer to home, but just as sunny! ☀️ and we'll be there

🎬 Fronteers DarkMode
A haven for frontend enthusiasts covering everything from #performance to #accessibility. It was nice to finally meet the many LinkedIn frontenders connections in real life! 

This was our second "cinema talk," held in the oldest city in the Netherlands: Dordrecht.

The organizers went the extra mile with a boat tour, a speaker + event dinner, and even custom posters to take home 😍

→ Next up: October 2nd, 2026. I’m really considering heading back, this time as an attendee! 🤓 

🕵️ SEO Benelux
The international hub for all things #SEO and live roundtables. I stepped in as a last-minute speaker (and might have gone a bit over time as a result 😬).

Top-notch venue and staff. Even the flaky Wi-Fi helped me prove my point about #sitespeed! 🐌 
Also, those cozy couches in the back are the perfect place for a speaker to get into the flow!

→ Note: Sadly, Mathias Noyez announced no third edition of the conference is planned, but the meetups are still going strong.

🦢 Sylius Conference 2025
A warm welcome into the Sylius and Symfony community in Lyon. I certainly didn’t feel alone with folks like Ksenia Zvereva (as if she was following us around every event, or other way around?) and others! 😀 

The venue was massive (see picture, a real conference hall), and Lyon was a revelation. I walked my shoes off exploring the city (twice), and the locals were patient enough to let me practice my French

→ Note: no signs of a new conference yet, but new Sylius meetups happening this + next month!

𝐎𝐭𝐡𝐞𝐫 𝐇𝐨𝐧𝐨𝐫𝐚𝐛𝐥𝐞 𝐌𝐞𝐧𝐭𝐢𝐨𝐧𝐬 (from a visitor perspective):

Although I didn't speak at these events (Karlijn did though at 1 & 2), we also visited:
1️⃣ Hyvä Camp Kiel by Splendid Internet GmbH which came with an on-site child-care service (happening again this year mid-June)
2️⃣ Meet Magento Italy 2026 by Andrea Saccà which came with a night walk through Rome 🏛️ (happening again June mid-June, LinkedIn name has been changed to 2026 already👌)
3️⃣ performance.now() conference, I mean: how can we not visit this one? It was our duo-talk in front of the biggest audience back in 2024. The people, pre-events, talks and on-site food is always top-notch!

What events are on your 2026 calendar and/or should I visit? 
My next one is by SEO Brein in De Kuip this month!

This client expected an LCP improvement, but got a regression instead.
A few days before the regression kicked in, they improved image delivery from their CDN.

So when I saw a regression, it immediately felt odd. That change surely should not have caused it!

But then:
→ they updated the Swiper library on a listing page
→ they didn’t tell me
→ and didn’t create an annotation for it either

𝐓𝐡𝐢𝐬 𝐡𝐚𝐩𝐩𝐞𝐧𝐞𝐝:
1️⃣ Swiper turned images into lazyloading, so `load delay` increased
2️⃣ As of that moment, images (and especially LCP) started to feel the pain of multiple simultaneous downloads, so `load duration` aka download time regressed as well due to increased download competition
3️⃣ After downloading, the image still had to wait for JS execution before rendering, massively regressing `render delay`  part

𝐋𝐞𝐬𝐬𝐨𝐧𝐬 𝐥𝐞𝐚𝐫𝐧𝐞𝐝:
With a single library update, multiple sub-parts regressed, leading to an 99.3% LCP regression for images on listing pages.

Just `render delay` alone ended up close to 2.5 seconds.
Which becomes a direct problem: Getting your overall #LCP below 2.5s becomes impossible if 1 out of 4 sub-parts is already reaching that number.

𝐋𝐮𝐜𝐤𝐢𝐥𝐲:
→ they reverted the change quickly (see the end of the chart)
→ so it will likely not impact their #CoreWebVitals nor SEO

Because: #SEO uses a 28-day rolling #CrUX window, while RUM shows problems in real time.

They thanked me for spotting this. But real credits go to detailed RUM breakdown aka sub-parts data from RUMvision - Core Web Vitals monitoring 🤓

Yes, you can build a Single Page Application. But should you (and how)?
Two important factors: #performance and real users

Attached is data from a South African site:
→ Internet speeds there are typically very different from many other regions
→ On top of that, 61% of visitors are NOT on a fast device

It's easy to imagine how that combination leads to bad #UX.

Things that also didn't help in this case to get rendering started:
→ shipping a bundled React #JavaScript file
→ of almost 10MB
→ without any compression

That combination will certainly push back rendering and seriously hurt your #CoreWebVitals 😬

Lesson: Before doing a heavy replatforming, get to know your audience (both their typical internet speed and used devices) via analytics (most RUM tools such as RUMvision - Core Web Vitals monitoring already tracks this by default)💡

taplio

Erwin Hofman's Linkedin Analytics

Erwin Hofman

Performance ScoreErwin Hofman's LinkedIn Performance

What is Erwin talking about?

Erwin Hofman's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Erwin Hofman's Linkedin Analytics

Erwin Hofman

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Erwin Hofman's LinkedIn Performance

What is Erwin talking about?

Erwin Hofman's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Erwin Hofman's Linkedin Analytics

Performance ScoreErwin Hofman's LinkedIn Performance

Erwin Hofman's Best Posts (last 30 days)