Pinned Tweet
Everything you need to know about Shopify site speed in 3 minutes
2 web site comparison
True Classics: not fast
Asos: fast (loadmaxxing, perf-chad)
I use a totally badass, radical tool in this video. It's called FasterSite and it's my baby.
First tweet👇
Also think this will bounce back.
Outcome focused is the largest moat, as this is a skill issue
But AI hype makes brands willing to test efficiency on their own and sells dreams
Everyone makes a saas, realised a saas is another job to hire for, kills it and pays for the trouble to be managed
I suspect the worst is nearly over
The agency model has drastically changed over the last 12 months.
Every friend I know in the agency space mentioned how difficult of a year it has been, and I can also attest to how much the buying behavior has changed.
Let me explain what changed:
There are two types of activities one can offer.
There are output-based activities, and then there are outcome-based activities.
An output-based activity is what your typical dev shop, ad creative agency, or video editing agency would offer.
"8 videos a month for $4k." Whether these videos perform, flop, or become viral hits, they do not take responsibility for the performance.
As long as their deliverables are approved, they get paid.
This type of activity has been completely eaten up by AI in the last 12 months.
Every agency that sold output has had a very difficult year.
On the other hand, there are outcome-driven activities.
These are typically more skilled specialists who would come into your business with clear KPIs.
"We'll increase your email conversions by 50%"
"We'll generate you 1m views in the first 90 days"
"We'll build you a paid ads funnel that pays for itself"
These are outcome-based roles that require industry knowledge, specialization, and years in the game.
Even that group felt some strain this year. But not as much as the output-based group.
The agency model has shifted towards outcomes. And that is much more difficult. Which is why I foresee a drop in new entrants building out agencies.
PS: The only group of output-based service providers that has not been disrupted are those with proprietary access to connections or data.
For example, if you want to speak on a TEDx stage or be featured on CNN, only a few people have that type of access. You can't Claude Code your way to a TEDx talk. Therefore, access and proprietary data are a moat that won't get replaced anytime soon.
Agents are great, but there is a point where faster / cheaper isn’t worth anything.
I will wait an extra 3 months and pay 2x for the doctor with the good rep.
Services need to live here
Also, LLM prices will go up and push casuals out. Few players will solve singular problems and leverage economies of scale
62
This code snippet shows you which pixels are slowing your store in 30s
Just tested this snippet and it caught 182 tracking events from 23 domains on one Shopify store (in about 30 seconds of clicking around).
The problem:
> the add to cart button is choppy / slow
> users see this slow experience
> this adds friction - they don't reach checkout
We measure this "delay" using INP, and INP has been linked to lower CVR by the shopify team themselves (2026 April, Shopify Perf Blog - "Every 32ms of INP = -1.5% CVR").
Sometimes your site is slow because your theme is old, but often - it's because you have +14 pixels all firing the moment you tap. Meaning this is totally fixable WITHOUT a 6 month re-theme project.
I've personally improved INP for brands, and seen instant CVR lifts (Escentuals dropped their INP 25% and CVR went up 6%).
But a typical speed score won't show you any of this INP stuff.
This is for you if:
> Add to cart or switching variants feels sticky on mobile
> You run ads, analytics, reviews and a heatmap tool at the same time
> Your dev says the theme is fine and you're not sure who to believe
It shows you which pixels fire on each tap and which ones double up. Then you can cut the duplicates (or move them server side) before you spend a cent on the theme.
It's a copy + paste:
> Copy the snippet
> Paste it into a draft theme
> Refresh your store and click around
The panel shows every event, where it went, and the exact moment the page got too busy to respond.
Watch the video to get the snippet!
3
4
133
I am becoming too powerful at my local Gail’s
The first one they didn’t let me pay. The second one was delivered randomly
1
11
393
How pixels work and why adding them kills your campaigns
INP = when your website lags.
High INP = lower CVR
(every 32ms of INP = -1.5% CVR)
... you can see where this is going
Pixels cause lag, because of their event-driven system. They wait for something to happen, then all fire at once.
Not a big deal on a single pixel, but this website has 14, and 12 of them fire code execution every click, and 8 of them then immediately send a network request.
Worst part, most of them are using duplicate data (ie, they all send screen size and page name). Which you maybe use only on 1 pixel.
There's only really 2 ways to fix this:
[1] Consolidate
Unless you're directly running ads with this pixel, try to assess if you're ACTUALLY using them. Like, do you make weekly / monthly reports with them? If not - bin them.
Don't have duplicates. You don't need heatmap + clarity (both do screen capture, pick 1).
[2] Server side tracking
Complicated to set up, and requires maintenance - but keeps all your data.
You basically create a singular pixel which captures ALL the data each pixel uses, then send it once to your server. Your server then distributes these events across your other pixels.
Simple - not easy.
--
This was top of mind because I'm doing research into pixels. If you want to be up to date with the latest changes in site speed and stay ahead of the curve - join the newsletter here:
izaac-barratt.kit.com/85ee6f…
It's where I document my test and learnings.
1
71
specific content is unmatched
have previously had weeks of doing 20k impressions, no calls booked.
yesterday, single post with 200 impressions booked a call.
the post was an audit of a brand I'm trying to work with...
My content strategy may be changing
74
Building something that stops Shopify apps slowing the site. Nobody wants to delete them, everybody wants faster pages.
Results so far, we maybe onto something
2
214
when trying to code/prompt something - I am overwhelmingly encouraged when it's not working
I am dealing with a real problem that cannot be fixed by somebody less technical than me
1
137
The safest way to lift CVR 5-10% for BFCM
1 - go to your theme and list all the apps currently active
2 - put them in a spreadsheet with columns (last used, feature used, currently needed?)
3 - disable everything not currently in use
4 - disable everything not used in > 30 days
5 - anywhere "feature used" overlaps, consolidate
Don't even need to delete - just turn them off on the store.
Pages will load 10-20% faster, ATC clicks will feel snappier.
Less traffic will bounce, more traffic will continue through the funnel.
Nothing visually changes on the site.
1
68
Keep Shopify fast even after years of old apps (5 steps I always use)
I spoke to 9 brands over the last 2 weeks, 5 of them considered 'old app code' to be their biggest performance blocker.
When you've been working on a site for more than 3 years, you will have app bloat.
Marketing adds an app for a single feature, then forgets to remove it.
Finance adds a tracking tool to generate a report. This puts snippets in the theme code that never get removed
And now, after 3 years of little changes, your site randomly becomes so slow even the internal team bounce (reload the page because they don't want to wait for it).
Nothing wrong with adding or removing apps, but if you don't have a process behind cleanup, it gets messy over time.
> Slow interaction on key pages (PDP variant changes)
> Very slow menu drawers or filter menus (sometimes broken)
> Images and videos very slow to load (vertical chunks)
As such, cleaning app code has become a big part of our internal process.
The 5 step process we use to keep Shopify stores fast for brands like Cozy Earth, FC Barcelona, Dryrobe, & David Protein.
Steps 1-4 are event based, step 5 should be routinely done.
--> On Adding an app
[1] Compare live websites
Look at the Shopify app listing page, and find some reviewers. Then look on their website and see how this app performs.
If using it feels slow, or this website itself happens to be slow - it's worth investigating more.
Prevention is better than cure.
[2] Compare before vs after
Never download an app before checking your current site speed (use Shopify analytics or Page Speed.dev for this). Check a few minutes after installing.
You don't need to be super granular with this, you just need to catch which apps have an impact, and get a feel for how much. Over time you will become more sensitive and instinctively know what is and is not worth keeping.
--> On Deleting an app
[3] Search HTML elements
Once removed, go to your page, right click -> inspect. In the network tab and the elements tab (at the top) - search for any mention of this app.
This finds any network requests still being made (like API calls), and catches any old snippets that have been left in the code.
[4] Search Theme code
In the code itself, search for any instance of the app you just removed. Search the domain, and search the name, then search any acronyms you can think of. For example, baseline commerce could be: bc, or baseline, or bcommerce.
Delete everything you find.
--> Quarterly
[5] Review apps
When doing app audits for brands, I suggest a giant list which covers:
> Who added the app (we need accountability)
> What feature you use (SPECIFIC features, not general purpose)
> How often it's used
Over time, you will start to get a lot of apps whose features overlap. Routinely look for opportunities to consolidate apps.
From experience when auditing Shopify stores have more than 7 apps active, LCP goes over Google's recommended threshold (2.5s LCP). So for now, I advise trying to be under 7.
___
Carry out all these steps, and you'll have no problem when adding / removing apps.
What did I miss?
2
3
138
POV: $10 theme update made your apparel brand +$30k in 2 days
The play-by-play:
> You moved 10% of your slow sessions out of the 'speed tax' zone.
> This 10% had an instant CVR lift
> You make an extra 30k that month
Sounds crazy, but this is from real data we tracked.
It happens because all stores have a speed limit where they start bleeding sales.
I call this the 'speed tax'.
And for the brand we were tracking, it's was a 30-50% drop (yikes).
It was causing by something I see on almost EVERY slow Shopify store... old / messy theme code.
If you've had a store for +2 years, you will have this.
Ways to know it's impacting you:
> Slow page load (videos / images don't show for several seconds)
> Menus randomly break (cart drawers / filter menus)
> Customers complain about things like ATC buttons not firing
It's one of the worst things to happen to a store, because it amplifies all other issues. Slow site + existing bugs = insufferable.
Even if ads, and community are tight - you still have to use the site to buy.
But you can do something before BFCM that totally fixes this and literally costs $10.
You can clean up your theme.
My name's Izaac and I run Baseline. We ONLY do site speed and we've done this for David Protein, Cozy Earth, Dermalogica, and FC Barcelona.
Over the last 9 months our avg results are:
> 75% faster pages
> 15% CVR lift
And for $10, you can get a theme clean which speeds up stores 10%
It's how we start every project, and is the safest work we do. The dental equivalent to a checkup and floss.
What it includes:
- Trim old app code
- Refine HTML structure
- Tidy basic CSS + JS
Delivered in 48hrs on a draft theme (publish or delete with a click) and WILL NOT touch:
> apps
> design
> features
Nothing deleted, just organised better.
Instantly faster.
The catch? You have to let us measure the impact it has on your site speed and make a case study.
We have 3 slots available and will be starting ASAP.
So if you'd like to turn $10 -> $30k
WATCH THE VIDEO TO ENTER!
4
3
182
born to say I'm building things beyond human comprehension
forced to admit there are 10-15 people in the world that care about this thing - but in this small circle, they would be blown away
1
56
Everything you need to know about pixels on Shopify (deep-dive video)
Site Speed Audit 11/1000 - Desktronic
First off - great work on LCP, team!
< 2s is great. No comments here.
But, INP could use some a bit of love. In particular, on PDPs when changing variants.
Big pixels and complicated CSS mean every event where you update the browser has extra baggage.
Most pixels wait for an event like a button click, or a page navigation, then fire a bunch of capture code.
Combined with the fact your code is already doing something tricky (swapping out product images), creates a backlog of tasks.
Browsers can only do 1 thing at a time, so if we interact during a large task - CPU is slammed and can't respond to our clicks.
If we have a bunch of big tasks that happen on click - CPU is slammed.
You get the picture.
The 80/20 recommendation:
[1] Look into your gtags and see if there is any code you can simplify. The less calculation code here, the better.
[2] Reduce CSS complexity.
Look into BEM CSS, it's a method of naming things specifically for their area. Doing this means any time an element updates, it only needs to check other elements in the surrounding area - not all elements with the same tag.
You want your CSS selector update count to be < 300 (ideally)
--
Hope that was useful!
If you want to have your site audited, drop a link and ask some questions
The worse performance the better! I love seeing the crazy outliers
148
3 theme changes made @CozyEarth 51% faster before BFCM
LCP = page load time
INP = interaction delay
When we first spoke, their site was slow. One PDP was +4s to load (1.5s over the LCP limit), and +600ms to respond to taps (+400ms over the INP limit!!).
Which is natural, after having spent years building on top of a theme. Mess builds up.
But eventually you reach the position where you have to either:
- re-theme
- go headless
- tackle all the clogged code
The cheapest and fastest is always the last one haha. But that doesn't make it easy.
After 3 weeks, we wen't live with a new build, and I'm pleased to report:
> 51% faster LCP (pages load much quicker)
> 34% faster interaction (much easier to navigate PDPs)
In particular, there were 3 things we did to get this result (organised by impact).
[1] - Clean up HTML
This one is surprising (it surprised me too). It's easy to get caught up with modern browser tech and focus on the complicated things. Sometimes, you just have to tighten the fundamentals.
They had a bunch of inline code, and complicated HTML. Doesn't sound like a big deal, but takes longer for the page to download AND takes long for the page to process.
Displaying the page is actually a huge strain on the CPU, so making it easier to display has a surprisingly large impact on INP.
[2] - Correctly ordering scripts
This is a staple of every project we do. The load order determines the load speed.
Priority things go first (and ideally only a few), heavy slow things go last. It's that simple.
Render blocking scripts stops your browser reading the page, so we only want these AFTER it's read enough of the page to start printing it.
[3] - Remove / defer invisible assets
You'd be surprised how many images are loaded on a page before they're visible. Cart images, menu images.
In one case, we actually found some old code which was part of an A/B test that had stopped running. We removed it and the HTML was 30% smaller instantly (literally, probably the single largest impact from only changing 1 file commit).
The crazy part is - we haven't even finished all the work!
We've got some liquid optimisation changes which should improve LCP, and more excitedly - we're migrating them to a server side approach for pixels. Aiming to get their INP < 150ms consistently!
Been a pleasure to work with the team and looking forward to how far we can push it!
If you want a second pair of eyes on your Shopify, grab yourself an audit here
baseline-commerce.com/offers…
See the top 3 things you can action before BFCM
2
79
site speed impact on paid traffic for 9fig brand over last 7 days.
after 2s, bounce rate climbs consistently and CVR falls off
if LCP p75 is > 2s, would seriously consider looking into this for BFCM
1
3
72
8fig Shopify is sells ~30% less when loading in +3s, and 51% less at +5s
Being able to measure session level speed and attribution is a game changer
This brand is mostly organic traffic - but in future, will be able to tell at the ad-campaign level, which pages need to be faster to increase CVR
1
1
89
3 ways to make Shopify fast without deleting your apps (proven on 9-figure brands)
Last 9 months our avg result was +70% faster LCP, and we worked with David Protein, Cozy Earth, FC Barcelona, & Dermlaogica.
There are 3 approaches that help us (an external team with no control over stuff like 'deleting apps') get best-in-industry results, and generate an extra $1.3M in CVR lift.
1) Sequence over size
It's more about how you load files than how small your site is. Sure, if you delete all your apps the site will be fast - but it's not feasible.
Make sure you load things in the correct order instead of leaving it to chance.
You can make a 6s website look like a 2s website if you aggressively prioritise what's above the fold (true story).
2) Squeeze the juiciest lemons!
Everybody looks at apps and external tools and blames them for speed.
Okay - now what? We can't control them.
What can we control?
Let's over index on that! Over compensate, make it SUPER lean. It's not about not having anything slow on the site, it's about reducing the total time.
An extra 1s taken from an area that is fast is still an extra 1s faster.
If you have 2 stones and 1 lemon to make lemonade.... what would you spend your time on?
3) It's a department wide effort
I know I just said you don't need to delete things... but it really would help.
Most times, the app buildup is from different departments.
Finanice, marketing.
Get everybody on the same page. Create a big spreadsheet of all the apps, and assign the individual (NOT DEPARTMENT - we need a name) who is responsible for each app.
Routinely ask them if they still need it.
-
And if you want the code snippets that we use to implement this stuff above - there are instructions in the video to get this!
What did I miss?
2
8
405
This pixel shows how many sales you lose to slow pages (literally down to the ad channel / device)
We had a crazy year making sites faster for David Protein, Cozy Earth, FC Barcelona, and Dermalogica.
Across all of these sites, we've had the same question - what is the EXACT point Shopify loses revenue from slow pages?
What this pixel does:
> Compares CVR to LCP down to the session level (per user)
> Where speed impacts bounce rate (how long before paid traffic quit?)
> Measure it down to the ad campaign, and user device
If you're curious about the ROI of fixing your site speed, or how much it's REALLY impacting users - drop me a comment / message.
You can get an air-tight answer you can build real plans from.
Exciting times!
What data would you want to see the most?
one of the things that enforces clean code is the fact it was tedious to write by hand. So we found areas to duplicate / reuse.
AI writes so quick, it will just rewrite a whole new class / function with 80% overlap.
Main thing I have to tell it is - reuse x here.
Wonder if this is being not worth our time anymore?
1
3
292






