Shopify Just Tied Every 100ms of LCP to a 3.5% Conversion Drop
Shopify published 28 days of conversion data on April 27, 2026 showing that every 100 milliseconds of slower Largest Contentful Paint costs roughly 3.5% in conversion across actively-selling stores. Stores hitting a 2.5-second LCP convert about 30% lower than stores at 1.5 seconds. Cumulative Layout Shift, the metric most CRO teams audit first, showed no meaningful correlation with conversion at all.
The full report by Mateusz Krzeszowiak landed on Shopify's enterprise blog and was written up the same week by PPC Land. The methodology is what makes it interesting: this isn't one merchant's funnel, it's the median across the platform's global ecosystem, with the slowest 5% of stores excluded so a few broken sites don't drag the curve.
The number every CRO team has been asking for without one
If you've ever sat in a sprint planning meeting trying to convince an engineering lead that page weight matters, you already know the problem. The Amazon "100ms = 1% of sales" line gets quoted constantly, but it's from a 2006 internal memo and Amazon never published the methodology. Mobify's 1.11% conversion lift per 100ms came from a single brand on a single template. Portent's "1-second site converts 3x faster than a 5-second site" was useful but didn't isolate variables.
What Shopify just published is the first defensible benchmark from a sample size that cuts across thousands of merchants, multiple verticals, and 175+ countries. That matters because, in my experience, the engineering pushback is rarely "we don't believe speed matters." It's "we don't believe your specific number." Now you have one with methodology you can actually point at.
What the LCP, INP, and CLS split actually says
The three numbers aren't equally useful and the report mostly buries that. Here's how I'd weight them as a CRO operator:
LCP: 100ms slower = ~3.5% lower conversion. This is the hero stat. LCP is measured at the moment your hero image, headline, or product photo finishes painting. It's the closest Core Web Vital to the user's gut feeling of "this page is slow." Cutting 200ms off LCP on a $1M store at a 2% baseline conversion rate is roughly $14,000 in annual revenue, before you account for any AOV uplift on faster checkout. It's a calculator I'd run before the next sprint planning meeting.
INP: 32ms slower = ~1.5% lower conversion. Interaction to Next Paint replaced FID in March 2024 and most teams haven't built the muscle to optimize for it. INP measures the worst-case lag between a user click and the next visual update. Third-party tag bloat absolutely crushes this metric, which is why Shopify's CWV pass rate quietly dropped when INP replaced FID, and why a lot of GTM-heavy stores still haven't recovered.
CLS: no meaningful correlation. This is the result that should rattle a few SEO consultants. Cumulative Layout Shift measures unexpected visual jumps as the page loads. Half the audits I've seen lead with CLS because it's the easiest metric to get to green. According to Shopify's data, fixing it doesn't move conversion. It might still help SEO, since CLS is a confirmed Google ranking signal in the official Core Web Vitals docs, but as a CRO lever it appears to be a dud on this dataset.
The CLS finding is the weirder story
Most CRO teams I've worked with would have predicted CLS to matter at the margin. The intuition is that visual jumps cause mistaps, frustrated bounces, and a vague feeling of cheapness. Apparently not, at least not at the resolution Shopify can measure.
One reading is that Shopify's themes have already engineered most layout shift out of the storefront. If 80% of stores are passing all three CWV thresholds, the variance between "passing" and "barely passing" CLS is probably too narrow for a conversion signal to emerge. That's plausible but it's also a reason to be careful about extending this finding beyond Shopify. A WooCommerce or custom-cart store with messy layout shift might still see CLS bite.
The honest take: I'd still fix bad CLS for the SEO and accessibility wins, but I'd stop letting it dominate a CRO audit. Move LCP and INP to the top of the queue.
Translating 3.5% into engineering hours
Here's the spreadsheet I'd build before the next stand-up. Take your store's annual revenue, multiply by 0.035, and that's the value of every 100ms of LCP improvement on a recurring annual basis. A $5M store is looking at $175,000 in conversion uplift per 100ms shaved. A $50M store is looking at $1.75M.
That math is the case for hiring a performance engineer or budgeting for a CDN upgrade. It's also the case for finally killing the four marketing pixels you stopped using in 2024 but never removed because nobody owned the tag governance.
The standard 100ms wins on a Shopify store, in roughly the order I'd attack them: lazy-loading hero images that should never have been deferred, deferring third-party scripts that don't need to fire on first paint, switching font loading to swap, and pulling the "above the fold" content into a single render pass. None of these are technically hard. The reason they don't get done is usually that nobody on the marketing side has put a number on the cost of not doing them. Now you can.
Where this study breaks down
A few things worth flagging before you walk into the meeting waving the chart.
First, this is correlation, not a controlled experiment. A faster store also tends to have a more competent operator, a cleaner theme, and probably better creative. The 3.5% per 100ms could be partly an artifact of the kind of merchant who keeps their LCP tight in the first place. Shopify says it controlled for other CWV metrics when isolating each signal, but it didn't isolate operator competence and probably can't.
Second, the bucket comparisons assume you can move from a 2.5-second LCP to a 1.5-second LCP cleanly. In practice, the last 500ms of LCP optimization is dramatically harder than the first 500ms, and the cost curve isn't linear. Don't promise a 30% conversion lift to your CFO and then deliver a 4% one because the engineering ROI tapered off.
Third, Shopify excluded the slowest 5% of stores. If you're in that bucket, your numbers might be much worse than the report suggests, and the fixes might be much cheaper. Run your own LCP measurement on real-user data via Chrome UX Report or Search Console before assuming the 3.5% slope applies to you.
The case to take to engineering on Monday
Pull your top-10 landing pages, get the 75th percentile LCP for each from Search Console's Core Web Vitals report, and identify the 2-3 pages over 2.5 seconds. Those are the ones eating the most revenue per page view and they're usually fixable in a sprint, not a quarter. If you've also been staring at the June 15 consent cutover wondering where you'll claw back conversion, this is one of the few levers left that doesn't require negotiating with Google.
I'd give it about three months before this 3.5% number is the standard line every speed-optimization vendor cites in their pitch deck. Your engineering team is going to hear it from somewhere. Better that it lands first as a credible request from marketing with a calculator attached, not as a Forrester report someone's CMO forwards down the chain six weeks late.
Notice Me Senpai Editorial
