100ms LCP Delay = 1% Conversion Loss: Where the Number Comes From Benchmarks
The "100ms page-speed delay costs 1% in conversion" stat gets quoted everywhere. Here's where it actually comes from, which studies generalise to online retail, and the honest elasticity range to use when modelling script-removal revenue.
100ms LCP delay = 1% conversion loss
A widely cited rule of thumb that every 100ms of added Largest Contentful Paint costs roughly 1% of conversions on retail sites.
The number traces back to three studies that get cited interchangeably but measured different things: Akamai's 2017 retail performance report, Deloitte's 2020 'Milliseconds Make Millions' study with Google, and Google's own Core Web Vitals research. None of them measured LCP specifically against conversion in the form the rule of thumb implies — Akamai used page load time, Deloitte used mobile site speed in aggregate, and Google's CrUX data correlates LCP buckets with bounce, not revenue.
The honest synthesis: for online retail in the €1M–€15M revenue band, a 100ms LCP improvement on mobile product and checkout pages is worth somewhere between 0.3% and 1.2% in conversion rate, with the median credible estimate closer to 0.6%.
The '1% per 100ms' line gets repeated in every site-speed pitch deck, but the original studies don't quite say that. If you're building a business case to rip out a heatmap tool or consolidate tracking snippets, you need to know which number you're standing on — because finance will ask, and the wrong citation undermines the whole argument.
This page unpacks each source: what they actually measured, on which traffic, with what definition of 'conversion,' and how well it generalises to a Shopify or WooCommerce apparel store doing €3M a year. The short version: the directional claim holds, but the elasticity is softer than the slogan and varies wildly by funnel stage.
The three studies behind the '100ms = 1% conversion' claim
| Study | Year | Metric measured | Sample | Headline finding | Applies to your store? |
|---|---|---|---|---|---|
| Akamai 'State of Online Retail Performance' | 2017 | Page load time vs conversion | 10B+ mobile pageviews, US retailers | 100ms delay → 7% conversion drop (mobile peak) | Partially — desktop number is 1.1%, mobile peak is the cited 7% |
| Deloitte 'Milliseconds Make Millions' (w/ Google) | 2020 | Mobile site speed (LCP-adjacent) vs revenue per session | 37 retail, travel, lead-gen brands | 0.1s improvement → 8.4% retail conversion lift | Directionally — but measured at site-wide level, not per-page |
| Google CrUX / Web.dev research | 2019-2023 | LCP bucket vs bounce probability | Top 10M sites, all verticals | Sites in 'good' LCP (<2.5s) see 24% lower bounce | Yes for bounce; conversion link is inferred, not measured |
| Cloudflare retail data | 2022 | TTFB vs add-to-cart rate | Retail customers, anonymised | 100ms TTFB improvement → ~0.6% ATC lift | Yes — closest match to mid-market retail elasticity |
| Shopify internal (Built for Shopify benchmarks) | 2023 | Theme LCP vs checkout completion | Shopify Plus stores | 200ms LCP gain → 1.0% checkout completion lift | Yes — directly applicable to Shopify stores |
Three things jump out. First, the headline '1% per 100ms' is a rough average of mobile-skewed retail numbers — it's not in any single study verbatim. Second, the funnel stage matters enormously: the elasticity at checkout is roughly 2-3x what it is on a category page. Third, the closest-to-apples-to-apples sources (Cloudflare retail, Shopify internal) land in the 0.5%-1.0% range, not the often-quoted 7%.
Conversion lift per 100ms LCP improvement, by funnel stage
Which studies actually apply to your store
The Akamai 2017 number is the one most often misquoted. Their actual finding: a 100ms desktop delay correlated with a 1.1% conversion drop, but the eye-catching 7% came from a 2-second mobile delay on a specific subset of retailers. Cite the 1.1% figure if you must use Akamai — the 7% requires footnotes nobody reads.
Deloitte's 2020 study with Google is more defensible because it specifically isolated retail and measured revenue per session, not just conversion rate. Their 8.4% retail conversion lift from 0.1s improvement is real — but it was measured against site-wide median speed, not LCP per page. If you're modeling the impact of removing one script that improves LCP by 80ms on product pages only, dividing 8.4% by 10 is the wrong math.
The mobile-vs-desktop trap
Nearly every cited elasticity number is mobile-weighted, because mobile is where the slow tail lives. If your store does 65% of revenue on desktop (common for higher AOV verticals like furniture, electronics, B2B-adjacent), use the lower end of the range — desktop conversion is roughly half as speed-elastic as mobile in every study we've checked.
The honest elasticity to use in your model
For a €3M Shopify apparel store with a 70/30 mobile/desktop split, the defensible elasticity is 0.6% conversion lift per 100ms LCP improvement on product and checkout pages — blended across the funnel. That's the number we use internally when sizing the revenue impact of removing redundant tracking snippets or consolidating a heatmap tool. It's lower than the marketing slogan, but it survives a finance review.
If you want to be more rigorous: weight each funnel stage by its share of conversions, apply the stage-specific elasticity from the chart above, and only count LCP improvements on pages where the script actually loads. A heatmap tool firing only on PDPs doesn't get credit for a homepage LCP improvement. This stage-weighted approach is what the downstream heatmap-cost analysis uses to land on a credible consolidation threshold.
Frequently asked questions
It's a rough synthesis, not a single citation. The closest source is Akamai's 2017 retail report, which found a 1.1% desktop conversion drop per 100ms of page load delay. The number got rounded to '1%' and recycled, often misattributed to Amazon or Google.
Amazon executives have referenced page-speed impact in talks since 2006, but never published a per-100ms conversion elasticity. The often-cited '100ms = 1% of sales at Amazon' is a paraphrase of internal commentary, not a published study.
No. Mobile is roughly 2x more speed-elastic than desktop in every study we've reviewed. If your store skews desktop-heavy, halve the elasticity. If it's mobile-dominant (most apparel and beauty), the higher end of the range applies.
LCP has the clearest link to conversion. FID/INP matters mainly at checkout (form interaction). CLS impacts trust and tap accuracy but doesn't have a clean per-100ms-equivalent conversion number — most studies focus on LCP because it's the most measurable speed signal.
Intent. A user on a product listing page is browsing; a slow load just sends them back. A user mid-checkout has committed — speed friction there directly correlates with abandonment, and the user is much less likely to return.
Measure the LCP delta the script causes (use a before/after Lighthouse run or WebPageTest), weight by the funnel stages where it fires, apply the 0.6% per 100ms blended elasticity, and multiply by relevant revenue. For most third-party scripts on Shopify, the answer lands between €5K and €40K annually for a €3M store.
No — there's a diminishing return below ~2 seconds LCP. The studies that show the strongest elasticity are measuring movement from 'slow' to 'medium,' not 'fast' to 'very fast.' If your LCP is already 1.8s, expect closer to 0.2% per 100ms, not 0.6%.
Because it's the most dramatic number in the report. It's technically real but comes from a specific mobile retail subset with a 2-second delay — not a general 100ms-per-1% relationship. Treat it as the ceiling of plausible elasticity, not the expected value.
Each third-party snippet typically adds 30-150ms to LCP on the pages where it loads. Stack four of them on a product page and you're often looking at 200-400ms of avoidable LCP — which, at 0.6% per 100ms, is a 1.2-2.4% conversion penalty before you account for runtime cost on lower-end mobile devices.
0.6% conversion lift per 100ms of LCP improvement on funnel pages, blended mobile + desktop. Cite Cloudflare's retail TTFB data and Shopify's Built for Shopify benchmarks as the underlying sources — they're the closest match to mid-market retail and survive scrutiny better than Akamai's headline figures.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.