The model
The estimate is one line of arithmetic, chosen to be conservative at every step:
loss = annual revenue × 50% new-visitor share × min(4% × seconds of LCP beyond 2.5s, 50%)
- 4% of conversions per second. Published measurements cluster between 4% and 8% per second (see sources). We use the bottom of the range.
- Only past Google's threshold. Seconds under a 2.5s LCP cost you nothing in this model. The clock starts where Google's "good" bucket ends.
- Only new visitors. Returning customers who already trust you wait longer, so we apply the loss to a 50% new-visitor share. Most stores acquire a larger share of new visitors than that.
- Capped at 50%. However slow the site, the model never claims more than half of new-visitor revenue, even though bounce studies would support more at extreme load times.
The conversion-rate field does not change the loss estimate; it re-expresses the same model as a rate. If speed suppresses half your visitors' conversions by the loss fraction, your observed rate understates your potential rate: potential = current ÷ (1 − 0.5 × loss fraction).
Where the numbers come from
| Source | Finding |
|---|---|
| Google / Deloitte, "Milliseconds Make Millions" (2020) | A 0.1s speed improvement raised retail conversions by 8.4% |
| Akamai, "The State of Online Retail Performance" (2017) | Every 100ms of added delay cost 7% in conversions |
| Google / SOASTA mobile study (2017) | Bounce probability rises 32% as load goes 1s→3s, 90% by 5s, 123% by 10s |
| Portent conversion study (2019, updated 2022) | Conversion rates drop roughly 4.4% per second of load in the 0–5s range |
| Our study of 269 Shopify stores (2026) | Median mobile LCP was 10.0s; 88% of stores in Google's Poor bucket |
What this model does NOT claim
- It does not claim any specific brand loses a specific amount. When we publish estimates next to a brand's measured speed, the dollar figure always describes a hypothetical store at a stated revenue, never that brand's actual finances.
- The underlying studies are large-sample and directionally consistent, but they are correlational. Your vertical, price point, and brand strength move the true number.
- Speed is not your only conversion lever. It is, however, the one that silently taxes every other one you optimize.
Frequently asked questions
How much revenue does a slow website lose?
A conservative estimate: about 4% of conversions per second of LCP beyond Google's 2.5 second threshold, applied to your new-visitor share. A store doing $100k/month with an 8 second LCP loses roughly $130k/year under this model.
Does site speed really affect sales?
Yes, and it is measured rather than theoretical: Google and Deloitte measured +8.4% retail conversions from a 0.1s improvement, Akamai measured a 7% conversion drop per 100ms of delay, and Google's mobile data shows bounce probability rising 32% between 1s and 3s of load time.
What is a good LCP for an online store?
Google's "good" threshold for Largest Contentful Paint is 2.5 seconds on mobile. In our study of 269 Shopify stores the median was 10 seconds, and 88% of stores were in the Poor bucket, so clearing 2.5s puts you ahead of nearly everyone.
Fix it
Run your URL through the free audit to get your real LCP and a CLAUDE.md fix plan your developer (or Claude Code) can act on. Shopify store? The app runs the same audit weekly inside your admin. The most common causes and fixes are in the audit library, starting with LCP itself.