# Lab data vs field data: why your speed score and your customers disagree Source: https://storefrontspeedjournal.com/blog/2026-10-20-lab-vs-field-data-shopify-speed.html Reviewed October 20, 2026 PageSpeed gives you two different stories: a controlled lab test and real visitor field data. Why they diverge, which one search rankings use, and how to work with both without chasing ghosts. ## Two instruments, two different truths Lab data comes from a controlled test: one device profile, one network speed, one location, run on demand. It is repeatable and diagnostic, which makes it the right tool for finding what to fix. Field data comes from actual visitors over the past weeks: their devices, networks, countries, and habits, aggregated into percentiles. It is messy and slow moving, and it is the data that describes what your customers genuinely experienced. Neither is the real speed in any simple sense. The lab is a simulation with the conditions chosen by the tool, often a mid range phone on a throttled connection that may be harsher or kinder than your actual audience. The field is reality with the labels worn off: you cannot rerun it after a fix and see the answer tomorrow, because it trails behind you by weeks. ## Why the two numbers diverge on Shopify stores Audience geography is the biggest wedge. A store whose customers are mostly domestic on modern phones and fast networks will post field numbers far better than the lab's default test suggests. A store selling internationally to mixed devices can see the reverse. Product mix matters too: field data aggregates every page type visitors actually opened, including the heavy app laden landing pages your lab tests never visit. Timing creates the rest of the gap. Lab tests measure the page as it is this minute, including the sale banner you added yesterday. Field data is a trailing window, so a fix you shipped two weeks ago is still diluting itself against the old slow pages in the average. Merchants routinely panic over field numbers that are simply the past, still draining out of the window. ## Which one rankings and decisions should use Search ranking systems use field data, the real visitor aggregates, assessed at the seventy fifth percentile: the experience of the slower quarter's boundary, not the average. That choice is deliberate; it measures what a substantial share of real people endured. So for visibility stakes, field data is the scoreboard, and lab data is the training ground where you improve what the scoreboard will eventually say. For decisions, use each where it is strong. Reproduce and diagnose in the lab, where you can test one change at a time and watch the waterfall. Judge outcomes in the field, over full windows, knowing the verdict lags your changes by weeks. The expensive mistake is mixing the roles: shipping changes until a lab number turns green while the field, and the customers, never move. ## A working rhythm for a real store Pick the handful of templates that carry your traffic, product, collection, and home, and record their field numbers monthly from the same report, seventy fifth percentile, mobile separated from desktop. Between readings, use lab tests to evaluate specific changes on those templates before and after. That pairing gives you a honest before and after for the change and a slow truthful trend for the business. Expect the field trend to move like a cargo ship. A genuine improvement shipped today finishes entering the data about a month from now. Write the ship date next to every speed change in your log, so when the field number moves you can attribute it, and when it does not, you know which change to interrogate. Speed work without that log is just weather watching. ## Related reading If this topic is useful, these earlier guides go deeper on connected questions: [Shopify font loading, layout shift, and the invisible tax](https://storefrontspeedjournal.com/blog/2026-10-19-shopify-font-loading-layout-shift.md), [Removing Shopify apps cleanly: the speed audit after uninstall](https://storefrontspeedjournal.com/blog/2026-10-19-removing-shopify-apps-cleanly-speed-audit.md), and [Lazy loading images below the fold on Shopify, without breaking the hero](https://storefrontspeedjournal.com/blog/2026-10-18-shopify-lazy-loading-below-fold-images.md). ## Frequently asked questions ### Why is my field LCP worse than my lab LCP? Usually audience and window effects: your real visitors include slower devices, weaker networks, and countries far from your test location, and the field window still contains weeks of older, slower pages. Check the field data by page template and device class before assuming a regression; the gap often describes your audience, not a new problem. ### Which number does Google rank with? Rankings use field data, the aggregated real visitor experience, evaluated at the seventy fifth percentile over a rolling window. Lab scores do not directly feed ranking. That is why a store can show a poor lab score and healthy rankings, or the reverse, and why field trends are the numbers to protect. ### How long after a fix does field data improve? Because the field window trails by roughly a month, expect improvement to phase in over several weeks and fully register about a month after the change reaches all visitors. Judge a fix at the four to six week mark in the field, and use lab tests for the immediate verdict on whether the change itself worked. ### Should I test on desktop or mobile settings in the lab? Mobile, for most stores: most traffic is mobile, mobile devices are the constrained ones, and the field data that matters most is usually the mobile set. Run desktop as a sanity check. If mobile lab results are acceptable, desktop almost always is too.