Roofers: build for the storm night, not the average Tuesday

Roofing demand is weather-shaped. Nothing happens for weeks, then a storm passes through and every search for emergency roof repair in your area happens inside the same evening, largely on phones, largely from people who have just found a wet patch on a ceiling.

That is a load test you did not schedule. It is also the only traffic that really matters commercially, which makes it worth engineering for specifically rather than hoping the site copes.

Two things decide whether it does: how much of your site can be served from cache without running any code, and how many megabytes each of those visitors has to download. Both are measurable this afternoon.

Anything with a per-visitor element tends to opt itself out. A quote calculator, a live availability widget, a form that generates a fresh token on render: any of these can mark the whole document uncacheable, and then every storm visitor runs the full PHP path.

The fix is to isolate the dynamic part rather than uncache the page around it. Load the calculator over a separate request after the page has painted, keep the surrounding content static, and the page stays cacheable. If a plugin refuses to work that way, weigh what it earns against what it costs you on the one night that matters.

First, uptime. The 99.9% commitment is backed by a pro-rated credit, and monitoring runs whether you are watching or not, but the practical value is knowing the site was up during the window you cared about rather than assuming it.

Second, field vitals. After a spike, check PageSpeed Insights field data for your emergency and area pages. Because that data is a rolling 28-day sample at the 75th percentile, a bad storm night shows up in it, and a good one confirms the work held. If the numbers stayed green under load, the caching did its job.