Last week I loaded 100 contractor websites on a phone connection and timed every one — HVAC, roofing, plumbing, solar, restoration. Not opinions about design. A stopwatch and a scale.

Thirty of the 100 crossed a three-second estimate. The worst fully weighed page was 19.2 megabytes. The sample describes the pages observed in August 2026; it does not establish market prevalence, lost calls or ad spend.

What I found

HOW HEAVY THEIR HOMEPAGES WERE 31 sites — under 1 MB 17 sites — 1 to 3 MB 15 sites — 3 to 5 MB 6 sites — over 5 MB (worst: 19.2 MB) 69 sites weighed in full · median 2.0 MB · a good page is under 1 MB

Lower transfer weight usually gives a page more room to perform well on constrained connections. Weight alone does not determine load time or visitor behaviour; server response, code, caching, device and network all matter.

The part that costs money

Say you spend $500 a month on Google Ads and it brings 100 people to your site. Nothing about your ads changes in this math — only the page they land on.

SAME $500 IN ADS · SAME 100 CLICKS OBSERVED PERIOD A calls = clicks × reach rate × call rate OBSERVED PERIOD B cost per call = $500 ÷ observed calls compare equivalent traffic · do not assign causality from two totals alone

Record real values for both periods. A difference in calls is evidence to investigate, not proof that speed caused it; campaign mix, seasonality, offer and tracking can move at the same time.

Check yours in 30 seconds

Go to pagespeed.web.dev, paste your website address, pick Mobile. Free, no signup. Two numbers matter: how long until the page is usable, and total page weight.

  • Under 2 seconds — a strong starting point; still verify real-user data and conversions
  • 2 to 4 seconds — inspect the slowest component and the affected traffic segments
  • Over 4 seconds — investigate; the number alone does not prove lost calls or revenue

What's actually slowing it down

Oversized photos are a common cause. One page in the sample contained images at 2.4, 2.7 and 2.3 MB. The correct delivery size depends on dimensions, format and visual tolerance; test the result instead of enforcing one universal kilobyte limit.

You don't need a new website to fix most of this. You need somebody to resize your photos and turn off the plugins you stopped using in 2021.

The other two culprits: a cheap slow server — 24 of the 100 sites took over a second just to answer, before a single image started loading — and a pile of plugins each dragging in its own scripts.

What to actually do about it

In order of how much time it buys you per hour of work:

  1. Resize every photo on the homepage

    Resize each image to its largest rendered size and test a suitable modern format. Browser tools such as Squoosh can do this locally; the saving depends on the source asset.

  2. Delete plugins you don't use

    Inventory active plugins and verify what each one loads. Remove only what is unused and test the site afterward.

  3. Check what your host charges you for

    If repeated measurements show slow server response, image work alone will not address that component. Inspect hosting, caching and server-side work; price does not prove quality.

Website speed is one part of the path from click to contact. Measure it beside message clarity, traffic quality and real conversion events.

Answers to the questions I get asked

How fast should my site load?

Under 3 seconds on a phone. A properly built contractor site does it in about 1 second and weighs under 1 MB.

Do I need a whole new website?

Often no. Compress the photos, drop dead plugins, move to decent hosting. If the site is 10 years old and won't work on a phone, then yes — start over.

Does speed affect Google rankings?

Yes, and it now affects whether AI tools mention you at all. A page nothing can load fast is a page nothing recommends.

One more observation: 90 of the 100 sites had no chat widget. That fact alone does not prove missed leads; phone, form and booking behaviour must be measured separately.