Find where your website loses people before you redesign it

Start with the page they land on, not the homepage

Open GA4, go to Reports, then Engagement, then Landing page, and sort by sessions. On most local service sites the homepage is not the top entry page. A single service page or one old blog post is doing half the work, and that is the page worth your attention.

Then look at two columns next to it: engagement rate and key events. A page pulling three hundred sessions a month and producing nothing is a specific, fixable problem. A redesign spreads your effort across thirty pages to fix that one.

The three failures behind most of it

First, the phone number is not tappable. We still find sites where the number sits inside a header image or as plain text, so a customer on a phone has to memorize it or copy it by hand. Some of them do not bother.

Second, the form asks for more than it has earned. Name, phone, and a sentence about the problem is enough to start. Street address, preferred contact window, and how you heard about us belong in the follow-up call, not on the page where someone is deciding whether to contact you at all.

Third, the page never says where you work. A visitor in Freehold who cannot tell from the page whether you cover Freehold will close it rather than call and ask.

Test the site on a phone you did not build it on

Hand your phone to somebody who has never seen the site and watch them try to reach you. Tap the number. Fill out the form. Then check whether the submission actually arrived.

Form email is the quietest failure mode there is. We have taken over sites where the contact form had been routing to an employee who left two years earlier, and nobody knew because the page still said thank you. Send yourself a test submission on the first of every month. It takes a minute and it has caught this more than once.

Speed matters, but not the way audit tools say

A performance score of 62 is not why you lost the job. What loses the job is a four megabyte hero photo loading over cell service in a driveway, and that is a different problem from the one an audit report puts at the top of the list.

Check the size of the largest images on your busiest pages and the number of scripts and fonts the site loads. Fix those two and ignore the score. Chasing a number into the nineties on a small brochure site is work that never shows up in your call volume.

What to do when there is not enough traffic to analyze

Under a few hundred sessions a month, analytics is mostly noise. Ten sessions on a page tells you nothing, and heatmap tools sold on the promise of insight will happily show you patterns that are not there.

Call five customers who hired you in the last quarter and ask two questions: what did you look at before you called, and was there a moment you almost called someone else instead. Owners are usually surprised by the answers. That conversation is better data than any tool you can buy at that traffic level.

How fast you answer is part of the website

The site's job ends at the submission. The sale is usually decided in the hour after it, and no design change compensates for a lead sitting unread until Thursday.

Time your own response for a month. Note when each form fill and each missed call came in and when somebody actually got back to the person. Owners who do this are routinely surprised, because the delays cluster around the hours nobody owns: the mid-afternoon stretch when everyone is on a job, and everything that arrives after Friday at four.

The fix is rarely technical. Forward form submissions to a phone as a text rather than an email, or give one person the standing job of clearing the inbox twice a day. In competitive trades, the first company to call back wins a disproportionate share of the work, and that has nothing to do with your website and everything to do with whether the website was worth building.

When a rebuild is actually the right call

Sometimes it is. If you cannot add a page or change a headline without paying a developer, if the site does not adapt to a phone screen, or if the platform it sits on is no longer supported, patching is throwing money at something that needs replacing.

Short of that, do the diagnosis first. In most of the audits we run, two or three specific changes account for nearly all of the recoverable loss, and a full rebuild is worth doing for the reasons above rather than as a way to guess at a conversion problem you have not identified yet.