A CRO guide for local business websites

Run conversion work as a loop: instrument the site, find the largest drop-off, change one thing, measure it against a pre-change baseline across a full cycle, then keep it or revert it. For seasonal trades, measuring across a full cycle is what stops seasonality from being credited as improvement.

Step 1: Instrument before you have opinions

Form field analytics, call tracking by page, and page-level exit data. Without these you are optimizing by taste, and taste is why most redesigns produce no measurable change.

Step 2: Find the largest drop-off

Look for the biggest absolute loss, not the worst percentage. A ten-point improvement on a page with four hundred monthly visitors is worth less than a two-point improvement on the page with four thousand.

Step 3: Change the offer before the ornament

Clarity about what you do, where you do it, and roughly what it costs moves more than any micro-optimization. Button color is the last five percent, and only after the first ninety-five is right.

Step 4: Measure across a full cycle

Seasonal trades will happily credit July for a change made in June. A full cycle, or a matched period year over year, is the only honest comparison.

Step 5: Revert honestly and log it

Anything that does not move the number gets reverted. Keeping changes because someone advocated for them is how sites accumulate clutter that nobody can justify.

What to test first, by trade

HVAC: replacement cost ranges and financing visibility on the install pages, and a shortened emergency form during season.

The traps

Changing several things at once, ending a test early because the first week looked good, and optimizing a page that receives almost no traffic are the three most common wasted efforts.

When CRO is the wrong project

If nobody is finding the site, conversion work raises a rate applied to almost nothing. Fix visibility first.

QUICK ANSWERS
How long should a test run?

Long enough to cover a full demand cycle for your trade, and never stopped early because the first week looked favorable.

How many changes at once?

One, unless the changes are on entirely separate templates with separate traffic. Otherwise attribution is guesswork.

What if the data is ambiguous?

Treat that as a result. Revert, keep the log entry, and move to the next largest leak rather than arguing about the interpretation.

NEXT STEP

Run the loop with someone who has run it before.

Book a CRO Review