How Do I Test Changes to Get More Calls and Leads From My Website?

By Bill Stilwell · Last reviewed October 18, 2026

The short answer

To improve how many visitors call, book, or fill out a form, test one change at a time. Write down what you expect to happen, pick one number to judge it by, run it for a set period, and record the result and your decision. Follow Google’s testing rules so the test doesn’t confuse search: show Google the same pages people see, and remove the test when it’s done.

Why it matters

Most small business websites get changed on a hunch. Someone moves the phone number, rewrites the headline, and adds a coupon all in one week. Calls go up or down, and nobody knows why. A dog groomer who tests one booking button change at a time learns what works and keeps it. One who changes everything at once learns nothing.

21.10.1 Hypothesis

What it means: a plain statement of what you think will improve results and why.

How to do it: write it as “If we change X, then Y will go up, because Z.” For example: “If we put the phone number at the top of the service page on phones, calls will go up, because visitors won’t have to scroll to find it.”

Common mistake: testing random ideas with no reason behind them. A reason tells you what to try next if it fails.

21.10.2 Change

What it means: one meaningful change.

How to do it: change one thing, such as the headline, the main button, the form length, or the photo at the top. Make it big enough to matter. Changing a button from dark blue to navy rarely moves anything for a small site.

Common mistake: changing several things at once, so you can’t tell which one made the difference.

21.10.3 Measurement

What it means: the one number that decides whether the change worked.

How to do it: pick a key event you already track, such as calls from the page, form submissions, or bookings (see 13.7.3 Form Tracking and 13.7.1 Analytics Setup and Measurement Plan). Decide before you start.

Common mistake: picking the number that looks best after the test ends.

21.10.4 Test period

What it means: how long the test runs.

How to do it: run it for whole weeks so weekdays and weekends are both included. Low-traffic sites need longer. If your page gets only a handful of leads a month, a test may need a couple of months, or you may be better off making a change and comparing month to month. Google’s guidance says to run a test only as long as needed and then update the site and remove the test parts.

Common mistake: stopping after three days because the early numbers look good. Early results swing a lot.

21.10.5 Segment

What it means: comparing the groups that matter.

How to do it: check results for phone and computer visitors separately, and for new versus returning visitors if you can. A change that helps phone visitors may hurt computer visitors.

Common mistake: looking only at the total and missing that one group got worse.

21.10.6 Result

What it means: what actually happened.

How to do it: record the before and after numbers, the number of visitors in each version, and anything unusual during the test, like a holiday or a storm that drove calls.

Common mistake: keeping results in your head. In six months you won’t remember.

21.10.7 Decision

What it means: keep, revise, or remove the change.

How to do it: keep it if it clearly helped. Revise it if results are mixed. Remove it if it hurt or did nothing. Then clean up: delete test pages, scripts, and redirects so only the winning version stays live.

Common mistake: leaving old test versions running for months. Google says long tests that serve a different version to many users can look like an attempt to deceive search engines.

21.11.1 Experiment ID

What it means: a unique label for each test.

How to do it: use a simple pattern like EXP-2026-01. Put it in your experiment log and in the name of the test in your testing tool.

Common mistake: naming tests “homepage test” over and over until you can’t tell them apart.

21.11.2 Objective

What it means: what the experiment is meant to learn.

How to do it: copy your hypothesis into the log in one or two sentences.

Common mistake: an objective so vague it can’t fail, like “improve the site.”

21.11.3 Baseline

What it means: current performance before the change.

How to do it: record the last 4 to 8 weeks of your chosen number for that page.

Common mistake: skipping the baseline, so there’s nothing to compare to.

21.11.4 Change

What it means: exactly what changed.

How to do it: describe it in words and add a screenshot of the old and new versions.

Common mistake: writing “new headline” without saving what the headline said.

21.11.5 Audience

What it means: who saw the change.

How to do it: note the page, the share of visitors who saw each version, and any limits, such as phones only.

Common mistake: forgetting that ad traffic was pointed at the page during the test, which changes who saw it.

21.11.6 Dates

What it means: start and end dates.

How to do it: record both, plus any days the test was paused or broken.

Common mistake: no end date, so the test never formally ends.

21.11.7 Result

What it means: the data and what you think it means.

How to do it: record the numbers from 21.10.6 and one or two sentences on why you think it happened.

Common mistake: recording the numbers with no interpretation, so the lesson gets lost.

21.11.8 Decision

What it means: what happens next.

How to do it: record keep, revise, or remove, the date the cleanup was done, and the next test idea.

Common mistake: deciding but never doing the cleanup.

When to get help from a pro

A web developer or conversion specialist can set up tests that follow Google’s rules. Questions to ask:

Will search engines see the same versions my visitors see?
Do the test versions use a canonical tag pointing to the original page, and temporary redirects if any?
How will you remove the test code and pages when we’re done?

Bill’s note

[Bill: add a short note. For example, a small change on a client’s service page that made a real difference in calls, or a test that surprised you by doing nothing.] If you’d rather have this done for you, SiteSmarty builds and manages websites for local businesses.

How I did this for SiteSmarty

[Date I did this:]
[What I actually did, step by step: the first test I ran on the SiteSmarty site and how I set it up]
[What it cost and how long it took:]
[What surprised me or what I’d do differently:]
[Screenshot or photo to add, with private info blurred: the experiment log row and before and after screenshots of the page]
[Result so far:]

Questions people ask

Can A/B testing hurt my Google rankings? Not if you follow Google’s guidance: don’t show Google a different page than visitors see, use a canonical tag on test versions, use temporary redirects, and end the test when you have an answer.

How long should a website test run? At least a few full weeks, longer for low-traffic pages. Stop when you have a clear answer, then clean up.

What should I test first? The weakest step in your funnel (see 21.7 Funnel Review). On many local service pages, that’s how easy it is to call or book from a phone.

What if my site gets very little traffic? Make one change, then compare a full month before and after. It’s less exact, but still better than guessing.

Audit check

Outline line	Question	Pass looks like	Fail looks like	Evidence to collect
21.10.1 Hypothesis	Does each test start with a written reason?	If, then, because statement	No reason given	Experiment log
21.10.2 Change	Is only one meaningful change tested?	One change	Several changes at once	Before and after screenshots
21.10.3 Measurement	Was the success number chosen before the test?	One key event set in advance	Picked after results	Log entry with date
21.10.4 Test period	Did the test run full weeks and end on time?	Whole weeks, planned end	Stopped early or never ended	Test dates
21.10.5 Segment	Were phone and computer results checked?	Both reviewed	Totals only	Analytics report
21.10.6 Result	Are before and after numbers recorded?	Numbers and notes	Memory only	Log entry
21.10.7 Decision	Was a decision made and test code removed?	Decision and cleanup date	Old tests still live	Log and live page check
21.11.1 Experiment ID	Does each test have a unique ID?	Unique IDs	Duplicate or missing names	Log
21.11.2 Objective	Is the objective specific?	Testable objective	Vague goal	Log
21.11.3 Baseline	Was a baseline recorded?	4 to 8 weeks of data	None	Log
21.11.4 Change	Is the exact change saved?	Words and screenshots	“New headline” only	Log and images
21.11.5 Audience	Is the audience described?	Page, split, and limits	Not recorded	Log
21.11.6 Dates	Are start and end dates recorded?	Both dates	Missing end date	Log
21.11.7 Result	Are data and interpretation recorded?	Numbers plus explanation	Numbers only	Log
21.11.8 Decision	Is the next step recorded?	Keep, revise, or remove with cleanup date	No decision	Log

Where this fits

Do this first: 13.7.1 Analytics Setup and Measurement Plan, 21.7 Funnel Review, 21.9 Customer Lifetime Value

Do this next: 22.1 Writing Standard Operating Procedures

Sources

Google Search Central, "Minimize A/B testing impact in Google Search" (checked October 18, 2026)
Google Business Profile Help, "Tips to improve your local ranking on Google" (checked October 18, 2026)

This page explains general rules and Bill's own experience. It is not legal, tax, or insurance advice. Talk to a qualified professional about your situation.