Last reviewed November 2026.
Short answer
You can A/B test a landing page with little traffic, but only a big change, and the answer will be a hint, not proof. Test one large change, set the end date first, count calls and forms, and write down what else changed. If the page gets only a few dozen visits a month, fix the obvious problems and talk to customers first.
A/B testing means showing two versions of a page and seeing which one gets more calls or forms. It needs a lot of visitors to give a clear answer, and a small local page rarely has them. This guide shows what to do about that, with a plan you can follow this week.
This guide is part of our landing page guide for small business.
In this guide
- What A/B testing is and why traffic matters
- A made-up example with the math
- What counts as low traffic
- Decide: test, compare or fix
- What to test first
- How to run a simple test, step by step
- Ways to learn when you cannot test
- Test ad text, where the volume is
- Keep search in mind
- Reading results honestly
- A one-page test log
- Common mistakes
- Frequently asked questions
- Next steps
What A/B testing is and why traffic matters
Two terms first. A/B test: you split visitors between version A (the page you have) and version B (one change). Conversion: the action you want, such as a call, a form or a booking. Your conversion rate is conversions divided by visits.
Results bounce around by chance, like flipping a coin ten times and getting seven heads. More flips give a steadier picture. More visitors do the same for a page. When you have few visitors, a gap between A and B may be luck. See what makes a landing page convert for the parts worth changing.
How many visitors do you need? It depends on three things: your current rate, the smallest gain you care about, and how sure you want to be. Evan Miller’s sample size calculator asks for those inputs. For its sample numbers (10.2% going to 13.2%) it returns 2,545 visitors per version. Instapage gives another example: a 5% rate, a 10% relative lift and 95% confidence need about 31,000 visitors per version.
Those numbers are far above what a local service page sees in a month. That is the whole problem, and the rest of this guide works around it.
A made-up example with the math
This example is made up. The numbers are round and are not from a real client. The sample sizes come from our own use of the standard formula, at 95% confidence and 80% power. Check them in a calculator with your own numbers.
Picture a paving or plumbing page with 200 visits a month and a 5% call-and-form rate. That is 10 conversions a month. You make a new headline and hope it helps. Here is how many visitors each version needs, depending on the gain you hope to detect.
| Gain you hope to detect | New rate | Visitors per version | Both versions | Months at 200 visits |
|---|---|---|---|---|
| 5% better (small) | 5.25% | about 122,000 | about 244,000 | over 1,200 |
| 20% better | 6% | about 8,200 | about 16,400 | about 82 |
| 50% better | 7.5% | about 1,470 | about 2,940 | about 15 |
| 100% better (double) | 10% | about 435 | about 870 | about 4.4 |
Read the table from the bottom. Only a huge change, one that doubles the rate, can be seen in about four months, and the page would need to run that whole time. A small gain cannot be seen at all. This matches what Optimizely describes when it compares a 5% gain with a 50% gain: the bigger the gain, the less data you need. Conversion Rate Experts says a 100% gain can be detected about four times as fast as a 50% gain.
The lesson: with 200 visits a month, a button color test is not worth running. The guide below says not to run an A/B test at all under about 300 visits. Learn from calls and customer notes instead.
What counts as low traffic
Portent describes low traffic as “a few hundred visits a day or less.” By that measure, most local service pages are low traffic, because they get far fewer than a few hundred visits a day.
The number that matters most is conversions, not visits. A page with 1,000 visits and 10 calls has little to compare. Count your calls and forms for the last 90 days. Write the number down. It tells you which row of the next table you are in.
Decide: test, compare or fix
Use this table to pick your starting point. Visits are per month to one page.
| Your situation | Best move | Why | Watch out for |
|---|---|---|---|
| Under about 300 visits and under 10 conversions | Fix and learn. Do not run an A/B test. | Chance will swamp any result. | Do not read much into week-to-week swings. |
| 300 to 1,500 visits | Before-and-after test of one big change. An A/B test only works for a near-doubling and takes months | You can only see a large gain. | Seasons and ad changes can mislead you. |
| 1,500 to 5,000 visits | A/B test one large change with a set end date | A big change can show up in a few months. | Do not peek and stop early. |
| More than that, or you can add ad spend | A/B test a few changes, one at a time | You have room to learn. | Keep tracking set up the same for both versions. |
| The page is broken, slow or off message | Fix it, then decide | Some problems are not questions. | Testing a broken page only tells you it is broken. |
The bands are a rough guide from our own reading of the sample-size math above, not a rule. Check your own numbers with a calculator. If the calculator says you need more visitors than you can get in four to six months, switch to the ideas below.
What to test first
Rank ideas by how big the change is and how easy it is to make. Start at the top. Portent gives a simple test for size: if you cannot squint and notice the thing you are testing, the change probably will not work on a low-traffic site.
| Idea | Size of change | Effort | Good for low traffic? |
|---|---|---|---|
| New headline that matches the ad | Large | Low | Yes. See headline formulas |
| A different offer, such as a free quote vs a booking | Large | Medium | Yes |
| Call-first page vs form-first page | Large | Medium | Yes. See call or form |
| Shorter form | Medium to large | Low | Yes. See how many fields |
| Real reviews near the top | Medium to large | Low | Yes. See social proof without fake reviews |
| Your own photos instead of stock | Medium | Medium | Maybe |
| Button text | Small to medium | Low | Rarely |
| Button color | Small | Low | No |
Test one variant against the original. Portent recommends the opposite of testing many variations on a low-traffic site, because each added version raises the time you need.
How to run a simple test, step by step
Track calls and forms before you begin. If you cannot count them, you have nothing to compare. See what to track in Google Ads and call tracking.

- Write one question. For example: “Will a headline that names the service and the town get more calls?”
- Pick one measure. Calls plus form fills, counted the same way for both versions.
- Check the page first. Fix anything broken. Use the speed and mobile check and the landing page mistakes list.
- Make one big change. Change the headline, or the offer, or the form. Not all three.
- Split traffic evenly. Most testing tools can do this. If yours cannot, run version A, then version B, for the same number of days.
- Set the end before you start. Pick a visitor count or a date. Evan Miller advises you to decide on a sample size in advance and wait until the experiment is over.
- Do not touch anything else. Keep ad text, budget and the phone line steady during the test.
- Record the result. Write the counts, the dates and anything else that changed.
- Choose. Keep the better version, or repeat the test if the gap is small.
If you are writing the new version, our landing page brief template keeps the message clear. If you want a team to build and test pages for you, see our landing page design service.
Ways to learn when you cannot test
When the math says you cannot test, you can still learn. These methods need people, not traffic.
1. Fix the obvious
A broken form, a slow page, a phone number you cannot tap and a headline that does not match the ad are problems to fix, not to test. Fix them today.
2. Talk to your leads
Ask each new customer two questions: “What made you call?” and “What almost stopped you?” Write down the answers. Ten short answers can point to the next change better than a month of weak data. Portent notes that you only need five to eight participants in user research to find patterns. Note that this is about research, not A/B testing.
3. Watch real people
Ask a friend or past customer who has not seen the page to use it on their phone while you watch. Ask them to find the price, the phone number and the next step. Where they pause is where the page loses people. Conversion Rate Experts lists user testing as its first choice for low traffic.
4. Compare before and after, with care
Run the old page for a set time. Then run the new page for the same length. It is weaker than a true A/B test, because the season, weather, ads and holidays can change. Keep a note of anything else that changed.
5. Test where you have more volume
Your ad gets more impressions than your page gets visits. That is the next section.
Test ad text, where the volume is
If your traffic comes from Google Ads, the ad has more views than the page. You can test two headlines in the ad faster than two headlines on the page. See how to write Google Ads that get clicks. The message in the ad and the headline on the page should match, so a winning ad often points to a page change as well.
Set a budget you can lose without harm. Our guide on how much to spend on Google Ads helps with that, and the ad budget reality check shows what a month of ads must earn to pay off.
Keep search in mind
If your landing page can also show up in Google Search, follow Google’s rules for tests. From Google Search Central:
- Do not show one set of URLs to Googlebot and a different set to people. That is cloaking, and it breaks Google’s rules whether or not you are testing.
- If you test with separate URLs, Google recommends rel=canonical on the alternate URLs rather than a noindex tag.
- If you redirect people to a test version, use a 302 (temporary) redirect, not a 301.
- Remove the test as soon as it ends. Google says it may see a test that runs too long as an attempt to deceive search engines.
These rules apply to any page Google can crawl, so follow them even if the page is mainly used for ads.
Want a landing page built for your ads? We design and build them for local service businesses. See our landing page design.
Reading results honestly
- Do not stop early because one version is ahead. Early leads can shrink or flip.
- Count small numbers as small. If A got 4 calls and B got 6, that is not a win. It is a coin flip.
- Check the context. Did ads, season or the phone line change during the test?
- Treat a lower confidence level as a lean. Optimizely says lowering your significance raises the chance of false positives. Conversion Rate Experts suggests 90 or 85 percent can be acceptable if the alternative is no testing. Use that room with care.
- Repeat a surprise. If a result looks too good, run it again.
- Count real jobs, not only calls. A version that brings more calls but fewer booked jobs has not won.
A one-page test log
Copy this into a notebook or spreadsheet. Fill it in before you start. It keeps you honest.

| Field | What to write |
|---|---|
| Question | One sentence. |
| Change | What is different in version B, in one line. |
| Measure | Calls plus forms, and how you count them. |
| Start date and end rule | A date or visitor count, set now. |
| Traffic source | Ads, search, or both. Keep it the same. |
| Other changes during the test | Ads, season, staff, phone line. |
| Result | Counts for A and for B. |
| Decision | Keep A, keep B, or repeat. |
| Next idea | What the result suggests. |
A test log also helps when you work with an agency. You can see what was tried and why.
Common mistakes
| Mistake | Why it hurts | Fix |
|---|---|---|
| Testing button colors on a tiny page | The change is too small to see in a few visits. | Test the headline, offer or form. |
| Changing three things at once | You cannot tell which one mattered. | Change one main thing. |
| Calling a winner after a week | Early leads can shrink or flip. | Set the end date first. |
| No call tracking | You have nothing to compare. | Set up tracking before the test. |
| Running many versions | Each one adds time. | Use one variant against the original. |
| Testing a broken page | You learn that it is broken. | Fix it first. |
| Changing the ads mid-test | Different visitors arrive. | Hold ads steady. |
| Leaving the test running | Google says to remove tests when done. | Remove test code at the end. |
| Ignoring customers | They tell you why they called. | Ask each new lead two questions. |
Central PA angle
A shop in Carlisle, Mechanicsburg or Harrisburg serves a small market. A page for one town may see a few dozen visits in a month. That is normal. It means your best tools are the slow ones: customer notes, phone calls and watching real people.
Seasons matter too. A snow plow page and a lawn page see different traffic in January and July. If you compare before and after, compare the same time of year. Our landing page guide for small business covers how the whole page fits together.
Frequently asked questions
Can you A/B test a landing page with low traffic?
Yes, but only for big changes, and the answer will be weak. With a few dozen visits a month, a small gain is hidden by chance. Test one large change against the old page, set an end date, and treat the result as a hint. Use calls and customer notes as a backup.
How many visitors do you need for an A/B test?
At a 5% conversion rate, you need about 435 visitors per version to see a doubling, and far more for smaller gains (see the table in this guide). The exact number depends on your current rate and the gain you want to detect. Instapage shows an example that needs about 31,000 visitors per version, and Evan Miller’s sample size calculator returns 2,545 per version for its sample inputs. Run your own numbers in a calculator before you start.
How long should a landing page A/B test run?
Run it until you reach the visitor count you chose in advance, and not a day sooner. Evan Miller advises you to decide on a sample size in advance and wait until the experiment is over. If that count would take months, the test is too big for your traffic.
What should a small business test first on a landing page?
Test the headline, the offer or the form, because these are big changes. A new headline that matches the ad is a good first test. Skip button colors and small wording tweaks. With low traffic, you can only see a gain if the change is large enough to notice at a glance.
Is a before-and-after test as good as a real A/B test?
No. A before-and-after test is weaker, because the season, the ads or the weather can change between the two periods. It is still better than guessing if you write down everything else that changed and keep both periods the same length.
What can I do instead of A/B testing?
Fix the obvious problems, talk to your customers and watch how real people use the page. A broken form, a slow page or a headline that does not match the ad needs a fix, not a test. Portent says five to eight user research participants can show patterns that improve the experience.
Does A/B testing hurt my SEO?
It does not have to. Google says not to show one set of URLs to Googlebot and a different set to people. If you test with separate URLs, use rel=canonical, use a 302 redirect and remove the test when it ends. These rules apply to any page Google can crawl.
Should I use a 90 percent confidence level instead of 95?
It is a trade-off, not a free pass. Optimizely says lowering your statistical significance raises the chance of false positives. Conversion Rate Experts suggests 90 or 85 percent may be fine when the alternative is no testing. If you do it, treat the result as a lean, not a fact.
Next steps
Here is the short plan. Count your calls and forms. Fix what is broken. Ask your last ten customers what made them call. Then pick one big change, set the end date and test it. If the numbers are too small, you still end up with a better page.
If you want a second pair of eyes, we will look at your page with you. There is no pressure to hire us. You can also start with a free tool. Our website teardown shows what a visitor sees on your site.
To see how we build pages for local businesses, read about our landing page design.
Related guides
- landing page guide for small business
- Landing Page for a Local Service Business: Call or Form?
- How to Brief a Landing Page: A Fill-In Template
- How Much Should a Local Business Spend on Google Ads?
- landing page design
See how we build landing pages for local businesses in Carlisle and Central PA.



