We test landing pages across every Google Ads account we manage. It's probably the single most reliable way to improve campaign performance without spending more. And yet, the most common mistake we see is judging a good page too quickly. Most businesses don't realise that Google's bidding algorithm needs time to respond to a new landing experience, and the data you see in the first few days can be genuinely misleading.
This guide covers how to set up, run and evaluate a landing page test in Google Ads. We'll also get into the silent adjustment period that most testing guides skip entirely, which is honestly the part that trips people up the most.
Before you start
A few things need to be in place before a landing page test is worth running. Skip these and you'll end up with results you can't trust, which is arguably worse than not testing at all.
Your conversion tracking has to be working reliably. That sounds obvious, but we've inherited accounts where "conversion tracking" was a tag that fires on every page load, not an actual verified conversion action. If you can't trust the data going in, the test result coming out is meaningless.
You also need enough traffic. A campaign generating five conversions a month won't produce a clear signal for weeks, possibly months. As a rough minimum, you want at least 30 to 50 conversions per month across the campaign you're testing.
And you need a reason for the change. "The page looks outdated" is a reason. "We want to see if a shorter form converts better" is a better one. The clearest tests start with a specific question, not a vague redesign.
Step 1: Start with a hypothesis, not just a new design
"Let's try a new landing page" sounds like a test, but it isn't really one. A proper test starts with a specific change and a reason you think it'll improve performance.
Good hypotheses tend to be narrow. Something like: "Reducing the form from seven fields to three will increase the conversion rate because we're asking for information we don't actually need at the enquiry stage." Or: "Moving the phone number above the fold will increase calls because mobile users aren't scrolling down to the contact section."
If you're testing a completely new page against the original, that's fine too. Just recognise you're testing everything at once. If performance improves, you'll know the new page is better overall. You won't know which specific element made the difference. That's a trade-off. For iterative improvement over time, isolating one variable per test gives you much clearer learning.
Step 2: Set up a landing page test in Google Ads
Google Ads has a built-in experiments feature that's designed for exactly this. It splits traffic at campaign level, so both pages see the same audience, keywords and ad copy. That level of control is what makes the test actually meaningful, rather than just comparing two pages that happened to run at different times or under different conditions.
To set it up:
- Go to Experiments in your Google Ads account.
- Create a custom experiment based on the campaign you want to test.
- In the experiment arm, change only the landing page URL. Keep everything else identical: keywords, bids, ad copy, extensions.
- Set your traffic split. 50/50 gets you to a clear result fastest. If the change feels higher risk (a completely different page structure or offer), 70/30 limits your exposure while still generating enough test data.
- Set the experiment to run for at least three to four weeks. You can always extend it. Ending early usually just means ending before the data tells you anything useful.
One thing worth flagging: if you're running Smart Bidding, both arms of the experiment will have their own bidding models. This matters more than you might think, and it's where the next step comes in.
Step 3: Give the algorithm time to adjust
This is the part most testing guides leave out, and honestly it's probably the most important thing in this entire article.
When you change a landing page on a Smart Bidding campaign (Max Conversions, Target CPA, Target ROAS), Google doesn't just serve the new page and carry on as normal. It needs to collect fresh data on how users interact with the new experience. Conversion rates shift. User behaviour signals change. The whole relationship between search queries and outcomes moves.
Google's bidding models start recalibrating. But they do it silently. Unlike a change to your bidding strategy or budget, a landing page change doesn't trigger the visible "Learning" status in your campaign. Google refers to this as a silent learning phase. The algorithm is adjusting, but there's nothing in the UI telling you that.
What this means in practice is that the bidding algorithm keeps operating on old data. If your previous page converted at 10% and your new page actually converts at 15%, Google doesn't know that yet. It's still bidding as if the conversion rate is 10%, so you won't see the full benefit of a better page straight away.
The reverse is true too, and this is where it gets painful. If the new page converts at 5% instead of 10%, Google may keep bidding aggressively because it still expects the higher rate. Your cost per acquisition spikes until the system catches up, and it's easy to panic and kill the test before it's had a fair chance.
How long does this adjustment take? Across the accounts we manage, it varies quite a bit:
- Moderate traffic accounts typically stabilise within 3 to 7 days.
- Lower-volume lead generation campaigns can take 1 to 2 weeks.
- High-volume accounts (ecommerce, high-traffic services) often adjust within a day or two.
Google's own documentation suggests roughly 50 conversion events or three conversion cycles for a bid strategy to fully calibrate to a new signal. For most service businesses running lead generation, that translates to at least a week of data. Sometimes two.
Performance Max campaigns tend to take longer, which makes sense. Google is optimising across more variables at once: audience signals, creative combinations and landing page content all feed into the same model. There's just more for it to figure out.
Manual CPC campaigns have no algorithmic adjustment period, so the effect on conversion rate is immediate. But you still need enough conversions to separate a genuine improvement from normal variance.
Step 4: Read the results properly
The temptation is to check the experiment every day and draw conclusions as soon as one page pulls ahead. We get it. But resist that urge.
Here's what we typically see:
Days 1 to 3: Conversion rates bounce around. Smart Bidding is still operating on old assumptions. Anything you see in this window is noise, not signal. It's tempting to read into it, but try not to.
Days 4 to 10: Things start to stabilise a bit. Bidding begins adapting to the new conversion patterns. You might see some directional signals forming, but it's still too early to commit to anything.
Days 10 to 21: With sufficient conversion volume, you can start to assess the true impact of the change.
The key metric here is conversions, not calendar days. For lead generation campaigns, we generally evaluate after the test has accumulated at least 50 to 100 conversions across both arms, where volume allows. At lower volumes, you may need to extend the test to three or even four weeks.
And don't look at conversion rate in isolation. Check these as well:
- Cost per conversion. A higher conversion rate doesn't help much if the cost per lead also went up. Smart Bidding can shift auction behaviour in ways that affect your CPC.
- Conversion quality. Are the leads from the new page as qualified as the ones from the original? If you track downstream outcomes (proposals sent, jobs won), check those too.
- Volume. Did the new page generate more or fewer total conversions at the same spend level? This one gets overlooked surprisingly often.
It's also worth knowing that industry data suggests only around 1 in 8 landing page A/B tests produces a statistically significant winner. That number might feel discouraging, but it's normal. Most tests show small differences that need more data, or they confirm that both pages perform roughly the same. A "no difference" result is still useful. It tells you that variable isn't the lever to pull, and you should test something else.
Step 5: Call it and roll out the winner
Once you have enough data and a clear winner, end the experiment and apply the winning page to your main campaign. Pretty straightforward.
But here's one thing that catches people out. If you launch a completely new page and immediately see a 10 to 20% drop in performance, don't panic and conclude the page is worse. If it's materially different from the original, give it at least 50 to 100 conversions (where volume supports it), or one to two weeks of stable traffic, before making the call. What you might be seeing is the algorithm's adjustment period, not the page's actual performance.
After rolling out the winner, keep an eye on things for another week. Performance in an experiment doesn't always replicate perfectly once 100% of traffic hits the new page. We've seen small discrepancies before. A week of post-rollout monitoring catches any issues before they become costly.
When to get help
Landing page testing is fairly straightforward when two things are true: your conversion tracking is accurate and your campaign generates enough conversions to produce a signal within a reasonable timeframe.
When either of those is missing, it gets harder. Unreliable tracking makes every result suspect. Very low conversion volume (under 15 to 20 per month) means tests can take months to reach a conclusion, and by then your market or your offer may have shifted.
Performance Max adds another layer of complexity. Google's optimising across more variables at once, which makes it harder to isolate whether a performance shift came from the landing page or from something else the algorithm decided to change. If you're running PMax for lead generation, the test results need more careful interpretation than a standard search campaign.
If you're running Google Ads and want to know whether your landing pages are leaving performance on the table, book a free audit call. We'll look at your setup, your tracking and your data, and show you where the opportunities are. No obligation, no pitch.