Google Ads Experiments: Test Your Way to a Lower Cost Per Lead
When a contractor says their Google Ads are underperforming, the first question worth asking is: what was the last change you tested? The answer is almost always the same. They changed something because it felt right, watched for a couple of weeks, decided it did not help, and changed it back. No baseline, no control group, no comparison. Just instinct and wasted budget.
Google’s Experiment Center, which consolidated campaign experiments and lift studies into a single interface in January 2026, eliminates the guesswork. It splits your campaign traffic automatically between a control version and a test version, tracks both simultaneously, and tells you with statistical confidence which version generates more leads at a lower cost. You do not have to interpret charts or guess whether a change helped. The system runs the comparison and gives you a clear answer.
For contractors spending $2,000 to $15,000 per month on Google Ads, a single successful experiment that reduces cost per lead by 15 to 20 percent is worth thousands of dollars per year in recovered budget. Here is how to run one correctly.
How the Experiment Center Works
The Experiment Center is in Google Ads under the Campaigns menu. When you create an experiment, Google clones your existing campaign into a control version and a treatment version. You make exactly one change to the treatment, then set a traffic split. The standard split is 50/50: half your campaign traffic goes to the control, half to the treatment. Both versions run simultaneously, exposed to the same search volume, the same competitor auctions, and the same demand patterns.
After the experiment runs for the set period, Google reports which version generated more conversions, at what cost per conversion, and whether the difference is statistically significant. A statistically significant result means the gap is large enough that it is very unlikely to be random noise. Significant wins tell you the change is real and you should apply it permanently. Inconclusive results tell you the variable you tested does not matter much and you should move on to something else.
What to Test First: Ad Headlines
Ad copy has the fastest path from change to measurable result. A headline swap shows meaningful CTR and conversion differences within 10 to 14 days in any market with normal search volume. Three types of headline changes produce the most consistent results for contractor accounts.
Urgency versus availability signals. Test a headline like "Same-Day Repair Available" against "Serving [City] and All Surrounding Areas." Emergency repair queries convert better on urgency. Maintenance and replacement queries convert better on geographic reach. Running both tells you which framing your specific market responds to rather than guessing.
Credential leads versus outcome leads. Test "Licensed HVAC Technicians in Phoenix" against "AC Fixed Right or We Come Back Free." Credential-first headlines work better in high-competition trades where homeowners have been burned before. Outcome-first headlines work better in markets where the primary concern is cost and contractors are less differentiated. You cannot know which applies to your market without a test.
Offer headlines versus no-offer headlines. Test "Free Diagnostic With Any Repair" against your current primary headline. Free offer headlines typically increase CTR by 10 to 25 percent on top-of-funnel searches but can reduce close rate if callers are fishing for free estimates with no intent to hire. Run this test only if you have call tracking set up to measure booked jobs, not just call count. A headline that doubles your calls and halves your close rate is not a win.
What to Test Second: Bidding Strategy
After confirming your ad copy is solid, the next highest-impact experiment for most contractor accounts is bidding strategy. The two most common comparisons worth running are listed below.
| Test | Control version | Treatment version | Best for |
|---|---|---|---|
| Bidding model | Maximize Conversions | Target CPA | Accounts with 25+ monthly conversions |
| CPA target level | Current Target CPA | Target CPA reduced 15 to 20% | Accounts already on Target CPA |
| Bidding model | Target CPA | Target ROAS | Accounts tracking job revenue by service |
For accounts generating fewer than 20 conversions per month, Target CPA typically starves the campaign of the data it needs to optimize. Maximize Conversions performs better at lower volume. For accounts above 25 monthly conversions, Target CPA usually reduces cost per lead while maintaining call count. The experiment tells you which applies to your account instead of making you guess.
If you drop the Target CPA by 15 percent in the treatment version and the campaign cannot hit the lower target, conversion volume will fall significantly in the first two weeks. That is a clear signal to end the experiment early and keep your current target. If call volume holds steady at the lower CPA, apply the new target permanently and recover that budget for other use.
How Long to Run the Test
The most costly mistake contractors make with Google Ads experiments is ending them in week one because the treatment looks worse, or in week two because it looks better. Both are wrong for the same reason: 10 days of data does not represent a full demand cycle.
Run every experiment for a minimum of four weeks, and ideally six weeks, regardless of what the early results show. Google Ads performance cycles within any given week (higher emergency repair searches on weekdays, higher planning and maintenance searches on weekends) and within any month (weather events, seasonal demand, competitor promotions). A test that ends after 10 days caught a random slice, not a representative one.
Google’s Experiment Center shows a confidence level for each metric. Wait until your primary conversion metric shows 90 percent or higher confidence before acting on the result. A result at 75 percent confidence means there is a 1 in 4 chance the difference is noise. At 90 percent, the chance drops to 1 in 10. For a change that affects every dollar of your ad spend for the next 6 to 12 months, 90 percent is the floor.
The Testing Sequence That Works
Run experiments in sequence, not in parallel. If you test headlines and bidding strategy simultaneously, you cannot know which change drove the result. Each test should inform the next: confirm the best headline first, then test bidding with the confirmed headlines in place, then test landing pages with the confirmed bidding strategy running.
- Message test (weeks 1 to 5): Test one headline change across all responsive search ads in the campaign. Primary metric: conversion rate. Secondary metric: CTR. Apply the winning version before moving to the next test.
- Bidding test (weeks 6 to 11): Test bidding strategy or CPA target change. Primary metric: cost per conversion. Secondary metric: total conversion count. Apply the winning version before the next test.
- Landing page test (weeks 12 to 16): Test a dedicated service landing page against your current page. Primary metric: call conversion rate. This test often produces the largest gains because most contractor campaign traffic still goes to generic homepages with no clear call to action above the fold.
Three tests run back to back takes roughly four months. A contractor who runs this sequence consistently finds two or three changes that compound over time. A 15 percent reduction in cost per lead from the headline test, combined with a 12 percent reduction from the bidding test, compounds to a 25 percent total reduction. That is not a rounding error on a $5,000 monthly budget. It is $1,250 per month recovered from the same spend.
Three Actions to Take This Week
- Open the Experiment Center in Google Ads and set up your first headline test today. Navigate to Campaigns, then Experiments. Select your primary lead generation campaign, click Create Experiment, set the traffic split to 50/50, and swap one headline in the treatment version. Set the end date four weeks out. Setup takes under 20 minutes. Google handles the traffic split automatically, and you can monitor results without touching anything until the experiment ends.
- Verify your conversion tracking before starting any test. An experiment comparing two campaign versions is only as reliable as the conversion data underneath it. If phone calls are your primary lead source but you are only tracking form fills, your experiment data does not reflect actual business results. Set up call tracking through Google Ads or a tool like CallRail before running the test. Make sure your call conversion threshold matches actual intent: a 60-second minimum call duration filters out accidental dials and disconnects. Confirm that call conversions and form fill conversions are both recording before the experiment starts.
- Write down your hypothesis before the test starts. Document in three sentences what you are testing, what you expect to happen, and why. This forces you to define success before results start arriving and prevents you from reinterpreting the goal mid-test when early data goes against your expectation. A clear hypothesis also makes it easier to apply the learning to other campaigns and other markets after the test concludes.
Google Ads experiments cost nothing to run beyond the normal campaign spend. The traffic split does not inflate your budget. The only cost is running both versions simultaneously for a few weeks instead of making a permanent change based on a guess. Contractors who test systematically pay less per lead than those who do not, because every confirmed win compounds into the next test. The businesses holding the lowest cost per lead in your market are not lucky. They have run the experiments.