Back to blog

Google Ads4 min read

How to run experiments in Google Ads

Google Ads experiments are for big changes: a different landing page or a different way of optimising. Most of ours test two landing pages. Here’s when to run one and how.

A Google Ads experiment: the original and the experiment split 50/50 for 60 or more days, sync off, then apply or abandon.

The short answer

A Google Ads experiment takes one of your campaigns and runs a second version of it alongside the original, with the traffic split between the two. The second version carries one big, meaningful change. Both run in the same auctions over the same period, so you can see exactly what that change did, then apply it or abandon it.

Use it for big changes only: a different landing page or a different way of optimising. Most of ours test two landing pages, because the page is the main lever on conversion rate.

What an experiment is for

Swap a landing page and leads go up the next month, but you don’t know if it was the page, the season, a competitor pausing or luck. An experiment takes the guesswork out: the only difference between the two campaigns is the thing you changed.

  1. Optimisation setup

    Max conversions, target CPA, a different conversion action.

  2. Landing page

    Two pages against each other. The main lever on conversion rate.

    Most of ours

An example: test Book a demo against Take the tour

The original is a one-field booker: Book a demo is the only thing the page asks for. The experiment is the same page with Take the tour as the primary call to action, a completely different conversion action. Same campaign, same keywords, same ads, a 50/50 split. Whichever wins, you know why.

Original · 50%
Book a demo is the only ask.
A landing page hero with a Book a Demo button in the navigation and a one-field Book a demo form, and no other call to action.
Experiment · 50%
Take the tour: a different conversion action.
The same hero with Take the tour as the primary button and Book a demo as a plain link beside it.

Test big changes, not small ones

Don’t test for the sake of it. An experiment is for a change big enough to be worth proving before it goes everywhere: a new landing page, or a different way of optimising, such as maximise conversions against target CPA, a different conversion action, or maximise conversion value.

A different ad isn’t big enough. You can have three ads live in an ad group, which is enough to test three different ads against each other without running an experiment. Google’s asset reports and a headline report already show which headlines and descriptions win, with nothing split.

Worth an experiment

  • A different landing page
  • Max conversions vs target CPA
  • A different conversion action
  • Maximise conversion value

Not worth one

  • A different adLaunch several ads and compare them
  • Headlines and descriptionsGoogle’s asset reports and a headline report

Every split gives Google less data

Every experiment is a new campaign, and every split halves the data each campaign learns from. The more volume behind the metric you optimise towards, the better the algorithm performs, on Google and on every other platform.

Only pay that price for a change worth proving. That’s why you don’t run many tests, and why you pick them carefully.

One campaign12 conversions

What Google learns from

Original6 conversions

What Google learns from

Experiment6 conversions

What Google learns from

How to create a Google experiment

In the client’s Google Ads account we go to Experiments, create a custom experiment for Search and pick the original campaign. Add a suffix like ‘- [Test Name]’ to the end of the name, copy the new campaign ID, and update the UTM tracking in the campaign settings with the new name and ID.

  1. 1 · Name it

    Add a suffix like “- [Test Name]” to the end

  2. 2 · Copy the ID

    The experiment gets a new campaign ID

  3. 3 · Update the UTMs

    New campaign name and ID in the settings

OriginalNon-Brand - High Intent - Marketing Services

ExperimentNon-Brand - High Intent - Marketing Services - Test Page

How to set up a landing page test

We swap the final URL on the experiment’s ads: ad group by ad group if each has its own page, all at once if not. Editing ads can strip their ad labels, so reapply them if needed and check the final URLs.

Original

Final URL/b2b-marketing-services

Experiment

Final URL/marketing-services-for-b2b

How to set up a conversion objective test

If you want to test a different conversion setup, go into the experiment campaign and change the settings: the conversion objective, and any of the goals you need to adjust. For example, you may be optimising to a booked demo, an SQL and an opportunity at the same time, so you’d change the objective and also the primary conversions for that campaign.

Original

Bid strategyMaximise conversions
Primary conversionBook a demo

Experiment

Bid strategyMaximise conversion value
Conversion values
  • Demo 1
  • SQL 10
  • Opportunity 100

The schedule settings we always use

Every experiment gets the same settings unless the client asks for something different.

  • A 50% split, so both versions get an equal share of the traffic.
  • 60+ day duration, so you can end it when you want, and it won’t end too early.
  • Sync turned off, so changes to the original don’t copy across.
  • Always update UTMs so you know in your CRM what the test drove.

Set the end date long, then end it early

Every experiment needs a start date and an end date, and once it’s running you can end it early but you can’t extend it. So set the end date two to three months out and end it the moment the data is clear. Set it short and let it lapse, and you’re building the whole experiment again.

StartSet both dates now
End it earlyOnce the data is clear
End date: 2 to 3 months outCan’t be moved later

Checks before it goes live

A setup mistake wastes the test. Always make sure to check:

  • Every ad group has the correct landing page.
  • All ad labels are back on.
  • Every UTM is updated and correct.
  • Keyword match types are right.
  • Bidding strategies are right.
  • The correct broad match negative keyword lists are applied.
Key takeaway

The takeaway

Experiments are for big changes: a different landing page or a different way of optimising. Split the traffic 50/50, track it properly, set the end date long and end it early, then apply or abandon on the numbers.

In this article
  1. The short answer
  2. What an experiment is for
  3. An example: test Book a demo against Take the tour
  4. Test big changes, not small ones
  5. Every split gives Google less data
  6. How to create a Google experiment
  7. How to set up a landing page test
  8. How to set up a conversion objective test
  9. The schedule settings we always use
  10. Set the end date long, then end it early
  11. Checks before it goes live
  12. The takeaway
Share this post
  • Drive More Meetings
  • Increase Pipeline
  • Strategies That Work

We run LinkedIn and Google Ads for B2B brands, and we’re good at it.

Book a CallGet the Library

More articles

Back to blog

Landing Pages

How to 2x your CVR on Google with a tour

Most Google landing pages only have a Book a demo button. Add a product tour as a second way to convert and the page can double its conversion rate.

Joseph Hill5 min read