Quick answer: Google Optimize is Google’s free tool for A/B and multivariate tests. You link it to a Google Analytics property, add a small snippet to your site, create variants in a visual editor (or with code), and let Optimize decide which version wins against a goal you already track in Analytics. Start with one high-traffic page and one clear change so you can trust the result.
Guessing which headline converts better wastes time. A controlled test shows you. Paid testing platforms have existed for years; Optimize matters because it is free for a solid set of features and it sits next to the Analytics account you likely already use. Google made the free Optimize product widely available in 2017.
This walkthrough covers a first A/B test from linking to readout. For the wider craft of experiments, see our guide to A/B testing.
What you need before you start
- A Google Analytics account with a web property on Universal Analytics
- A goal or metric worth improving (form submits, clicks to checkout, time on page is usually a weak primary goal)
- Enough traffic that a test can finish in a reasonable time
- Ability to add JavaScript to your site (theme file, tag manager, or plugin)
- Permission to change the page you will test
If goals are not set up yet, configure them first. Our article on Google Analytics goals explains the common lead and sales patterns.
Step 1: Create an Optimize account and container
- Go to Google Optimize and sign in with the Google account that can access your Analytics property.
- Create an account (your company) and a container (usually one per website).
- Note the Optimize container ID (it looks like GTM-XXXXXX but for Optimize).
Keep naming clear: “Main website” beats “Container 1”.
Step 2: Link Optimize to Google Analytics
Inside the Optimize container, link the Analytics property that collects data for the site. This connection lets Optimize use Analytics goals and lets reports appear in both products.
Also enable the Optimize integration in Analytics under Administration where the linking UI asks for it. Linking AdWords or other products is separate; for a first test you only need Analytics.
Step 3: Install the Optimize snippet
Optimize gives you a snippet to place on every page you might test. Many teams install it sitewide through Google Tag Manager, which keeps tags in one place. Others paste it next to the Analytics tracking code.
Two practical tips:
- Load Optimize as early as you reasonably can so visitors do not see a flash of the original page before the variant appears (flicker). Optimize documents anti-flicker snippets for this.
- Test the install with Optimize’s diagnostic tools and by previewing an experiment before you start it.
If you use gtag.js (the global site tag introduced in 2017) or analytics.js, follow the install instructions that match your tag. Do not mix conflicting Analytics snippets on the same page.
Step 4: Pick a sensible first experiment
Good first tests:
- Homepage headline or primary button label
- Landing page hero image versus a simpler version
- Form length (fewer fields versus more)
- Pricing page order of plans
Poor first tests:
- Tiny colour tweaks on low-traffic pages
- Five changes at once (you will not know what worked)
- Pages that change daily for campaigns
If you need ideas rooted in user behaviour, run a simple CRO audit or watch session recordings before you invent variants.
Step 5: Build the experiment in Optimize
- Click Create Experiment and choose A/B test.
- Enter the page URL you will test.
- Create Variant 1 and edit it in the visual editor (text, images, hide/show) or with custom CSS/JavaScript for more control.
- Set targeting: usually “URL equals” that page. You can also target by device or Analytics audiences once you are comfortable.
- Choose an objective: an Analytics goal, or a pageview of a thank-you URL, or built-in metrics. Prefer the business outcome (lead, purchase) over vanity metrics.
- Set traffic split. A 50/50 split is fine for two variants including original.
Name variants clearly (“Control”, “Shorter headline”) so reporting is readable later.
Step 6: Preview, then start
Use Preview to see each variant as a visitor would. Check mobile and desktop. Confirm that links, forms and tag tracking still work on the variant.
When you start the experiment, leave it alone. Stopping early because one day looks good is a classic mistake.
Step 7: Read results without fooling yourself
Optimize will show which variant is leading and whether it has enough confidence to declare a leader. Keep these rules in mind:
- Let the test run through full business cycles (weekdays and weekends) when your sales pattern varies by day
- Do not peek and relaunch repeatedly; that biases outcomes
- Segment carefully after the fact (mobile vs desktop) without turning every slice into a new “winner”
- When a variant wins, implement it in the real template or CMS, then retire the experiment so you are not running Optimize edits forever
Pair Optimize data with your normal Analytics behaviour flow so you understand why something won, not only that it won.
Document every test
Keep a simple log of each experiment: the page, the hypothesis, the change, start and end dates, traffic, the result and what you shipped. After a few months that log becomes a record of what your audience responds to, and it stops the team from rerunning old ideas by accident.
Limits of the free tier (practical view)
Free Optimize covers A/B and modest multivariate testing with Analytics integration. Very large programmes sometimes outgrow it and move to paid testing suites. For most small and mid-size sites, free Optimize is enough to learn and to ship meaningful wins.
Also remember that Optimize changes the page in the browser. Extreme personalisation and heavy DOM rewrites can affect speed. Keep variants lean.
Privacy and consent notes
Testing tools set cookies and alter content. If you serve EU visitors, plan how Optimize fits your cookie and privacy notices before GDPR applies in May 2018. Do not treat “it is a Google product” as automatic compliance; your use case and disclosures still matter.
FAQ
Do I need Google Tag Manager?
No, but it makes installs and future tags easier. Direct snippet install works.
How long should a test run?
Until Optimize reports a clear result with adequate sessions, and at least one full week for many B2B sites. Low-traffic sites may need longer or bigger changes.
Can I test WordPress pages?
Yes. Install the snippet theme-wide or via GTM. Avoid fighting a page builder’s dynamic markup with fragile visual edits; prefer clear text and layout changes.
What if traffic is too low?
Test bigger changes on your highest-traffic template, improve traffic quality first, or use qualitative methods (short onsite surveys, customer interviews, usability sessions) until volume supports experiments.
Run one clean test
Google Optimize removes the price barrier to proper A/B testing. Link Analytics, install the snippet carefully, change one meaningful thing on a busy page, and wait for a trustworthy result. When you are ready to turn winning ideas into a steady programme, see how we help teams turn visitors into customers with research-backed CRO, not random tweaks.
