Google Optimize

Google Optimise was good at one thing above all others: letting you run a browser-side A/B test without touching your server code. If you already had Google Analytics installed, the setup was a snippet or a Tag Manager tag and you could start splitting traffic in minutes. I used it for a client who wanted to test a red button against a green one on their product page. The integration with Analytics meant the goal was just a click event I already tracked. The test reached significance after four weeks and showed a 12% lift in add-to-cart clicks. That result was clean and actionable.

For multivariate tests the interface was straightforward. You selected elements on the page, typed in variants and Optimise randomised combinations. I once ran a three-element test on a landing page – headline, CTA colour and hero image – and the tool handled the nine combinations without complaining. The report flagged the winning combination automatically.

The redirect test type was useful for testing completely different pages. If you had a new checkout flow you could send 50% of traffic to the new URL via a redirect test. I did that for an e-commerce site and the test ran for two weeks before I called the winner.

Optimise also offered basic personalization: you could serve different content based on segments like new vs returning visitors or device type. For a SaaS client I personalised the homepage headline for returning visitors. The session replay tool [Hotjar](/hotjar/) showed that returning visitors interacted more with the personalised headline. It wasn't sophisticated – no machine learning – but it was better than nothing and it was free.

For a free tool the reliability was decent. Google kept the servers running and the reports updated near real-time. If you needed a simple test fast and already lived in the Google ecosystem, Optimise was the path of least resistance.

The biggest awkwardness is that Optimise no longer exists. Google shut it down on 30 September 2023. If you are reading this in 2025, the tool is gone. That makes any advice about setup or best practices purely historical. I once linked an old tutorial that recommended Optimise for personalization, and a reader tried to implement it six months after the sunset. They wasted an afternoon before they realised.

Even when it was alive, Optimise had limitations that made it awkward for serious testing. It relied entirely on Google Analytics data for targeting and goals. If your Analytics implementation was messy – duplicate events, wrong attribution models, low session quality – your test results were unreliable. I helped a friend who had set up Optimise on a site where Analytics was tracking the same pageview from two different accounts. The test split was useless.

Performance was another issue. The Optimise snippet loaded synchronously by default and could delay page rendering. Google recommended async loading via Tag Manager, but that added complexity. I saw a client's page speed drop by 0.8 seconds on mobile when they activated Optimise. They abandoned the test because of the impact on [website speed](/website-speed-test/).

Sample size requirements were often underestimated. Optimise's built-in calculator told you how many visitors you needed, but many marketers ignored it. I ran a test on a low-traffic blog post hoping to see a lift in email signups. After three months the test was still 10% away from significance. I had to kill it. That is a waste of effort.

Another awkward point: the free version capped you at three simultaneous tests and 5,000 monthly unique visitors for personalization. If your site had more traffic you hit the limits fast. The paid Optimise 360 solved that but cost thousands per year and required a Google Analytics 360 contract. Most of my small clients could not justify that.

Finally, the interface felt dated. Editing variants in the visual editor was buggy. I once had a variant that broke because the editor wrapped my HTML in extra divs. I lost two hours debugging.

Because Optimise is dead, you need a new tool. Here is what I use now and why.

For free heatmaps, session recordings and basic A/B testing, my first pick is [Microsoft Clarity](/microsoft-clarity/). It is completely free, has no traffic limits, and the recording quality is excellent. You cannot run a controlled A/B test inside Clarity – it is an observation tool, not an experiment tool. But for understanding user behaviour and prioritising which page element to test, Clarity is unbeatable at the price.

For true A/B and multivariate testing on a budget, I point people to Hotjar. The free tier lets you run one active test and record up to 35 daily sessions. I use Hotjar for small client projects where the sample size is under 10,000 visitors a month. The visual editor is more stable than Optimise's ever was. Last month I set up a CTA colour test in 15 minutes and it ran without errors for six weeks.

If you need advanced testing features like server-side experiments, fractional factorial designs or full personalisation engines, you have to pay. VWO and Optimizely are the industry standards. I have used VWO on a few larger sites and the segmentation abilities beat Optimise by a mile. The downside is cost – expect $500+ per month. For serious conversion rate optimisation programmes that price is justified.

For analytics without testing, I switched to [Fathom Analytics](/fathom-analytics/) for one of my own sites. It is lightweight, privacy-friendly and costs a flat monthly fee. No testing features, but if you only need reliable pageview and event data, Fathom is simpler than Google Analytics.

If you rely on Google data, [Bing Webmaster Tools](/bing-webmaster-tools/) gives you free search performance data and can act as a secondary source, though it is not a replacement for Optimise.

My decision rule: if your total monthly visitors are under 50,000 and you need testing, use Hotjar's paid plan. If you are on a strict zero budget, combine Microsoft Clarity for insight with manual server-side tests using a feature flag tool like LaunchDarkly. That gives you control without monthly fees.

I would not recommend any tool that claims to be a direct Optimise clone with free forever pricing. Those usually have hidden limits or poor support. Pay for what you actually need.