Optimization types

Your website needs to drive results. But how do you know your site is actually doing that? And how do you make it better?

With Webflow Optimize, you can serve different versions of your site, watch how real visitors respond, and let their behavior show you what works.

In this video, we're going to cover the optimization types: testing and personalization. Both can be driven by you manually or by AI automatically. We'll walk through all of that, and also cover when to use which, so you can make an informed decision.

One quick note: You can build all your optimizations, but you need tracking on to start collecting data. Just be sure you're compliant with how you're tracking visitor data first. But once you do, in the Insights tab, go to Tracking. Then switch on tracking. And you'll want to do this for every site you plan to optimize.

Okay, with that, let's start with testing.

Testing lets you compare different versions of page elements to see what performs best, like a different button, different heading copy, or maybe a different image. You put two or more versions out into the world, and let real visitor behavior pick the winner.

Let's say we want to create a test to see which button is more likely to get clicked on.

Here we are in Webflow and let's go up to the context bar and create a new test optimization to compare two versions of our button. We'll give it a name, then we choose the type: a traditional test, where you set the traffic split and wait for a winner, or an AI-optimized test, where AI splits the traffic for you. We'll start with traditional.

Now let's build our variations. Up in the context bar, you can see which variation you're working on. There's the base, that's our original version, what's currently on our page, and then our new variation, which we'll rename to something more descriptive. Right on the canvas, we select the button, edit the content in this variation, and type in something more action-oriented. Okay, that's our first variation. Then we'll add a second variation the same way: name it and edit the content. At any time, we can switch between the variations in the context bar to preview them.

With our variations built, let's set our traffic and our goal. Under settings and goals, we choose how traffic is allocated. We can choose an even split, or a manual split that totals a hundred percent. And we set our goal here, the action we're measuring, which in our case is a click on the button. That's how Optimize knows which version is actually working.

Then we can go back and launch our test. And to take our variations live, we publish the site. Now our test is live. Depending on how much traffic your site receives, getting meaningful results will take time. But once you do, you can go back to your optimizations and see how your test performed. Once it's reached statistical significance, meaning the data's reliable enough to trust, you can end the test and act on what it found: either set the winning variation as your new base if it won, or try a new idea if it didn't.

Now here's where AI comes in. Instead of managing that traffic split yourself, you create an AI-optimized test and Optimize continuously tests your variations, automatically directing more people to the highest-performing ones. You don't need to set a split. It adjusts in real-time based on how visitors behave. And there's no declared winner here. Rather than ending when one version wins, an AI-optimized test keeps running, always sending each visitor to whatever's working best for them.

And at any time, you can add new ideas and remove existing ones to ensure you’re aligned with your evolving business objectives.

So that's testing. Let's talk about personalization.

With personalization, you can specify which audiences should see which variation, so you can tailor the experience for the people who meet your criteria.

Let's say we want to show returning visitors a hero section that's ready to convert them. But new visitors still get the base variation.

Same place: Insights. All optimizations. And we'll create a new optimization, but this time we'll add a new personalization, and give it a name.

Next, we need to define our audiences, and we can get really granular here. We can create a new audience based on any rule set: their device, their location, where they came from. For our case, we'll create an audience for returning visitors, people who've been to our site before. All other visitors will just see the base variation.

Now we just need to build the variation this audience will see. Right on the canvas, we'll update the hero: change the heading since these folks already know us and don't need the introduction. And we'll swap any other element inviting them straight to the product.

And then, same as testing, we set our goal, QA, and launch. Again, getting some data will take time, but instead of one winner for everyone, it shows how the variation performs with its assigned audience, how that specific group responds.

And just like with testing, AI can drive this too. With an AI-optimized personalization, you still build the variations, but you let AI decide which audiences see what, adapting over time based on how well each one converts.

So that's personalization. But how do you choose between testing and personalization?

Create a test for when you want to find one best version for everyone. You're comparing a few variations and letting the traffic decide the winner.

And create a personalization for when you want to deliberately give different visitors different experiences. So you're tailoring the page to who they are, not necessarily looking for a single winner.

And you can layer AI on top on either one. But when is using AI the right call?

For testing, AI is great when you want to understand how your ideas perform against the base, but you don't need to lock in a single winner. AI constantly adjusts to how visitors behave, so what's shown is always what's working best.

And for personalization, AI can automatically find the segments your ideas actually resonate with. You might assume a variation will land with a certain audience, but AI confirms whether it really does, based on how visitors respond. Basically, AI handles the balancing so you don't have to.

So that's testing and personalization, but you can also use the AI Assistant to suggest what you want to test and personalize.

With AI-powered suggestions, you can use the AI Assistant to recommend copy to test or personalize, and what variations to create, so you're not starting from a blank slate. And once your optimization is running, the AI Assistant can also help you make sense of the results.

So those are the ways to optimize your site with Webflow Optimize.

Remember: optimization isn't a one-time fix. It's an ongoing process. You try an idea, learn from the results, apply what worked, and repeat. So keep revisiting your strategy and testing new variations. And ideally, your conversion rate gets a little better each time.