Squarespace Killed Its Blank Template — and Built Something Better From the Wreckage

Picture the moment a product team has been waiting for. Squarespace — the platform where 3 million customers build their websites — ships a blank template. No pre-designed layout, no guardrails, just a white screen and total creative freedom. The early dashboards light up: more users than ever are entering the CMS. Then the team looks one metric downstream, and the celebration stops.

"This was actually a bit of a disaster," says Lina Blackman, Director of Product Analytics at Squarespace. "It was not a good user experience. We saw early signals that we were increasing the number of users coming into our CMS — but they didn't end up converting."

On The Experimentation Edge, Blackman walked host Ashley Stirrup through what happened next — and why one of Squarespace's worst-performing launches became one of its most valuable. Her account is a case study in what separates mature experimentation programs from teams that simply run tests: the discipline to treat a loss as data, not as a verdict.

🎧 Listen to the full episode →

Three Million Customers, Two Kinds of Users

The blank template wasn't an unreasonable bet. Plenty of users say they want full control, and the most technical ones mean it. But when Blackman's team dug into why entry rates rose while conversion fell, they found something more useful than a verdict on one feature. They found a fault line running through the entire customer base.

"We realized that we have two very distinct user types," Blackman explains. "One that is more technical and wants the customization, and then a whole lot of users who actually need a guided way to design their templates." The second group — small business owners with no time to dig into HTML — weren't failing to use the blank template. The blank template was failing them.

The segmentation that emerged, learners versus builders, did more than explain one bad result. It reshaped how teams thought about the entire onboarding flow. The blank design was killed forever. The thinking it forced became Blueprint, the AI-guided builder now live on squarespace.com.

"Sometimes if you ask a user what they want," Blackman says, "what they tell you they want is not actually what they want."

That single line is the quiet engine of the whole story. Stated preference said: give us freedom. Behavior said: give us guidance. Only a controlled experiment could refute that disagreement — and only a team prepared to interrogate a loss could hear the answer.

The Two Questions That Rescue a Losing Test

Squarespace runs 150 to 200 experiments a year, with analysts embedded in every product team and a centralized program — shared test repo, briefs, decision matrices — holding it together. At that volume, most tests will not win. What matters is what happens next.

"I heavily believe that failed tests are just as good," Blackman says. When no variant wins, her analysts ask two questions. First: are there granular segments where this experience actually worked? That's the analytics deep dive — the kind that surfaced learners and builders. Second: should we continue investing in this idea at all?

The second question is the one most teams skip, because it's not a statistics question. It's a portfolio question. "There's a million things that a product team can be building," Blackman notes. "Experimentation's just a good way to focus and revisit all of the user problems that you could be solving and figuring out which one's most impactful."

Every closed test, win or lose, also feeds a knowledge library: this worked for our users, these kinds of experiences don't. Not the final say on future decisions, Blackman is careful to add, but compounding context that makes the next hypothesis sharper than the last.

One or Two Big Wins a Quarter Is the Healthy Number

There's an emotional dimension here that experimentation leaders will recognize. Teams root for their variants. Even Blackman's analysts — whom she affectionately calls the "spies" embedded in product teams — get attached. The antidote isn't detachment; it's calibrated expectations.

"Teams only need one or two big wins a quarter," she says. "It's just not sustainable to have a million hits."

That math changes how a loss reads. If you expect most tests to lose, a losing test isn't a failure of the program — it's the program working. The ideas that would have quietly hurt the business get caught before they ship to 3 million customers. The fault lines in your user base get mapped. And the rare big win arrives with the statistical confidence to bet on it.

If AI Runs the Test, Why Run It?

The conversation turns sharpest when Stirrup asks where experimentation at Squarespace goes next. Blackman's answer starts where everyone's does — AI — and then takes a turn most don't.

Yes, AI is speeding up the A/B testing workflow, and her team is already experimenting with it. But she's wary of automating away the thinking. "The tests that are less risky, sure — have an AI run the analysis, have an AI run the test brief," she says. "But then I might question: why are we running it then?"

Her emerging dividing line: hand AI the mundane parts of the analyst workflow — tracking, assignment setup — and keep the methodical reasoning human. "We'll still have to approach some parts of the process manually, because we need to be thinking methodically through each of the steps."

The same clear-eyed pragmatism applies to AI-powered features themselves. With every company releasing a chatbot, the launch decision is often already made. "We know we wanna roll this out because that's what the industry is doing," Blackman says. "But we could definitely leverage experimentation for optimizing the experience or the entry points." The test isn't whether to ship — it's whether the prompt quality, the entry points, and the long-term value actually hold up once users arrive.

The Wreckage Is the Asset

The blank template story tends to get retold as a redemption arc: failed launch becomes beloved AI builder. But the more useful reading is structural. Squarespace didn't get lucky. It had embedded analysts close enough to the product to dig past the surface metric, a program disciplined enough to ask the two questions, and a culture honest enough to kill a design forever without killing the learning.

That's the real argument for experimentation — not that it produces wins, but that it converts losses into direction. The blank screen didn't survive. What Squarespace learned from it is now the front door to the product.

Ready to put more rigor behind your own losing tests? GrowthBook is the open-source feature flagging and experimentation platform built for teams that want to ship winning experiments — and learn from the rest. Start for free or get a demo at growthbook.io.