Baking

The baking channel in our Intelligems Slack popped off over the holiday week. 

We had six different types of rolls, a loaf of bread, three different style turkeys, a sourdough stuffing, four flavors of ice cream and a pizza. (We take some liberties with what’s allowed in #baking.)

It’s one of my favorite channels, because baking has been a shared experience at Intelligems since the very beginning: When Drew and I started Intelligems, we had a newsletter that went to investors, friends and family. At the end of it, we included a section called “Star Baker” that highlighted what one of us had baked recently. 

It’s sort of stuck as part of our culture, I think, because the parallels between baking and testing are so strong.

Anyone who has baked bread knows the mix of science and art that defines the whole hobby.

There are rules you must follow — hydration ratios, gluten development, fermentation time. But then you do your best with the ingredients available and the time you have, then you close the oven door and wait. 

Whether you nailed it or not depends on a lot more than just following the rules: Humidity swings, the kitchen runs a little warm or a little cool, the dough behaves differently. Sometimes a loaf falls short not because the baker made a bad decision, but because the context changed. And you can’t taste and tweak halfway through like you can with pasta sauce.

You learn only at the end whether your choices were right, which means the feedback loops are long.

Drew’s favorite parts of baking, for instance, are shaping and scoring sourdough loaves, but they’re also the ones you get the fewest reps at. You spend hours preparing just to practice that motion once; you’re lucky if you get the chance to try it again tomorrow.

And then there’s taste. You can follow every best practice and still discover that what you think is ideal isn’t what your audience wants.

I have a friend whose brother is opening a bakery in Cincinnati. Their goal is to be known for the best chocolate chip cookie in the city. 

How hard would that be!?

Of all the different permutations that you can have for a chocolate chip cookie, you have to choose one. If you tried to do even two different types, people wouldn’t even know which to get. That’d be too much for them.

Testing looks a lot like that.

Every brand is working with slightly different contexts: different shoppers, different product margins, different seasonality, different design language. Even a “proven” site design won’t necessarily resonate with a different audience or if the market shifts. The goal isn’t to find the universally perfect product detail page — that doesn’t exist — but to keep moving toward versions that work better for the people actually visiting your site.

Because of that, you end up with that same tension between science and art. There are formal rules — sample sizes, significance levels — and ignoring them is like skipping salt in a loaf of bread. But even within those structures, there’s room for taste. Two winning variants can emerge in a tie, and you still have to decide which one better fits the experience you’re trying to create for your customers.

Testing and baking, I guess, are really just ongoing exercises in curiosity. You form a hypothesis, you commit to it, and you learn from the outcome. You have to respect fundamentals rules, but still allow space for creativity, taste, and the occasional adjustment that comes from knowing your environment (and when to break some of the rules).

And you don’t really ever get to declare the work finished, because conditions change and preferences shift.