Somewhere in the first week of every identity project, we produce a page of marks that are, frankly, not very good. Obvious metaphors. Overworked lettering. A monogram that looks like three others we’ve seen this year. We keep every one of them.

This isn’t a failure of the process. It is the process. The bad logo’s job isn’t to be chosen — it’s to rule something out, cheaply, before the team has spent three weeks and a lot of goodwill discovering the same thing the hard way.

Divergence before judgment

The biggest risk in any identity sprint is converging too early. Once a team falls in love with a direction in week one, every later idea gets measured against it unfairly, and genuinely better solutions never get a fair hearing. So we deliberately overproduce in the beginning — quantity as a hedge against premature certainty.

We also show rougher work than most studios are comfortable sharing. A polished mockup makes a mediocre idea look finished, which makes it harder to kill. A pencil sketch makes a strong idea look promising, which makes it easier to build on. We’d rather clients react to the idea than the render.

What the bad logos taught us, this time

For Novaflor, a small floral-adjacent lifestyle client who wanted to feel less like a delivery service and more like an actual walk through the garden their arrangements come from, the first round leaned hard into literal iconography — a rose outline, a stem, a watering can stacked three ways. All perfectly fine. All forgettable.

The mark that shipped — a four-petal flower built from clean, continuous linework, sitting white against a motion-blurred bouquet — came from a much rougher sketch two rounds earlier that nobody on the client side even flagged as a favorite. It just kept surviving the cut, which turned out to be the more honest signal in the room.

By the time we reach the version that ships, it rarely resembles anything from that first page. But it couldn’t exist without it. The bad logos aren’t a detour on the way to the good one. They’re the map.