Our favorite mistakes
We never understood treating failure like a label. Instead we learned that failing fast is the defining skill of the age of rapid iteration.
First, we should probably define rapid iteration for anyone who hasn’t lived inside it, because the phrase gets thrown around like everyone agreed on a meaning.
The loop
Rapid iteration is a loop, and the loop is simple.
- Ideate. Have the idea.
- Test. Ship something real as fast as possible: not a deck about the thing, the thing.
- Learn. Let reality respond.
- Refine. Keep what held, let go of what didn’t, and run the loop again until the signal is clear.
The loop itself isn’t new; lean and agile teams have run versions of it for decades, from Eric Ries’s Lean Startup back to Deming’s plan-do-check-act. What’s new is how fast it spins.
For most of this industry’s history you had no choice but to front-load everything. Long strategy phases, careful prototypes, approvals in triplicate, all because a minor failure discovered late could mean catastrophic impact. Blown budgets, missed launches, relationships strained past repair.
The evolution of new tools and AI flipped that math. You can now iterate fast and fail fast with minimal pain, and the failures usually pay for themselves: each one carries learnings that make the next pass measurably better. By the time most agencies finish their first version of something, we’ve already failed ten or twelve times, and finished not just that project but three or four others we failed our way through to get there.
Which brings us to failure, and why we’ve never understood treating it like a label.
In most companies, a failure is a verdict. Something went wrong, so something must be wrong with the plan, the team, or the person. But inside the loop, a failure is just a measurement. It’s the moment reality answers a question you’d otherwise still be debating in a conference room. The campaign that lands flat tells you something true about your audience that no amount of planning would have surfaced. The layout that tests badly is reality doing you the favor of disagreeing early, while the stakes are still small. You paid for that information. The only actual waste is refusing the receipt.
Ten thousand ways not to make a lightbulb
The story goes that Edison, asked about all his failed attempts at the bulb, said he hadn’t failed at all; he’d found ten thousand ways that don’t work. The quote has been polished by a century of retelling, but the posture underneath it is the whole thing. He wasn’t reframing failure to feel better. He was describing his actual method. The ways-that-don’t-work were the product. Each one narrowed the search. The bulb wasn’t discovered in spite of the ten thousand; it was discovered through them, one eliminated wrong answer at a time.
Nobody remembers attempt 4,217. Everybody has a light on in the room they’re reading this in.
The skill is speed
Trying has never been cheaper, and when trying is cheap the constraint stops being “can we afford to fail” and becomes “how fast can we find out.” The teams winning right now aren’t the ones failing less. They’re the ones failing sooner, on smaller stakes, with better notes.
Failing slow is the expensive version.
The failure costs the same; you just paid extra for the delay. Learning to fail quickly is a genuine skill, and like most skills it looks like a personality trait from the outside. It means shipping the rough version while your instinct begs for one more week of polish. It means designing the test so the answer arrives in days, not quarters. It means saying “we were wrong about that” in a meeting without ceremony, the way you’d report the weather, so the room can move to what’s next instead of managing anyone’s feelings about it. Around here that sentence costs nothing. That’s on purpose. The moment being wrong gets expensive socially, people stop reporting their measurements, and then you’re flying on instruments you know are lying.
None of this is an argument for carelessness. There’s a difference between a fast failure and a sloppy one: the fast failure was designed to teach something and did; the sloppy one taught nothing because nobody defined the question. Fail on purpose, in the direction of a question. Sloppy is just noise with a budget.
So no, we don’t keep a highlight reel of our mistakes, and we don’t keep a blooper reel either, but we do keep the lessons…
