We didn't start it because it was easy

There is an old line about going to the moon: we choose to do these things not because they are easy, but because they are hard.

Ours is the wrong version. We didn't start this because it was easy. We started it because we thought it was easy.

That's the honest story. A few people who are good at data science, sitting in a moment when the tools suddenly got strange and powerful, thinking: how hard can it be to run a company where the machines do a serious amount of the work? Answer: quite hard, actually. Harder than the maths, which nobody expected. We are keeping the joke, because the joke is the plan.

It's the hipster version of running a company, and we're fine with that description. We experiment. We try the odd thing first, because the odd thing sometimes works, and because when it doesn't we find out in an afternoon instead of in month three. We throw prototypes away on purpose. We have broken a leg or two, publicly, and told everyone about it, which is what hipsters do when they fall off something.

Here is the difference between that and chaos: we are committed to delivering. The experiment is how we find the answer, not the answer itself. When the odd thing works, it ships. When it doesn't, you will hear about it early, in plain language, from a person, and then we do the boring version and ship that.

I'll give you the shape of it. Someone arrives with a problem that sounds enormous: forecast this, clean this, automate this, make sense of the pile of things we've been ignoring since 2019. Two responses are available. One is forty slides and a discovery phase. The other is a small ugly script that answers the real question by Thursday, plus a sentence explaining what we still don't know.

Which of the two you get is a decision, not a preference. We can build the forty slides, and sometimes that is the right answer: a programme with a committee behind it, a decision that has to survive a reorganisation, a question whose data nobody has opened yet. Say the word and we will. But we will also say when the small ugly script is the honest answer — before you have paid for the deck, not after. We would rather be the supplier who tells you the cheap truth than the one who invoices you for the beautiful version.

Which is the offer, really: we take the part that can be wrong cheaply, before you spend money on the part that cannot.

What we're actually selling, underneath, is problem solving. Not enthusiasm, not a deck, not a "journey". We are built for the awkward middle of a problem — the part where the data is wrong, the definition is unclear, and the person who wants the answer is not the person who owns the data. That middle is where companies lose months. It's also the part most suppliers would rather not discuss.

And creativity here is a method, not a personality. It means we keep a list of dumb ideas on purpose. It means the person who cleans the data is allowed to suggest a different question. It means the smallest possible thing that could prove us wrong gets built first, because being wrong cheaply is a skill.

We didn't start it because it was easy. We started it because we thought it was easy, discovered we were wrong, and stayed anyway. That last part is the bit we'd like to be judged on.