Esmé — We build the proof
What if you could prove that…
- If we generate images of paint defects, then we can train a computer vision model to find the real ones.
- If a toy brick carries an accelerometer, then we can tell jumping from driving, and predict what the child built.
- If we know the venues, cities, hotels and airports, then we can plan every football match, for every age group, across a whole country.
- If survey data serves as ground truth, then we can predict what a real audience would say.
- If we train a language model on research into marriage, then it can act as a marital counsellor.
- If we process historic legal documents, then we can write a playbook for a legal department.
- If a phone camera can judge the wear on an excavator’s track shoes, then we can calculate what the excavator is worth second-hand.
- If we generate population data for fictional neighbourhoods, then we can furnish entire video game cities automatically.
Esmé builds the proof your business needs to reduce risk and scale up.
Each of these started as someone’s hypothesis. We built the proof, and found out.
If we generate images of paint defects, then we can train a computer vision model to find the real ones.
If an HR email comes in, then we can work out the preventative care that fits it best.
If a toy brick carries an accelerometer, then we can tell jumping from driving, and predict what the child built.
If we know the venues, cities, hotels and airports, then we can plan every football match, for every age group, across a whole country.
If we fit special solar panels, then the northern lights can add a little power to a green fleet of arctic vessels in winter.
If survey data serves as ground truth, then we can predict what a real audience would say.
If we 3D-scan your head, then anthropology can tell us which glasses frame suits you best.
If we train a language model on research into marriage, then it can act as a marital counsellor.
If we process historic legal documents, then we can write a playbook for a legal department.
If a toy recognises a child’s voice, then the child can lock their toys.
If a phone camera can judge the wear on an excavator’s track shoes, then we can calculate what the excavator is worth second-hand.
If we generate population data for fictional neighbourhoods, then we can furnish entire video game cities automatically.
There are no names here. Most of what we build is competitive information, and we keep it that way. If one of these makes you curious, get in touch.
One question, one proof
Here’s how this works. You tell us what you think might be possible. We spend a few weeks finding out, usually about four. At the end you see it running, and you know what it would take to go further.
Hypothesis
If we do this, then that. We start by making it precise enough to test.
Proof
We pick the technology the problem needs and build the smallest thing that settles the question.
Demo
Working software, live, unscripted. Plus what we learned on the way: what it would take to scale, who you’d need, what it would cost.
Decision
Yours, with evidence in hand. Everything we built is yours to keep.
The budget is fixed and sized so that the person who owns the problem can sign it off. And within weeks, there’s something real to show upstairs.
On your side of the table
When the proof holds and the scope is clear, we build the whole thing. We work next to your engineers and product owners while we do it.
Everything is yours: the source, the repositories, the research. You can look at our work at any time. We check in every two weeks, and every quarter you decide how we continue.
What we build is made to run without us. By the time we hand it over, your team knows the system as well as we do. And if we notice something else worth testing along the way, we’ll flag it.
What to build next
Sometimes a proof raises a bigger question: where should all of this go? Then we write a roadmap. The next few months in detail, the coming year as a plan, the year after as a direction. Specific where it can be, open where it should be.
The next few months
In detail
The coming year
As a plan
The year after
As a direction
Our roadmaps have been included in M&A documentation, where they give both parties a clear view of where the business can go next.
About us
Esmé is a group of very senior creative technologists, with fifteen to twenty years each. Between us, we’ve worked at Monks, DEPT, Code & Theory and Google, building for the biggest brands and for the companies that make the technology. We’ve also won most of the awards this industry gives out.
We scale to fit the proof.
The name comes from “subject matter experts”. Throughout our careers, people like us were the ones called in to explain technical things in ways anyone could follow, and we got called SMEs. Say it out loud often enough and it turns into a name. We kept it.
Questions
What if the proof shows it won’t work?
Then you know before you’ve invested in it, and you know why. If we suspect a hypothesis won’t hold, we say so before we start, with our reasons. Sometimes we’d rather not take the work.
Will you tell us if we should spend less?
Yes. A science publisher asked us to adapt copy for forty identified audiences. When we tested the output, the versions were so alike that the differences weren’t worth making.
The research pointed to reading level instead: even highly educated readers prefer simple copy. We suggested three versions instead of forty. That meant fewer hours for us and more value for them, and we’re still working together.
What if the technology changes before we’ve finished?
We’ll tell you if it’s about to. One client wanted a tool for high-end post production. We tested a different workflow that new technology already made possible. They didn’t build the tool, which would have been out of date by the time it was done.
What if something is technically true, but unwise?
We’ll say so. We once proved that solar panels could draw power from the northern lights. The power was real, but too small to claim it made a difference. We advised against building a public story on it.
What does a proof sprint cost?
It depends on the scope. It’s sized so the person who owns the problem can sign it off without asking for approval. The budget for a proof sprint is fixed: if it takes us longer than we planned, that’s on us.
Tell us your hypothesis and we’ll give you a number we’ll stick to.
How long does a proof sprint take?
Usually about four weeks. Some questions settle faster, and some need longer. We agree on the length before we start.
Who owns what you build?
You do: the source, the repositories, the research. You can look at our work at any time.
What happens after the proof?
Usually a roadmap of what to build next, a build retainer, or both. The retainer is typically one full-time equivalent, with updates every two weeks and a decision every quarter. We build it with your team and hand it over.
Can we see examples of your work?
Most of what we build is competitive information. We respect our clients, so we don’t publish it. The library above shows the kinds of questions we’ve answered. If you’d like to know more about how we work, get in touch.
Your hypothesis
Tell us how yours goes. We’ll tell you whether we think it holds, and how we’d find out.
Or write to us