This is not a page.It is a map.
Verso Atlas is a five-person research practice. Everything we have published sits on this canvas — drag to move, scroll to zoom, or use the list on the left.
Drag anywhere · Scroll to zoom · Press F to fit
Verso Atlas is a five-person research practice. Everything we have published sits on this canvas — drag to move, scroll to zoom, or use the list on the left.
Drag anywhere · Scroll to zoom · Press F to fit
Five people — two designers, an ethnographer, a statistician and a policy lead. We take three commissions a year, all of them about how public systems are actually used rather than how they were specified.
A ticketing redesign that tested beautifully and failed in the concourse. We watched 4,100 transactions and found the problem was not the interface at all.
A benefits application with a 62% abandonment rate. Rebuilt around what applicants actually had to hand, rather than what the department wanted to know first.
Across four unrelated public services, completion collapses somewhere between five and seven minutes of continuous attention. Below six, most people finish. Above it, most do not.
Wayfinding for people who are frightened, in pain, or arriving in the dark. Tested with people who had actually done it.
People are reliable narrators of what they want and unreliable narrators of what they do. Both matter, but only one can be observed.
A pilot with no defined end is a way of avoiding the decision to roll out or to stop. We now write the end condition into the first week of every engagement.
No unmoderated remote testing as primary evidence. No surveys as a substitute for observation. No recommendations we have not watched someone try.
We take on three pieces of work a year and we are usually booked two quarters out. Tell us what is going wrong and we will tell you honestly whether we are the right people.
We were founded in 2018 after a transport authority asked us why a redesigned ticketing system had made queues longer. The answer took four months and involved standing in stations at six in the morning. That is still roughly the method.
We publish everything we can. Where a client will not let us publish the work, we publish the method.
The new machines were faster per transaction and slower per passenger. People arrived in groups, and the old machines allowed one person to buy for four. The new flow did not, so a group of four became four queue positions.
We recommended a group-purchase path and a physical reorganisation of the concourse. Queue length fell 38% within a month of the change, before any software was rewritten.
The original form asked for a National Insurance number on page one. Roughly a third of applicants did not have it in front of them, left to find it, and never returned.
We moved identity verification to the end, made every field saveable, and cut the form from 44 questions to 19 by taking the rest from records the department already held. Completion rose to 81%.
This holds for a benefits form, a vehicle registration, a school application and a clinical questionnaire. It appears to be about the length of an uninterrupted moment in an ordinary day, not about the difficulty of the task.
The practical version: if your service takes longer than six minutes, it must survive being abandoned and resumed. Most cannot.
We interviewed 60 people within a week of an unplanned hospital visit. Almost none remembered reading a sign. They remembered following someone, asking someone, or a colour.
The redesign leant on colour and floor markings rather than text, and put one staffed point in the entrance rather than four unstaffed signboards. Time from door to correct department fell by a third.
We ask afterwards, never instead. The pattern of a study is: watch for a week without talking to anyone, then interview using what you saw as the prompt.
It is slower and it embarrasses everyone slightly, and it is the only reason our recommendations survive contact with a real building.
Three of our early projects ended as permanent pilots — running, funded, unloved and unscaled. The fix was procedural, not analytical: we agree in advance what result would justify rolling out, and what result would justify stopping.
Each of these is a way of getting a number quickly that will not survive the building it is meant to describe. We will use all three as supporting evidence and none of them as the argument.
This costs us about one commission a year, which we consider a reasonable price.
Write to studio@versoatlas.example.com with what is happening, what you have already tried, and who inside your organisation cares about the answer. That last one matters more than the budget.
We reply to everything within a week, including the ones we turn down.