--- title: A year of research notes date: 2025-10-02 description: What twelve months of writing up every research session in the same short format taught me. tags: [research, writing] --- In October last year I started writing up every research session in the same format, in a plain text file the day it happened. Twelve months later I have 61 of them. I had expected the archive to be the benefit. It turned out the benefit was the habit of writing, and the archive was a pleasant side effect. ## The format Each note has five parts, and I do not skip any, even when a session went badly. Who the participant was and what they do. What I asked them to do. What they did. What surprised me. What I would ask next time. That is it. Most are around 250 words, and the shortest is 80, from a session in which the participant's laptop died after four minutes. The "what surprised me" line is the one that matters. In the first month it was often empty, which told me something about how closely I was listening. By March I was filling it in nearly every time. ### Why bother with the "next time" line When I wrote the first notes, I treated them as records. Then in January I opened the old notes before planning a study with Ravi, and found that three of my "next time" questions from the autumn were exactly what we were about to ask again from scratch. Those lines are a to-do list written by someone who had just watched a real person use the thing.[^why] ## What the archive showed Once I had 40 notes, I read them in a row. A pattern appeared that I had not seen from inside any single session: people almost never read the explanatory text on a screen. They looked at it, I could see their eyes move, and they carried on. Of the 61 sessions, 47 had at least one moment where a participant skipped text that would have answered their question. That changed how I argue in design meetings. I used to say "people don't read", which people dismiss as an opinion. Now I say "in 47 of 61 sessions", and then I stop talking. Lucia, who is usually the person asking for another paragraph of help text, asked for the list of sessions. I sent it. She did not ask for the paragraph again. ### A caution about counting I am a bit wary of that number. The 61 sessions are not a random sample. They are the ones I ran, with the customers who agreed to talk to me, mostly on desktop, mostly in English. The count says something about those sessions and only a little about everybody else. I try to say that out loud when I quote it. ## What I would change The format is too tilted toward moments of trouble. I notice friction and write it down, and I notice smooth moments much less. When Ravi read the archive, he asked where the evidence was for the things that worked. There was almost none, because I had not been writing it down. I added a sixth line in July: "what went smoothly". It is already shorter than the other five and I think that is accurate. I would also write the notes in front of the participant when I can. A few of the best sessions this year ended with me reading back my notes aloud and asking whether they were right. Twice the participant corrected me, and both corrections were better than what I had written. The interviews feed into this too. The questions I keep coming back to are in a [separate note](/notes/interview-questions), and they have not changed much, which I find reassuring. ## Practical bits The files live in one folder, named by date. I search them with the same search I use for everything else. Every few months I read the last ten in a row, without taking notes, to see what repeats. It takes about forty minutes and I look forward to it more than I expected. If you start one of these, the only advice I have is to write the note before you eat lunch. Notes written after lunch are, in my archive, about a third shorter and almost always say "went well". [^why]: The label is a habit from when I tried writing footnotes with numbers and kept renumbering them after edits. A word never needs renumbering.