I Spent Three Months Coding 80 Interviews by Hand. Then I Found Out There Was Another Way. I Am Still Angry.
I Spent Three Months Coding 80 Interviews by Hand. Then I Found Out There Was Another Way. I Am Still Angry.
Nobody warned me about the coding phase.
My supervisor mentioned it. My methods textbook described it. My PhD cohort talked about it in the vague way people talk about something unpleasant that is still far away.
None of that prepared me for sitting at my desk at 11pm on a Wednesday in October, on interview number 34 of 80, realising I had been doing this for six weeks and I was not even halfway through.
I had 46 interviews left to code. I had a chapter draft due in eight weeks. I had a supervisor who was kind but firm about deadlines. And I had a growing suspicion that the thing I had spent four years training to do, which was think carefully and rigorously about complex human experience, was being almost entirely consumed by a mechanical, repetitive, mind-numbing process that had nothing to do with thinking.
I was not analysing. I was sorting.
What Manual Coding Actually Feels Like
If you have never done qualitative research, here is what manual coding in NVivo or ATLAS.ti actually looks like at two in the morning six weeks in.
You read a sentence. You decide what it means. You assign it a code. You read the next sentence. You decide what it means. You assign it a code or create a new one. You read the next sentence.
You do this for every sentence in every transcript. Eighty interviews. Each one forty-five to sixty minutes long. Each transcript fifteen to twenty pages. You are making thousands of individual decisions, each one feeling consequential, each one feeling like it might be the wrong call, each one sitting in a growing structure of categories that you built in week two and are no longer sure accurately represents what the data is actually saying.
By week six you are not a researcher. You are a very slow, very tired, very expensive text-sorting machine.
The thinking part, the part you did a PhD to do, the part where you look at what human beings said and work out what it means and why it matters, that part is scheduled for later. After the coding. After the months of organising. After the process that was supposed to support the analysis has consumed the time and energy that was supposed to go into it.
I finished my 80 interviews. It took eleven weeks. I submitted the chapter. My supervisor said the analysis was good but she wished I had gone deeper on three of the themes.
I did not have the energy to go deeper. I had spent it all on the coding.
The Message That Changed Everything
Four months after I submitted that chapter, a friend in my cohort sent me a message at 9am on a Monday morning.
It just said: "have you tried qinsights. genuinely asking"
I had not. I had heard the name somewhere, maybe in a methods forum, maybe in a conference side conversation. I had not tried it because I had assumed it was another software tool with a steep learning curve and a free trial that would take two weeks to set up properly and still not solve the actual problem.
I was wrong about all of that.
What Qinsights Actually Is
Qinsights is an AI qualitative research platform built by Dr. Susanne Friese. If you are in qualitative research you may know her name. She wrote the definitive methodological guide to ATLAS.ti. She has been training researchers in qualitative data analysis software since 1992. She is one of the most recognised methodologists in the field.
She built Qinsights because she spent thirty years watching brilliant researchers spend most of their time doing something a machine could do better, faster, and without getting tired at 2am.
The platform is built around what she calls Conversational Analysis with AI. You bring your transcripts. You talk to them. You ask questions. What are the dominant themes across these interviews? Where do participants contradict each other? What is the tension between what people say they believe and how they describe their actual behaviour? What am I missing?
The AI reads across your entire dataset simultaneously. It surfaces patterns. It brings evidence. It links every single insight back to the exact line in the transcript, the exact timestamp in the recording, the exact speaker in the focus group. Every claim is traceable. Nothing is invented. Nothing is inferred without a source.
And you stay in charge. That is the part I want to emphasise because it is the part I was most sceptical about before I tried it. Dr. Friese built this platform with thirty years of understanding of exactly where the line is between what AI can genuinely do and what only a trained researcher with disciplinary knowledge and interpretive judgment can do. Qinsights handles the volume and the structure. You handle the meaning.
What Happened When I Used It
My next project after that chapter was a set of 45 semi-structured interviews for a collaborative research project. I uploaded the transcripts to Qinsights on a Tuesday morning.
By Thursday afternoon I had a thematic framework, a set of sub-themes, supporting evidence for every claim, and a synthesis document that my co-investigator described as the most clearly structured analysis she had seen from any project we had worked on together.
I had not coded a single line manually.
I want to be precise about what I mean when I say the analysis was good. I do not mean it was fast and acceptable. I mean it was rigorous. Every theme was grounded in the data. Every claim had a traceable evidence chain. My co-investigator, who had not been involved in the data collection, said she felt she understood the dataset from reading the analysis in a way she rarely felt from reading traditionally coded outputs.
The themes I found were not the themes I would have found after eleven weeks of manual coding with my analytical capacity running on empty. They were the themes I found when I had the cognitive space to actually think about what I was seeing.
I am angry because nobody told me earlier.
I am angry because I spent eleven weeks doing something that took two days. I am angry because the energy I spent sorting text in the early hours of October and November was energy I could have spent on the interpretive work I trained for years to do. I am angry because my chapter could have been deeper and I would have had the capacity to make it deeper if I had not been so depleted by the time I got there.
I am not angry at NVivo. I am not angry at ATLAS.ti or MAXQDA. They are tools that were built for a world where what Qinsights does was not possible. They were the best option available. They were not good enough.
I am angry at the gap between what qualitative research could have been for researchers like me and what it was, for years, because the field kept accepting a bottleneck as a permanent feature of the work instead of a solvable problem.
The bottleneck is solved. Has been for a while now. And too many researchers are still sitting at their desks at 2am in October, on interview number 34 of 80, not knowing that.
What I Want You to Do With This
If you are a qualitative researcher and you are in the coding phase right now, or dreading the coding phase of your next project, or avoiding starting a project because you know what the coding phase costs you, I want you to try Qinsights before you spend another week doing it manually.
Not because it is faster. Because when you are not exhausted from eleven weeks of sorting, you do better research. You go deeper on the themes that matter. You notice the tensions your manual process would have flattened. You bring your actual expertise to the analysis instead of bringing whatever is left of it after the process has taken the rest.
Dr. Susanne Friese built this platform because she believed qualitative researchers deserved better than spending most of their time on the part of the job that should require them the least.
