One shortcut in qualitative fieldwork that almost everyone has taken at some point - instead of transcribing an interview, listen to the recording and key in the respondent's answers straight into an Excel grid. It eliminates one step from the process, the analysis is arrived at sooner. On paper - it's a cheaper workaround. It is a widespread practice, and defensible until someone asks a question about the data.

Here's what happened when a healthcare research agency worked out what this shortcut was costing them, and what changed when they stopped taking this shortcut.

The agency

A specialist healthcare research agency, with over $22 million in annual turnover and employing more than 50 researchers across their US and UK offices. They run both qualitative and quantitative work, largely for big pharma clients. Their reputation rests on the quality of interpretation their research directors deliver.

The problem

Simply put - analysis with nothing underneath it. The agency's fieldwork partner was using the standard shortcut. Recordings came in, responses were keyed directly into a grid, and no transcript was ever produced. The obvious loss was the moderator questions. When you key in answers only, what remains is a column of responses - detached from how the moderator prompted, no probes, no follow-up, no indication of whether the respondent volunteered a view or was walked toward it.

For a research director writing a report for a pharma client, that was a serious problem, and it showed up in three ways:

1. Nuance disappeared. Meaning in a depth interview often sits in the sequence - what was asked, how it was answered, what the moderator did next. A grid cell preserves the answer, but discards the sequence.

2. Confidence eroded. Directors were writing findings from a dataset where - in their own assessment - half the interview had already been dropped, even before analysis began.

3. There was nowhere to go back to. This was the issue that mattered most. When a question arose about how a finding had been interpreted - internally, or from the client - there was no source document. No transcript to search, no passage to re-read, no way to check whether an interpretation held. As one of their researchers put it - they had no avenue to turn to.

The shortcut had not removed work from the process. Instead, it had moved the work onto the most expensive person in the chain (Director level), and stripped away the evidence they needed to do that work.

The solution

The agency brought myTranscriptionPlace, a 20-interview study to start with. The requirement was specific: full verbatim transcripts and a coded analysis grid, without adding days to the report-writing schedule.

Our process ran in four steps:

1. AI-drafted and human-corrected transcription. AI produced the first draft of the transcript. A human reviewer worked through it (using the audio) and corrected it. Identifiable information was redacted at this stage.

2. The discussion guide became the analysis frame. We gave the AI the discussion guide and had it parse out the individual questions and topics.

3. The client corrected the frame, before any coding began. This is the step that made the process work. The agency's researchers reviewed the parsed topic list and reshaped it to match how they actually intended to analyse the study - merging certain topics, splitting others, reordering to fit their reporting logic. Only once they had signed off, did any coding happen.

4. Coding, then human review of placement of data under various codes. The AI located the relevant passages for each topic across every interview. Our analysts then reviewed the output to confirm the data had been placed correctly.

Each study was delivered as verbatim transcripts, a coded Excel grid - with respondents as rows and guide questions as columns - and topic-level summaries with notable quotes.

Why the solution worked for the client

The AI never decided what the study was about. Instead, the researcher defined the frame, the machine populated it, and humans checked both the transcript and the data placement. What the agency received was an analysis built on their logic, traceable back to a verified source document.

Secondly, healthcare qualitative work carries obligations that a general transcription service is not set up for. On this engagement:

• myTranscriptionPlace's ISO 27001 certified and HIPAA-compliant processes stood up to the client's procurement norms.
• An NDA was signed alongside the agency's MSA.
• Every transcriber working on the account was trained on adverse event reporting, tailored to the client's specific requirements.
• Identifiable data was handled and redacted by our team.
• Data processing was conducted on US servers. The correction and review team worked from India.
• On completion of every project, we obtained written confirmation of data deletion from everyone who had touched the files.

The result

	Pre engagement with myTranscriptionPlace	After partnering with  myTranscriptionPlace
Verbatim transcript	Not produced	A standard deliverable
Per-interview turnaround	2–3 days, transcript alone	2 days, transcript and coded analysis
Last interview to full analysis	5–7 days	3 days
Cost vs. keyed-in Excel with no transcript	Baseline	20% lower
Cost vs. transcription and analysis bought separately	Baseline	50% lower

The agency's analysis went from roughly a week after fieldwork closed, to three days; all while gaining the transcript layer they were entirely missing out on.

The cost comparison is the part worth considering. Adding transcription back into the process made it cheaper than the shortcut that had removed it, by 20%. Against the conventional alternative of commissioning transcription from one vendor and content analysis from another, the saving was 50%.

Most importantly, the analysis became traceable. When an interpretation was questioned, there was now a document to return to.

One of the agency's research directors wrote to us mid-project: "I just want to thank you and your team for the really nice work on the topic level summaries section of the CA [content analysis]. This report is a major crunch to finish on my end right now and I am primarily using that section to report-write right now. It would take me so much longer if I didn't have the summaries to start off from."

What followed

The 20-interview trial became an ongoing engagement: over 1,000 interviews a year for two years running, spanning English-only and multi-market studies (Spanish, Italian, Japanese, German, French and several other languages). 

When requested to provide a reference to industry body BHBIA, one of our contacts at the agency wrote: "We have been working with them [myTranscriptionPlace] on several projects. We have commissioned them for transcript and content analysis services on global projects, covering several languages and markets. The team has displayed full cooperation in learning our compliance requirements and produced quality work so far."

The takeaway

Transcription looks like a line item in the research process. More often than not, it is treated as one, and is usually the first thing compromised on, when facing a compressed timeline. However, a transcript is not merely a deliverable. Instead, it is the audit trail that researcher interpretation rests on. Remove it and the work does not disappear - it reappears later, as a research director struggles to reconstruct context from memory, or defend a finding they cannot evidence.

This agency did not trade speed for rigour. They got both, for far less than the shortcut had been costing them.

---

If you are running healthcare qualitative work and want to see what this looks like on your own data, send us a single recording and a discussion guide. We will return the transcript and a coded grid so you can judge the output rather than our pitch.