Key Steps in Data Presentation: Editing, Coding, and Transcribing

When researchers finish collecting data – whether through surveys, interviews, focus groups, or field observations – they don’t immediately have findings. What they have is a pile of raw material: messy, incomplete, inconsistent, and often difficult to interpret. Before any analysis can happen, that raw data must be processed. The three core steps in this process are editing, coding, and transcribing. Together, these steps transform unorganized primary data into accurate, consistent, and analysis-ready information. Understanding each step – what it involves, why it matters, and how it’s done – is essential for any researcher who wants their work to hold up to scrutiny.

Why data processing matters before presentation 🔗

Data processing sits between data collection and data analysis. It is not a minor administrative task – it is the stage that determines whether your findings will be credible. Raw data collected from questionnaires, interviews, or observation schedules is rarely clean. Responses may be incomplete, answers may contradict each other, or the handwriting may simply be illegible. Without processing this data carefully, even the most carefully designed study can produce flawed conclusions.

The processing stage typically includes editing, coding, classification, and tabulation. This post focuses on the first three, which are the most foundational steps – and the ones most directly tied to accuracy, consistency, and homogeneity of data.

Step one: data editing 🔗

Editing is the first thing a researcher does with collected data. Editing is the process of examining collected questionnaires or schedules to detect errors and omissions, and to correct them so the data is ready for further processing. The goal is to ensure that every data point is accurate, complete, consistent with other responses, and entered in a uniform format.

According to Wikipedia’s guide on data editing, the process involves reviewing data for consistency, detection of errors, and outliers – values that are extremely unusual and may indicate entry mistakes. Editing also verifies that all required fields are filled and that no response has been duplicated in the dataset. These checks form the quality foundation that every subsequent step depends on.

Field editing vs. central editing 🔗

There are two main types of editing, and both serve distinct purposes. Field editing is carried out by the investigator or enumerator shortly after data collection – often on the same day. This is the time to fix abbreviated or illegible entries while the context of the interview is still fresh. The key rule here: field editing should never involve guessing or fabricating data to fill gaps.

Central editing, on the other hand, happens after all completed forms have been returned to the office. A dedicated editor reviews the entire dataset systematically. As noted by the Food Safety Institute, central editing involves checking for completeness, consistency, and logical errors across all records. For instance, if a respondent says they are 30 years old but also states they have been working at the same company for 35 years – that’s a logical inconsistency that requires attention. Editors can correct obvious errors or, when the correct answer cannot be determined, simply record “no answer” rather than inserting uncertain data.

Types of checks used in editing 🔗

Statistics Canada’s methodology guide outlines several types of editing checks that researchers apply:

• Validity edits check each field individually to ensure values fall within an acceptable range and are in the correct format (e.g., a numerical field shouldn’t contain text).
• Duplication edits verify that each respondent appears in the dataset only once, preventing any single response from skewing results.
• Consistency edits compare different answers from the same respondent to detect contradictions between fields.
• Historical edits compare current data to previous rounds of a survey to flag unusual changes that may indicate errors.

Each editor should initial and date every completed form after review. This creates an audit trail and maintains accountability throughout the editing process.

Step two: data coding 🔗

Once the data has been edited and cleaned, the next step is coding. Coding is the process of assigning numerals or symbols to responses so that they can be grouped into a limited number of categories for efficient analysis. It is what allows a researcher to move from a long list of individual responses to a manageable set of categories that can be counted, compared, and interpreted.

For closed-ended questions, coding is relatively straightforward – response options are predefined, and each one simply receives a numerical value. For open-ended questions, the process requires more judgment. Researchers read through all responses, identify recurring themes, and develop a coding scheme or codebook that defines how different types of answers will be categorized. For example, if a survey asked “What barriers did you face at work?”, responses might be grouped under codes like “1 = lack of resources,” “2 = poor communication,” “3 = insufficient training,” and so on.

Properties of a good coding scheme 🔗

Regardless of whether coding is done for quantitative or qualitative data, the coding categories must meet two fundamental criteria. First, they must be mutually exclusive – each response should fit into only one category, with no overlap. Second, they must be exhaustive – every possible response must have a category to go into. When unexpected responses arise, researchers can include an “Other” category as a catch-all, but this should be used sparingly and documented clearly.

It is also important that codes are assigned consistently. The Food Safety Institute recommends preserving as much detail as possible during coding, rather than immediately collapsing responses into overly broad categories – doing so early can cause researchers to lose nuance that becomes important during analysis.

Qualitative coding: open, axial, and selective 🔗

In qualitative research, coding goes through progressive stages. Open, axial, and selective coding – developed within grounded theory methodology by Strauss and Corbin – represent a structured approach to extracting meaning from non-numerical data such as interview transcripts or field notes.

Open coding is the starting point. The researcher reads through the data line by line, identifying concepts and assigning preliminary labels. These initial codes are loose and tentative, meant to break the data into its smallest meaningful components. According to The Craft of Sociological Research, once a short list of codes has been refined, each should have a clear definition – often recorded in a codebook – to ensure consistent application across the entire dataset.

Axial coding follows. Here, the researcher examines how the open codes relate to each other, grouping them into broader categories and subcategories. The process moves from description toward explanation – researchers look for causal relationships, patterns, and connections between themes that weren’t visible when codes were examined individually.

Selective coding is the final stage. The researcher identifies a single core category or central theme that integrates all the others into a cohesive narrative or theory. As described by Quirkos, this is where the researcher moves toward answering the original research question with the full weight of the organized, coded data behind them.

Intercoder reliability 🔗

When multiple researchers are involved in coding the same dataset, consistency becomes a concern. Intercoder reliability refers to the degree of agreement between different coders when they independently assign codes to the same data. Achieving high intercoder reliability requires a detailed codebook, pilot testing of the coding scheme on a small data sample, and regular calibration discussions between coders. Without it, the same response could be categorized differently by different researchers, undermining the reliability of the entire analysis.

Step three: data transcribing 🔗

Transcription is the process of converting non-digital or non-textual data into a written, accessible format. It is the final preparatory step before analysis begins, and it applies most directly to audio recordings, video interviews, focus group sessions, and handwritten field notes. As the University of Illinois Library Guide on qualitative data analysis notes, while transcription is often treated as part of data collection, it is also an act of analysis – the choices made during transcription shape the data that will later be analyzed.

Types of transcription 🔗

ATLAS.ti’s guide to research transcripts identifies three main types of transcription that researchers choose between based on their analytical needs:

• Verbatim transcription captures every word spoken, including filler words (“um,” “uh”), false starts, repetitions, pauses, laughter, and other non-verbal cues. This is used when the manner of speaking matters as much as the content – for example, in discourse analysis or studies of emotional responses.
• Clean verbatim transcription records every word but removes fillers and stutters, producing a more readable transcript. It is suitable when the focus is on content rather than speaking style.
• Intelligent (edited) transcription goes further, reorganizing and condensing the text for clarity while preserving the core meaning. This is used when readability is prioritized over exactness of wording.

Accuracy and best practices in transcribing 🔗

Transcription is time-intensive. Research published on BCcampus notes that a one-hour interview can take six to seven hours to transcribe, depending on the transcriber’s experience, the audio quality, and the tools used. Researchers can use automated transcription software to speed up the process, but no software is completely accurate – human review and editing are always necessary, especially when audio quality is poor or speakers use technical terminology.

For accuracy, researchers are advised to transcribe in short segments rather than long stretches, mark inaudible sections clearly (e.g., with a blank line) rather than guessing at words, label non-verbal cues like laughter or pauses using standard notation, and begin a new paragraph with each new speaker. Insight7’s guide to qualitative data analysis also stresses that confidentiality must be maintained throughout – participants should be anonymized during transcription using pseudonyms or codes, and all files should be stored securely.

It is also worth noting that once a transcript is complete, it feeds directly into the coding process. SpeakWrite’s transcription guide notes that coding is typically iterative, requiring multiple passes through the data to refine codes and themes – meaning the quality of the transcript directly affects the quality of the coding that follows.

Accuracy, consistency, and homogeneity: the three pillars of good data presentation 🔗

Running through all three steps – editing, coding, and transcribing – are three core requirements that any researcher must keep in mind.

Accuracy means that the data reflects what respondents actually said or reported, without distortion from errors, guesswork, or careless transcription. Every edit, code, and transcript should be cross-checked rather than assumed to be correct on the first pass.

Consistency means applying the same standards throughout. The same type of response should always be edited the same way, coded into the same category, and transcribed using the same notation. Inconsistency at any stage introduces noise that can distort the final analysis. This is why codebooks, editing guidelines, and transcription protocols all exist – to enforce consistency across an entire dataset.

Homogeneity refers to the grouping of data that shares common characteristics so that comparisons are meaningful. In coding especially, grouping unlike responses together – or splitting similar ones across different categories – can make patterns invisible or create false ones. Well-designed coding categories preserve the natural groupings in the data.

Researchers who document every decision made during editing, coding, and transcribing create a transparent audit trail that allows others to evaluate, replicate, or build on their work. This documentation is not optional – it is what separates research that can be trusted from research that cannot.

The relationship between these three steps 🔗

Editing, coding, and transcribing are not independent tasks – they form a sequence in which each step depends on the quality of the one before it. Poorly edited data produces unreliable codes. Poorly transcribed interviews produce inaccurate text that is then coded incorrectly. The errors compound at each stage. Conversely, when these three steps are carried out carefully and systematically, researchers arrive at a clean, structured dataset that is genuinely ready for analysis – and for meaningful presentation to an audience.

Whether a researcher is working with quantitative survey data or qualitative interview material, these preparatory steps are what allow the data to speak clearly and honestly. They are not bureaucratic hurdles between collection and analysis – they are the process by which raw information becomes research evidence.

What do you think? How might inconsistent coding or rushed transcription distort the conclusions of a research study – and at what point in the process do you think errors are most likely to go unnoticed? If you were designing a multi-researcher project, what safeguards would you put in place to ensure consistency across the editing, coding, and transcription stages?

References

1. https://ebooks.inflibnet.ac.in/hsp16/chapter/processing-operation-editing-coding-classification/
2. https://www.mbaknol.com/research-methodology/methods-of-data-processing-in-research/
3. https://en.wikipedia.org/wiki/Data_editing
4. https://foodsafety.institute/research-methodology/validity-editing-coding-data-collection/
5. https://www150.statcan.gc.ca/n1/edu/power-pouvoir/ch3/editing-edition/5214781-eng.htm
6. https://delvetool.com/blog/openaxialselective
7. https://viva.pressbooks.pub/sociology-research-methods/chapter/11-1-coding-qualitative-data/
8. https://www.quirkos.com/blog/post/open-and-axial-coding-qualitative-software/
9. https://insight7.io/systematic-coding-in-qualitative-research-how-to-do-it-right/
10. https://guides.library.illinois.edu/qualitative/transcription
11. https://atlasti.com/guides/qualitative-research-guide-part-2/research-transcripts
12. https://pressbooks.bccampus.ca/undergradresearch/chapter/transcribing-and-coding/
13. https://insight7.io/how-to-transcribe-and-code-data-in-qualitative-research/
14. https://speakwrite.com/blog/data-transcription/