Quantifying qualitative data means converting coded themes, categories, and observations into counts, frequencies, and cross-tabulations that can be reported numerically and compared across subgroups. Common techniques include frequency coding (how often a theme appears), presence/absence matrices, intensity ratings, and cross-tabulation by participant metadata. These approaches sit at the heart of mixed methods research, where qualitative and quantitative findings complement each other.
Why would you quantify qualitative data?
Qualitative data is rich by nature: interview transcripts, field notes, open-ended survey responses, and documents contain nuance, contradiction, and context that numbers alone cannot capture. So why count things at all?
Three situations make quantification worth doing. First, when you need to communicate findings to a sceptical audience. Saying "10 of 18 participants mentioned price as a barrier" lands differently with a leadership team than "several participants mentioned price." Second, when you want to compare subgroups: do junior employees express different concerns from senior ones? Does the pattern differ between customers who churned and those who renewed? Third, when you are integrating qualitative findings into a broader study that includes quantitative data.
The risk is doing it badly. Quantifying qualitative data too aggressively collapses the richness that made qualitative research worth doing. The goal is to use numbers to clarify and communicate, not to pretend that themes are the same as survey response categories.
What are the 5 main techniques for quantifying qualitative data?
1. Frequency coding
Frequency coding counts how often each theme or code appears across your dataset. It is the most straightforward form of quantification and the most common.
The basic output is a count per theme: "Trust in AI tools was mentioned by 14 of 22 participants (64%). Concerns about data privacy were raised by 11 (50%). Speed of analysis was cited as the primary motivation by 17 (77%)."
You can report frequencies in several ways:
- Count of documents/participants mentioning a theme
- Count of total mentions (a single document can mention a theme multiple times)
- Percentage of participants who raised a theme
- Rank order of themes by frequency
When it works well: summarising findings for stakeholders who are used to percentages and counts, comparing theme prevalence across two waves of research, reporting on large-scale surveys with open-ended components.
Watch the denominator. A theme mentioned by 3 of 5 participants (60%) sounds very different from 3 of 50 participants (6%). Always report the base. And remember that frequency does not equal importance: a theme mentioned by only 2 participants may be analytically more significant than one mentioned by 15, if those 2 participants represent a key segment.
2. Presence/absence matrices
A presence/absence matrix codes each theme as either present (1) or absent (0) for each document or participant. The result is a data table with one row per document and one column per theme.
This creates a form of structured data that supports further analysis: you can calculate the percentage of participants who showed each theme, identify co-occurring themes, cluster documents by theme pattern, and feed the matrix into quantitative analysis tools.
| Participant | Theme A | Theme B | Theme C | Theme D |
|---|---|---|---|---|
| P1 | 1 | 1 | 0 | 1 |
| P2 | 0 | 1 | 1 | 0 |
| P3 | 1 | 0 | 0 | 1 |
The matrix approach works well in content analysis, where you are coding a defined set of categories across a large number of documents. It is less suited to grounded theory or interpretive phenomenological analysis, where themes are not predefined.
3. Cross-tabulation by participant metadata
Cross-tabulation connects theme occurrence to participant or document characteristics: age group, department, tenure, country, product used, customer segment. This is where quantified qualitative data becomes analytically powerful.
A simple example: in an exit interview study, you might find that concerns about career growth were mentioned by 80% of employees under 35, compared with 30% of those over 45. That pattern would not appear from reading the transcripts in sequence. It requires systematic cross-tabulation.
The mechanics are the same as a standard cross-tab: theme presence on one axis, participant characteristic on the other, cell counts or percentages in the cells. For large datasets, you can apply chi-square tests to assess whether the distribution is statistically different from chance.
Skimle's Statistics View handles this automatically. Once you have added metadata variables to your documents (role, segment, date, country, or custom fields), the platform cross-tabulates theme occurrence by those variables and flags statistically significant differences. The metadata analysis guide explains how to set this up and interpret the output.
4. Intensity and sentiment coding
Rather than a binary (present/absent), intensity coding assigns a scale to each theme occurrence: positive/neutral/negative, or a 1-to-5 scale indicating how strongly a theme was expressed.
Sentiment coding is the most common form: each coded excerpt is rated as positive, neutral, or negative in tone. Aggregated across participants and cross-tabulated by metadata, sentiment coding produces structured output that quantitative researchers can work with directly.
More granular intensity coding requires more judgement and more time to apply reliably. A codebook specifying what counts as "strong" versus "moderate" expression of a theme, combined with inter-rater reliability checks, is essential for defensible results.
Use case: a product team analysing 400 customer support tickets might code each ticket for the feature mentioned and the sentiment expressed, producing a feature-by-sentiment matrix that product managers can use to prioritise fixes.
5. Saturation tracking as a quantitative signal
Tracking the rate at which new codes appear as you move through your dataset provides a quantitative signal for theoretical saturation. If you code your first 10 interviews and identify 40 distinct codes, then analyse 10 more and add only 4, then 10 more and add 1, you have quantitative evidence that saturation has been reached.
This is useful for academic write-ups, where reviewers increasingly ask for empirical evidence of saturation rather than the assertion "analysis continued until no new themes emerged." It is also useful for consulting projects, where a client might ask whether interviewing more people would change the findings.
The measure is simple: track new code appearances per batch of documents. When the curve flattens, saturation is near.
When should you NOT quantify qualitative data?
Quantification adds value in specific circumstances. It adds noise in others.
Avoid quantifying when your sample is very small. Reporting "4 of 6 participants mentioned X (67%)" implies statistical precision that a sample of 6 cannot support. For small, purposive samples, report patterns in qualitative terms: "Most participants mentioned X, with two notable exceptions who..."
Avoid when the richness is the point. A phenomenological study exploring how five nurses experience moral distress should not be converted into a theme frequency table. The value is in the texture of the experience, not its prevalence across a small purposive sample.
Avoid treating co-occurrence as causation. Two themes appearing together in many documents does not mean one causes the other. Qualitative data is not structured for causal inference; quantification does not change that.
The honest framing: quantification supports communication and comparison. It does not make qualitative research more rigorous, because rigour in qualitative research comes from analytical depth, transparency, and appropriate methods.
How does this work in a mixed methods study?
Recommended reading
Mixed methods research: 4 designs, examples and when to use it
Mixed methods research integrates qualitative and quantitative data in a single study. Quantified qualitative data is often the bridge between the two.
The four main integration patterns are:
QUAL → QUAN (sequential exploratory): Qualitative research first identifies themes and constructs, which are then operationalised into survey items for quantitative testing. The coded categories from your qualitative phase become the variables in your quantitative instrument.
QUAN → QUAL (sequential explanatory): Quantitative results identify a pattern or anomaly, and qualitative research explains why. You might find that customer satisfaction drops in one segment and then conduct interviews specifically with that segment to understand the cause.
Concurrent triangulation: Both strands run simultaneously and are integrated at the interpretation stage. Frequency tables from the qualitative strand are compared with survey data; convergence or divergence tells you something about the stability of your findings.
Embedded design: One strand (typically quantitative) is nested inside the other. For example, a large survey might include open-ended questions whose analysis produces a frequency table embedded in the quantitative results.
The mixed methods research guide covers these designs in detail, including sample sizes and timing considerations.
Practical example: quantifying 60 customer discovery interviews
A product team conducts 60 interviews across three customer segments (enterprise, mid-market, and SMB). The interviews are coded using an inductive approach, producing 28 distinct codes grouped into 7 themes.
Step 1: Frequency coding. Count how many participants mentioned each theme. Theme 3 (integration complexity) was mentioned by 47 of 60 participants (78%). Theme 7 (pricing clarity) was mentioned by only 12 (20%).
Step 2: Cross-tabulation by segment. Create a segment by theme matrix. Integration complexity was mentioned by 95% of enterprise participants, 80% of mid-market participants, and only 40% of SMB participants. Pricing clarity was mentioned by 5% of enterprise, 15% of mid-market, and 52% of SMB. This reversal is the insight that drives the product roadmap.
Step 3: Sentiment coding. For the two highest-frequency themes, code each mention as positive (feature working well), neutral (observation), or negative (pain point). Integration complexity is mentioned negatively in 85% of enterprise cases; pricing clarity is split roughly evenly between positive and negative.
Step 4: Report. "Integration complexity is the primary concern for enterprise and mid-market customers (mentioned by 88% combined), with negative sentiment in 85% of cases. Pricing clarity is uniquely salient for SMB customers (mentioned by 52%, vs 5% of enterprise), with mixed sentiment suggesting both a pain point and a growth opportunity."
This is an output that a product roadmap meeting can use. It came from qualitative interviews, but the delivery is structured enough to sit alongside quantitative conversion and retention data.
Frequently asked questions
Is quantifying qualitative data always appropriate?
No. Quantification works best when your sample is large enough to support percentages (generally at least 20 participants), when you need to compare subgroups, or when you are integrating with quantitative data. For small purposive samples in interpretive research, qualitative descriptions of patterns are more honest than percentages.
What is the difference between content analysis and thematic analysis when it comes to quantification?
Content analysis is designed for quantification from the start: it uses predefined categories applied systematically, producing frequency counts by design. Thematic analysis typically produces themes inductively and does not assume frequency coding, though you can apply frequency analysis after themes are identified. The epistemological commitments differ: content analysis is generally positivist, while thematic analysis can sit within multiple paradigms.
Can I use statistical tests on frequencies from qualitative coding?
Yes, with caution. Chi-square tests can be applied to presence/absence or frequency data from qualitative coding, testing whether theme distribution across subgroups differs from what you would expect by chance. This is legitimate when your sample is large enough (minimum expected cell count of 5 per cell), your coding is reliable (inter-rater reliability checked), and you treat the result as one signal among many rather than a definitive finding.
How does inter-rater reliability work with quantified qualitative data?
If two coders independently apply the same coding scheme to a subset of documents, you can calculate percentage agreement or Cohen's kappa to assess coding reliability. A kappa above 0.7 is generally considered acceptable for quantitative reporting. Where reliability is lower, report findings as "preliminary" or address it in your limitations section.
How does Skimle handle the quantification of qualitative data?
Skimle's Data View produces theme frequency tables automatically as part of the analysis. Once documents are coded into categories, you can see how often each theme appears, cross-tabulated by any metadata variable you have set. The metadata analysis feature tests whether subgroup differences are statistically significant and surfaces them in the category summaries, so you do not need to build the cross-tabs manually.
Ready to move from qualitative themes to structured, reportable findings? Try Skimle for free and see how the Statistics View and metadata cross-tabulation turn your coded interviews into quantified insights without losing the qualitative depth.
Related reading: Mixed methods research: 4 designs, examples and when to use each | Using metadata variables to discover patterns in qualitative data | Inductive vs deductive coding: when to use each
About the authors
Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile
Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile
Sources
- Using thematic analysis in psychology — Braun & Clarke (2006), Qualitative Research in Psychology
- Mixed Methods Research — Creswell & Plano Clark (2017), SAGE Publications
- Content Analysis: An Introduction to Its Methodology — Krippendorff (2018), SAGE
- Inter-rater reliability and Cohen's kappa — McHugh (2012), Biochemia Medica
- Grounded Theory — Glaser & Strauss (1967), Aldine de Gruyter
