Overview
Researchers often use 'content analysis' and 'discourse analysis' interchangeably, but they rest on different assumptions. Quantitative content analysis, in the tradition of Krippendorff and Neuendorf, treats texts as data to be coded systematically and reliably, so it needs a codebook, unit definitions and an agreement statistic. Qualitative content analysis, as described by Mayring, Schreier or Hsieh and Shannon, is a systematic but interpretive way of reducing and categorizing meaning. Discourse analysis, in its many forms, treats language as social action and focuses on how meaning, identity and power are produced.
We start by matching the method to your research question and corpus: a study counting how often newspapers frame migration as a security issue needs something very different from a study of how a government speech legitimizes austerity. Then we help you build the sampling plan, coding frame or analytic framework, pilot it and document it.
For content analysis we can run reliability testing and descriptive or inferential statistics on your coded data. For discourse analysis we work as a methodological sparring partner, reviewing your analysis of sample texts and helping you connect textual detail to social context without overreaching.
Who it’s for and the problems it solves
This is for you if you are…
- Students in media, communication, politics, education and business analyzing documents, news or social media
- Doctoral researchers applying critical discourse analysis and needing their framework to be explicit
- Health and policy researchers coding documents or open-text responses with a codebook
- Research teams that need an intercoder reliability procedure before coding at scale
Common problems we help with
- You are unsure whether your study is content analysis, thematic analysis or discourse analysis
- Your coding frame has overlapping categories and coders keep disagreeing
- A reviewer asked for Krippendorff's alpha and you do not know how to calculate or interpret it
- Your CDA reads as opinion because the link between linguistic features and wider context is not shown
- You have thousands of posts or articles and need a defensible sampling strategy
What’s included — and what isn’t
Included
- Method selection: quantitative content analysis, qualitative content analysis or a named discourse-analytic approach
- Corpus definition and sampling: time frame, sources, search strings, units of analysis and inclusion rules
- Codebook development with category definitions, decision rules, examples and borderline cases
- Pilot coding, coder training materials and intercoder reliability testing (Krippendorff's alpha, Cohen's kappa, percent agreement with caveats)
- Statistical analysis of coded content: frequencies, cross-tabulations, chi-square or regression where appropriate
- For discourse analysis: framework guidance (Fairclough's three dimensions, van Dijk's socio-cognitive approach, Wodak's discourse-historical approach, Foucauldian or discursive psychology traditions)
- Feedback on your analysis of sample texts and on how you evidence claims with textual features
- CAQDAS setup for document coding in NVivo, ATLAS.ti or MAXQDA
Not included
- Producing the interpretive analysis of your texts for you to submit as your own
- Adjusting codes after the fact to inflate reliability scores
- Collecting social media data in ways that breach platform terms or your ethics approval
- Legal advice on copyright or data protection for your corpus; we flag issues so you can check with your institution
What you receive
- Method and sampling memo
- A concise rationale for your chosen approach, corpus boundaries, sampling procedure and unit of analysis that you can adapt for your methods chapter.
- Codebook
- A structured coding frame with definitions, inclusion and exclusion rules and anchor examples, refined after pilot coding.
- Reliability report
- Agreement statistics per variable with confidence intervals where available, notes on problem categories and the revisions made.
- Analysis outputs
- For content analysis: tables, charts and test results with annotated syntax. For discourse analysis: written feedback on your sample analyses.
- Walkthrough session
- A live explanation of the results or framework so you can present and defend your choices.
How it works
Clarify the question
We confirm what you want to know about your texts and which analytic tradition answers it.
Build the frame
We agree the corpus and sampling, then draft a codebook or analytic framework with you.
Pilot and test
Coding a pilot subset exposes vague categories. For content analysis we calculate agreement and revise until it is acceptable for your purpose.
Analyze and review
You or your team code the full sample; we run statistics or review your discourse analysis and return feedback and outputs.
Traditions, frameworks and tools
- Quantitative content analysis
- Systematic, replicable coding of manifest content with predefined categories, following Krippendorff and Neuendorf, reported with reliability coefficients.
- Qualitative content analysis
- Directed, conventional or summative variants (Hsieh and Shannon) and Mayring's or Schreier's category-building procedures for interpretive but rule-guided reduction of meaning.
- Krippendorff's alpha
- An agreement coefficient that handles multiple coders, missing data and nominal to ratio data; commonly preferred over percent agreement, which ignores chance.
- Fairclough's critical discourse analysis
- Three linked dimensions: text (vocabulary, grammar, cohesion), discursive practice (production, distribution, interpretation) and social practice (ideology and power).
- Other discourse approaches
- Van Dijk's socio-cognitive model, Wodak's discourse-historical approach, Foucauldian discourse analysis, discursive psychology and multimodal analysis of images and layout.
- Software
- NVivo, ATLAS.ti or MAXQDA for coding; R or Python for reliability statistics and corpus counts; Excel for small codebooks.
Content, thematic or discourse analysis?
A simple test is to ask what kind of claim you want to make. If you want to say how often something occurs and compare across sources or years, you need content analysis with a reliable codebook. If you want to describe patterns of meaning across participants, thematic analysis may fit. If you want to show how language constructs identities, legitimizes policies or reproduces power, you are doing discourse analysis.
- 'How often did tabloids use crisis framing in 2020 to 2024?' → quantitative content analysis
- 'What concerns do patients raise in complaint letters?' → qualitative content analysis or codebook TA
- 'How does the minister's speech position teachers as responsible for failure?' → critical discourse analysis
Making reliability meaningful
Reliability only means something if categories are clearly defined and coders are trained. We help you test on a subsample coded independently, report alpha per variable rather than one averaged figure, and explain what you did with categories that fell short: redefining, merging or reporting them as exploratory. Commonly cited benchmarks are a guide, not a rule; the acceptable level depends on how the codes will be used.
Keeping critical discourse analysis rigorous
CDA is sometimes dismissed as the analyst finding the ideology they expected. The defense is transparency: explain why texts were selected, show the linguistic evidence (modality, nominalization, passive agency, presupposition, intertextuality) and connect it explicitly to context. We review your sample analyses with that standard in mind and help you acknowledge the limits of your interpretation.
What the work can look like
Illustrative example — fictional data
Illustrative codebook entry and reliability note
A fictional master's project coded 600 online news articles about electric vehicles for dominant frame. The category and figures below are invented to show how a codebook entry and reliability result are documented.
Codebook entry: ECONOMIC_OPPORTUNITY
Definition: the article presents EVs primarily in terms of jobs, investment, growth or competitiveness. Include: headlines or leads on factories, exports, stock performance. Exclude: cost-of-ownership for consumers (code CONSUMER_COST). Borderline: government subsidy stories are coded here only if framed as industrial policy.
Reliability result
Two coders independently coded a random subsample of 90 articles. Krippendorff's alpha for the frame variable was 0.71 on the first round; after clarifying the subsidy rule it rose to 0.83 on a fresh subsample. The write-up reports both rounds and the rule change.
Illustrative only: the project, category and numbers are invented to show documentation style, not a real client's data.
What you’ll need to provide
- Your research questions and any theoretical framework you are committed to
- The corpus or a description of where it comes from, with access details
- Any draft coding frame, existing codebook or prior studies you want to follow
- Coder availability if reliability testing with a second coder is planned
- Ethics approval details for online or sensitive texts
You don’t need everything ready to ask for a quote — tell us what you have and we’ll say what’s missing.
What affects price and turnaround
Price depends on
- Size and variety of the corpus and whether sampling must be designed from scratch
- Number of variables or categories in the codebook
- Whether reliability testing with multiple coders is required
- Statistical analysis of coded data versus framework feedback only
- Languages involved and whether texts need translation before analysis
Turnaround depends on
- Access to the full corpus in a usable format
- Number of codebook revision and pilot rounds
- Availability of second coders for reliability testing
Every project is quoted individually. How we price our work.
Questions people ask before starting
Do I need a second coder?
For quantitative content analysis, usually yes, because reliability is part of the method's claim. For qualitative content analysis or discourse analysis it depends on the tradition and your department. We help you decide and describe the choice.
Can you collect the articles or social media posts?
We can help design search strings, sampling and data-collection procedures, and advise on database exports. Collection must follow platform terms, copyright rules and your ethics approval; you remain responsible for the data.
Which discourse analysis framework should I use?
It depends on your discipline and question. Fairclough's CDA suits policy and media texts where power is central; discursive psychology suits talk and interaction; the discourse-historical approach suits long-term, historically situated debates. We explain the trade-offs so you can choose.
Can you analyze texts in languages other than English?
We can support the method for many languages if you supply accurate translations or can work with us on the original texts. Tell us the languages at the start so we can confirm what is realistic.
Will you write my analysis chapter?
No. We provide codebooks, statistics, feedback and explanation; you write the chapter. We can review your draft for accuracy and coherence with your method.