Vichara QDAUser Guide
Academy Open the Workbench

Vichara QDA: User Guide

A quick, practical guide to running a qualitative study end-to-end. Vichara supports the full workflow: set up your study, import and code data, build themes, run analyses, and export a report.

1. What you get, and what it costs

Create an account and you start on a 14-day free trial with everything switched on. After that, a subscription is ₹999 a month, and that is the price everywhere; there is no separate overseas price. Any international Visa or Mastercard, debit or credit, can pay it, and you are charged in your own currency: your bank converts the rupee amount, so someone in London is debited the equivalent in pounds. There is no annual lock-in and nothing renews automatically — you are never charged without pressing the button yourself.

If you stop paying, you do not lose your work. Your projects live in your own browser. Coding, the codebook, themes, memos, every analysis and every export keep working exactly as they did before. What stops is saving to the server, sharing with collaborators, Manan, and starting a new project. Finish the study you are in the middle of; pay again when you start the next one.

Credits are a separate thing

A subscription buys the software. Credits are used by Manan, the optional assistant, and by nothing else — coding, the codebook, themes, the analyses and all exports cost nothing and never touch your balance. Every account gets 150 free credits on sign-up. Top-ups are 100 credits for ₹200, with larger packs slightly cheaper per credit. Credits do not expire. One credit is about a thousand words of text read or written, and your balance is shown in the sidebar.

Students

Students get 50% off. Open Account → Plan and send a request with your institution, course and year, plus either an institutional email address or a link to a student ID or enrolment letter. A person reads every one of these and you hear back either way by email.

Your account and your projects

Projects belong to the account that made them. Signing in as someone else on the same computer shows that person's projects, not yours. Your plan and the days remaining are shown in the header beside your name, not buried in a menu.

2. Set up your study

  • Create a project (sidebar → + New). Enter your title, aim, objectives and data type (IDI, FGD, etc.). This frames your work and appears in reports. Edit anytime via Documents → ✎ Edit details.
  • Invite collaborators as you create it — enter their email addresses in the new-project dialog and the project is saved to the server and shared with them straight away. You can also do this later from Cloud & Sharing.
  • Use multiple projects for multiple studies; switch between them at the top of the sidebar.
  • The project checklist on the dashboard tracks what is done and what is still missing — aim and objectives, documents, a codebook, coded passages, themes — so you can see where the study actually is.

3. Manan: your coding assistant

Manan (मनन, “reflection”) is an optional assistant that reads passages and proposes codes from your codebook. It sits in the sidebar directly after Project set-up.

Three things worth being clear about before you use it.

It is optional. Every part of this workbench works without it. Coding every passage by hand is the most defensible analysis of all, and plenty of people do exactly that.

It suggests; you decide. Nothing Manan proposes is written into your data. Every suggestion lands in Review, where you accept or reject it one at a time.

You are told how well it does before you rely on it. Its agreement with your own coding is measured and shown as a number, not claimed.

Train it on your study first

Rather than turning a general-purpose model loose on your transcripts, Manan is specialised on your project before it codes anything. Fill in your title, aim and objectives on Project set-up and have a codebook with codes in it, then press Train Manan on this project. Training builds a brief — and you can read and edit every word of it:

  • a coder persona: who it is being asked to be, in the context of your study;
  • this study's vocabulary: the terms, local words and abbreviations that appear in your data;
  • a coding rubric derived from your own code definitions and inclusion/exclusion criteria;
  • practice cases it writes for itself, which it is then marked on.

No model weights are changed and nothing is carried between projects. Training produces instructions for this study, which Manan then follows on every request.

Calibration: two different numbers, and only one belongs in a paper

  • The self-test runs at the end of training on the invented practice cases. It measures whether Manan applies its own brief consistently. It can run before you have coded anything, which is its whole use — but it is not evidence about your data.
  • Check Manan against your own coding is the real test. Manan re-codes 24 passages that you have already coded, without being shown your answers, and the two codings are compared with Cohen's κ and Krippendorff's α — the same statistics you would use between two human coders. This is the number to quote.

If κ comes back low, read the practice cases before blaming the assistant. The confusions it makes are usually the ones a new human coder would make, and they point at the codebook: codes that overlap in meaning, or definitions that restate the code's own name instead of saying when to apply it.

Agreement depends mostly on how many codes are in play. On a real study we measured κ around 0.6 with 10–30 codes in scope, and close to zero with 150 or more — no human coder does well picking from four hundred either. If your codebook is large, set Level of detail → Top-level codes only so Manan sorts passages into parent codes and you do the fine distinctions yourself. That division of labour is both more accurate and more honest.

What you can ask it to do

  • Auto-code with Manan (Coding view) reads the passages you have not coded and proposes codes from your codebook. It can work through one document or the whole project. Everything goes to Review.
  • Draft codebook with Manan (Codebook view) proposes a starting codebook from your protocol and your transcripts, with definitions and criteria you then edit. Useful when you are staring at a blank codebook; never a substitute for reading your data.
  • Ask Manan about this project answers questions about what you have gathered so far.

What leaves your computer, and when

Two different things are worth separating here.

  • The analyses — frequency, co-occurrence, retrieval, sentiment, κ, the matrices, the write-up draft — are computed on the Vichara server. The text is sent, calculated in memory and discarded. It is not stored and no outside company sees it. This is ordinary arithmetic on your data, not a model.
  • Manan is the only part that sends your text to a language-model provider outside Vichara, and it sends only the passages needed for the request you pressed — never your whole project, and never anything at all unless you press something. Those passages are processed and discarded, and are not used to train anything.

So if your ethics approval does not allow participant text to reach a third party, do not use Manan. Everything else in the workbench is unaffected, and a study coded entirely by hand is the stronger one to defend anyway.

4. Bring in your data

  • Transcription template (Documents view) gives you a clean Word template with speaker turns on separate lines. Transcribing in this format imports perfectly. Avoid putting transcripts inside Word tables.
  • Import .docx or .txt files, or Paste text directly.
  • Case attributes (Documents view): define variables such as Gender, Age group or Site, then tag each document. These power subgroup comparisons.

The Transcriber

The Transcriber is a workspace for turning a recording into a transcript. Load an audio or video file, or record straight from the microphone, then play it back at your own speed while you type: speaker labels, renaming speakers, timestamps and export into a Vichara project are all built in. This part works for everyone and needs nothing installed.

Automatic speech-to-text is a separate matter. It runs on a speech engine on your own computer (port 8000), not on the Vichara server, so it is available only if you are running that engine locally. The page tells you which situation you are in before you start — “Speech engine connected” or “not detected, manual transcription available” — rather than letting you upload an hour of audio and then fail. If you need a transcript today, type it here; nothing else in Vichara depends on the engine.

5. Code your data (Coding view)

  • Select a passage in the transcript, then click a code in the pop-up or the right-hand panel to apply it.
  • In-vivo coding creates a code named after the participant's own words.
  • Passages can carry several codes (overlaps are shown). Click a highlight to review or remove its codes.
  • Manage codes in Codebook: hierarchy (parent and child codes), colours, merge and rename.
  • Each code carries a definition, inclusion criteria ("apply when…"), exclusion criteria ("do not apply when…"), an anchor example, its origin (inductive / deductive / in-vivo) and a sentiment. Record your analytic framework and coding approach under "Framework & details" so the codebook is publication-ready.
  • Import / export the codebook: Word (for a thesis appendix), Excel/CSV (editable, and the import format: download the template), or REFI-QDA .qdc, the open standard that NVivo, ATLAS.ti and MAXQDA read and write. Imports show a preview with merge options before anything changes.

6. Thematic analysis: Braun & Clarke (Themes view)

Vichara follows the six phases of reflexive thematic analysis. Open Themes → Show 6-phase guide to track your progress:

  1. Familiarisation: read and re-read; note first impressions as Memos.
  2. Initial codes: code systematically across the data (Coding view).
  3. Searching for themes: create themes here and group related codes under each.
  4. Reviewing themes: test them with Retrieval and Co-occurrence; do they hold together?
  5. Defining & naming: write a clear definition for each theme.
  6. Producing the report: pull vivid extracts and export (below).

Approach & framework — inductive, deductive or hybrid

Open Themes → Approach & framework and record how you are deriving themes. This is written into your thematic-analysis report as part of the audit trail.

  • Inductive: themes built bottom-up from the data.
  • Deductive: themes come top-down from an established framework.
  • Hybrid: a framework as scaffold, plus inductive themes for what it misses.

For deductive or hybrid work, choose one of the built-in frameworks — CFIR, TDF, COM-B, socio-ecological model, Health Belief Model, RE-AIM, NPT, WHO health system building blocks, Levesque access to care or Andersen's behavioural model — and press Apply as themes. Vichara creates one theme per domain with the domain definition as its coding rule, and each framework carries its original citation for your methods section.

A coverage table then shows how many codes and segments sit under each domain, and flags domains with no evidence yet. An empty domain is a finding worth reporting, not a failure. Download it with Coverage matrix (CSV).

Thematic analysis report

Themes → ⬇ TA report produces a standalone thematic-analysis output: your approach and framework, the six-phase audit trail with the notes you wrote for each phase, every theme with its definition and origin (inductive or a-priori), its codes, segment and document counts, illustrative extracts, and framework coverage. Choose how many quotes per theme, then download as Word, or take the theme × code matrix as CSV. The same downloads sit under Export & Backup → Thematic analysis.

7. Analyse (Analysis view) & how to interpret

  • Code frequency: how often each code appears, and in how many documents. High counts show salient topics; interpret with care: frequency ≠ importance.
  • Co-occurrence: how often two codes are applied to the same passage. Strong overlaps suggest related concepts or candidate themes.
  • Retrieval: pull every quote for a code, optionally filtered by attribute (e.g., only women). This is your evidence base for writing up.
  • Text search: find a word/phrase across all transcripts with surrounding context.
  • Word frequency: the most common words (stop-words removed) across the corpus or within one code: a quick sense of language and salience.
  • Matrix query: codes × an attribute (e.g., codes × Gender): compare how sub-groups talk about each code.
  • Framework matrix: the classic cases × themes grid of framework analysis (Ritchie & Spencer). Each cell shows how many extracts that case contributed to that theme; click it to read them and write your own summary, or press Auto-generate (single cell) or Auto-summarise cells (whole matrix) to draft them from that case's own words. Auto-drafts are flagged draft until you edit and save them, so you always review before use. Comparing summaries across rows is how framework analysis builds its findings. Download as Word or CSV.
  • Sentiment & emotion: the valence of each coded extract on a five-point scale (−− very negative, − negative, 0 neutral, + positive, ++ very positive), with negation handled (so "not helpful" reads negative), plus an emotion profile (fear, anger, sadness, hope, shame/stigma, trust) and your strongest extracts. Treat these as suggestions to review, not verdicts: set the authoritative sentiment yourself on each code.
  • Evidence table: your quotes laid out as theme → code → supporting extracts with the source and case attributes, ready to paste into a manuscript. Download as Word or CSV/Excel. While coding, click a highlighted passage and press ☆ to mark a key quote, then tick "key quotes only" to build a table of just your best extracts.
  • Word tree (KWIC): enter a word and see how participants talk around it: a branching tree of the most common continuations, plus a full keyword-in-context concordance you can export.
  • Code map: a network diagram where each circle is a code (sized by how often it's used) and lines connect codes applied to the same passages — it reveals which ideas cluster together. Downloadable for publication.
  • Analyse by Codes or Themes: on frequency, co-occurrence, code map and matrix, switch the toggle to roll your codes up into their themes for a higher-level view.
  • Reliability (κ): inter-rater reliability, in three steps. (1) Press Coder pack — this exports your documents and codebook with none of your coding, so your second coder is not primed by seeing where you already coded. (2) They import it (Export & Backup → Import project), code independently, and export their project. (3) Load their file here. You get Cohen's kappa per code and overall with Landis & Koch strength bands, and a reconciliation list of the exact passages you disagreed on — work through those together and write the rule that settles each one into the code's inclusion or exclusion criteria. Both the κ table and the disagreements export as CSV, and the run is recorded in your audit trail.
  • Every chart offers PNG (hi-res) and SVG (vector) download for publication.

8. Write-up: draft your Methods and Results

The Write-up view drafts the two hardest sections of your paper from the project you have already coded. It is a first draft to revise, not a finished manuscript.

  1. Open Write-up → Write-up details… and fill in the facts a Methods section needs that are not in your coded data: sampling method and rationale, setting, eligibility criteria, data collection period, transcription, number of coders, ethics approval, trustworthiness strategies, participant demographics (which becomes Table 1) and your Cohen's kappa if you ran one.
  2. Press Generate draft.

You get a subheaded Methods (study design · participants and setting · data collection · data analysis · rigour and trustworthiness · ethical considerations) and Results (participant characteristics · overview of themes · one subsection per theme), plus Table 1 participant characteristics, Table 2 theme frequency, Table 3 illustrative extracts, charts of theme coverage, and a reference list containing only the sources your recorded choices actually call for.

It will not invent anything. Every figure, quote and citation traces back to your project or to something you typed. A fact it needs but does not have appears as [[MISSING: …]] in the draft and is listed as a prompt above it. Fill those in and regenerate.

Download as Word or save as PDF. Read and revise before submission, and check each number against your coded data.

9. Export, share and back up

  • Report as Word (.doc) or PDF (editable / print-ready): includes study overview, themes, codebook, frequencies and all coded extracts.
  • Thematic analysis: the TA report (Word), theme × code matrix (CSV) and framework coverage (CSV).
  • CSV exports of coded segments and codebook open in Excel, SPSS or R.
  • Project backup (.json): export regularly. Projects live in this browser unless saved to the cloud; the backup lets you restore, move computers, or share with a colleague.

What “the cloud” actually means here

By default your work is stored in your own browser, on your own computer. Nothing leaves it. That is why Vichara works offline, and why a lapsed subscription cannot take your coding away — it was never on anyone's server to withhold.

Save to the cloud means: also keep a copy on the Vichara server. There are three reasons to want that, and one reason not to.

  • Another device. Your office computer can open the same project.
  • Sharing. A colleague can only work on a project that exists somewhere you both reach.
  • Survival. Clearing your browser data, or a dead laptop, takes a browser-only project with it.
  • Against it: the transcripts are then sitting on a server. For sensitive work — identifiable participants, children, clinical records — a browser-only project, backed up to your own encrypted drive as a .json, is the stronger answer to an ethics committee.

It is a per-project choice, not a setting. Keep the sensitive study local and put the rest on the server if you like.

Sharing is per project

Save a project to the cloud from Cloud & Sharing — or invite people when you first create it — then invite more by email or with a link. Access applies to that one project only — your other projects stay private, and each is shared separately. The share dialog names the project you are sharing so there is no ambiguity. Collaborators can be editors or viewers.

Deleting a project

Each project card on the dashboard has a delete control in its footer. What it does depends on the project: a cloud project you own is deleted for everyone it is shared with; one shared with you only removes your access; a browser-only project is removed from this device. You are always asked to confirm — export a backup first if you may want it later.

10. Learn the method: Vichara Academy

The Academy (sidebar, opens in a new tab) is a free ten-module course in qualitative research: foundations, choosing a design, sampling, interviewing and focus groups, coding and the codebook, thematic analysis, framework analysis, rigour and reliability, ethics, and reporting. Each module has worked examples, diagrams and flowcharts, a quiz you need to pass, and an assignment you complete in your own project.

Sign in to track progress. Complete every module and you can download a certificate in your name — a PDF carrying a certificate ID and a QR code that anyone can scan to verify it at /verify. Quizzes are graded on the server; assignments you mark yourself.

Good practice

  • Cite the software in your methods, and name the version you used. Ready-made references in APA, Vancouver, MLA, BibTeX and RIS are on the How to cite page, along with a methods sentence you can adapt.
  • Write Memos as ideas emerge: they become your audit trail.
  • Keep coding rules in each code's description for consistency (and inter-coder work).
  • Back up (Export project .json) at the end of every session.

Your research data (documents, codes, memos) stays in your browser unless you choose to save a project to the server, in which case it is stored there so it is backed up and can be shared with collaborators you invite. Passages sent to Manan are processed and discarded immediately, and are never used to train anything. Server backups taken for disaster recovery deliberately exclude project contents: the administrator can confirm your project exists and is being backed up, and cannot read a word of it.

Vichara QDA version 1.0 · Developed by Dr G Hari Prakash · How to cite · info@vicharaqda.in

© 2026 Vichara QDA. All rights reserved.