Turn survey verbatims into coded variables using your own codebook or one proposed by AI. Review uncertain responses, reuse the codebook across waves, and export numeric codes with value labels to SPSS or Excel.
Read the 75-comment example and compare the initial and reviewed Excel files without signing up or uploading data. Create an account to try the workflow yourself.
A typical study leaves 3,000 to 20,000 verbatims per wave. Coding them by hand needs a team of coders, a lead adjudicating disagreements, and days of calendar time after fieldwork closes — and the codeframe still drifts between coders and between waves, while multi-mention counting varies with whoever handled the batch.
Days of coding per wave, all of it after fieldwork closes
Codeframe drift between coders and between waves
Multi-mention counts and category nets nobody can audit
Survey Coder Pro is verbatim coding software built for market research teams: it applies your codeframe to every verbatim, flags what it can't call with confidence for an analyst, and returns a file your DP team can tab.
AI applies the codebook; your team reviews uncertain responses and can check any assignment
Import an Excel codebook with code numbers, definitions and examples, or build one from the data
Multi-mention and category nets handled by an explicit, auditable rule
Reuse code numbers and definitions across waves, and review changes before comparing results
Import, code, review and export, with the codebook and assignments available to inspect
Import verbatims from .sav, Excel or CSV and select the response columns. Review empty-value markers before import so respondent text is not silently treated as missing.
Import your codeframe with categories, subcategories, definitions and examples, or generate one from the data. Every code keeps its definition — that's what makes the coding reproducible.
New tracking waves inherit the codebook, including code numbers, definitions and examples. Review emerging themes and definition changes before comparing periods.
Choose from 17 response languages and reuse a common codebook across markets. The interface is available in five languages. Check a reviewed sample in each study language.
Export numeric codes with value labels to SPSS, or download Excel. Choose coding results only, or add coded variables to a retained original .sav through the beta enriched-file workflow.
Filter responses that need review, inspect confidence and correct assigned codes. The review threshold can include low confidence, low and medium confidence, or every response.
SurveyCoder codes open-ended comments into a controlled codebook, lets your team review the assignments, and exports results to Excel or SPSS. Research agencies and CX teams can use their own codebook or review one proposed by AI, then reuse it on the next wave.
The comments and labels below are synthetic, manually prepared reference decisions. They are not customer results or an AI accuracy benchmark.
Banking · score 4
“The app works well but the transfer fee is too high.”
Reference codes: 101 — Digital experience and 102 — Fees and charges. Keep both themes; a low recommendation score does not make every part of the comment negative.
Airline · score 3
“My flight was late and nobody explained the delay.”
Reference codes: 103 — Service and communication and 104 — Waiting and delays. Separate the delay from the communication issue so both can be counted and reviewed.
Theme differences between promoters and detractors describe associations; they do not establish what caused the score. SurveyCoder supports the coding workflow; use your survey system to collect the feedback.
These CSVs illustrate a handoff, rather than the exact product export. For a pilot, measure preparation time, review time and accepted codes on your own sample. Request a coding pilot.
We processed the same 75 English retail comments and original codebook used by the welcome project in an isolated QA project. The comments are synthetic; this coding run and its native Excel exports are real.
The initial run classified 72 comments and left 3 unclassified. Codex-assisted editorial review changed 15 responses and added 7 codes. The reviewed export contains all 75 classified comments. See the original output, the reviewed output and the reason for each change.
Run on 6 October 2026 in STANDARD mode, using 75 credits. This is a workflow demonstration, not independent human validation or an accuracy benchmark. The codebook contains examples from the sample, and the dataset has no NPS scores.
A verbatim is a respondent's answer exactly as it was captured — in the field, on CATI, or in an online panel. Verbatim coding is assigning each one a code, or several, from a controlled list, so the text enters the tab run as a variable and can be banner-crossed against age, region, brand, or segment.
It's a craft with its own rules, and that's what separates it from "summarise the responses with an LLM": a coder doesn't interpret, a coder applies a frame. Fixing that frame is the work that has to happen before the first verbatim is touched.
The structure: which categories exist, which codes sit under each one, what number each code carries, and which codes are residual ("Other", "Don't know / refused" — conventionally 97/98/99). It's what gets signed off with the client and what drives the layout of the deliverable.
The codeframe plus the rules for applying it: each code's definition, examples of verbatims that fall inside and outside it, and the edge cases already adjudicated. A codeframe without definitions gets applied differently by each coder; a codebook with definitions and examples is reproducible — and it's what Survey Coder Pro keeps versioned.
Survey Coder supports response coding in 17 languages: English, Spanish, Portuguese, French, German, Italian, Dutch, Polish, Russian, Chinese, Japanese, Korean, Hindi, Indonesian, Thai, Vietnamese and Arabic. The interface is available in English, Spanish, Portuguese, German and French. Review a sample in each study language; language support is not a guarantee of equal coding quality across languages.
Use the needs-review filter to inspect flagged responses and correct their assigned codes. The manual-review threshold can include low-confidence responses, low and medium confidence, or all responses. A confidence flag is a review aid, not proof that an unflagged assignment is correct. Check a sample of unflagged responses too, and record agreed definitions before reusing the codebook.
One sentence usually touches two or three themes: "price is fine but the app keeps crashing and the branch staff were rude" is three mentions, not one. Coding multi-mention means assigning all three codes and keeping mention order, so the tables can later be read on a mentions base or a respondent base. The rule has to be explicit, because the two bases don't add up to the same percentages and the client will ask which one they're looking at.
A net is the category column: the share of respondents who mentioned at least one of the codes under it. It is not the sum of those codes — a respondent who mentioned two codes in the same category counts once in the net. Survey Coder Pro computes nets on unique respondents per category subtree, supports a three-level hierarchy (Category › Subcategory › Code), and lets you switch the net column off per question when the client's layout doesn't carry it.
In a tracker, wave 4's coding isn't worth anything on its own — it's worth something if it's comparable to waves 1 through 3. The real risk isn't coding badly, it's coding differently: a code whose definition quietly shifts mid-year manufactures a trend that isn't there. So the codebook is inherited across waves with its code numbers intact, and emergent themes the new wave brings in are surfaced separately for an analyst to rule on — open a new code, or send it to a residual. That loop is what holds the time series together.
Survey Coder Pro exports coding results to Excel or to SPSS .sav with numeric codes and value labels. You can also add coded variables to a retained original .sav. This enriched-file workflow is in beta: it runs structural checks and can hold a file for review before release. Choose a supported layout and check the generated file before handing it to your data-processing team.
Start coding open-ended responses with AI today. Free trial with 250 responses.