Definition
A codeframe is the defined set of codes that open-text answers are tagged with. The survey and questionnaire design guide gives the method in a sentence: code open-text answers into themes, using a codeframe built from a sample and then applied to the rest. Each answer can carry more than one code, and what is reported is the number of respondents per code, not the number of mentions.
The frame is a document as well as a list. Each row names a code, defines what it covers, says what it excludes, and gives an example answer. Related codes can be grouped into nets, such as several cost codes under one cost heading, so a report can show the broad picture and the detail beneath it. Most frames also carry codes for answers that fit nothing, and for blanks or answers to a different question.
Building, defining and checking the frame
The tutorial on coding open-text answers by hand works through the procedure. What makes the result a codeframe rather than a list of impressions comes down to a few properties:
- It comes from the answers. Codes are drafted from a sample of what respondents wrote, in words close to theirs, not from what the team hoped to hear or the questionnaire assumed.
- Its codes have boundaries. The exclusion line does most of the work, because overlapping codes only separate once each says what belongs to its neighbour.
- Two people apply it the same way. A second coder codes a portion independently; differences usually point to a definition that allows two readings, which is sharpened and the affected answers recoded.
- It is frozen with a date. A frame saved with its definitions can be reapplied in the next wave, which is what makes counts comparable over time. Changes are dated, not made silently.
The counting rule completes it: respondents per code against a base that is stated beside every figure, with a note that shares add up to more than the whole when answers carry several codes.
In practice: surveys, support contacts and AI-assisted coding
The same discipline shows up wherever free text has to be counted. The tutorial on tagging support contacts by reason starts from the codeframe method and adapts it for agents who choose a reason at the close of every contact, under time pressure, which pushes the list to be short and stable for months.
At larger volumes, the guide to AI for research synthesis builds the codeframe before the model touches the full set. A model may propose candidate codes from a sample, but a person chooses, merges and defines them, and the model then codes from the frame rather than inventing its own categories, with every excerpt traceable to its source and a sample of its coding checked against the definitions. The voice of the customer guide uses the same split: comments coded by hand for small volumes and with text analysis for large ones, with automated themes checked by reading samples and counts taken from the coded table rather than a tool's summary.
A garden centre chain counts what its customers complained about
Bramwexle Garden Centres ran a customer survey across its 5 sites and received 1,240 answers to "What could we do better?". The first report was a word cloud, in which "price" and "expensive" were the largest words. The insight lead then drafted a codeframe from a random 150 answers, with 14 codes in 4 nets, each defined with an exclusion and an example, and a colleague coded 200 answers independently to check it. Once all 1,240 were coded and respondents counted per code, parking and queues at the car park entrance were mentioned by 31% of the 1,086 who gave a codable answer, and price by 17%. The word cloud had counted words; the frame counted people.
Related terms
A ScreenerThe short questionnaire that confirms each research candidate fits a study, asking about recent behaviour without revealing which answers qualify. and a Discussion GuideThe plan for a research interview or group: a warm-up, the episode to reconstruct, open questions about past behaviour and probes, used as a plan, not a script. shape what a study collects in the first place; the codeframe is how the open material it produces is turned into themes afterwards. A Contact ReasonThe main reason a customer contacted support, tagged by the agent from a fixed, defined list so volume and other service metrics can be read by cause. is a code from a support team's fixed reason list, a codeframe built for agents to apply live. A Leading QuestionA survey or interview question whose wording contains or suggests its expected answer, producing measurement bias rather than evidence. distorts the answers before any coding starts, and no frame can repair that: codes describe what respondents wrote in response to the question they were asked.
Common confusions
- A codeframe is a word count. Word clouds and keyword tallies count words; a frame counts respondents per defined theme, so the same complaint in different words lands in one code.
- The codes can be written before reading the answers. A frame drafted from assumptions forces answers into the team's categories. The answers come first.
- More codes make a better frame. A frame too long for a coder to hold in mind is applied inconsistently; a code nobody would act on can usually be merged.
- A model can build the frame itself. It can suggest candidates. Choosing, defining and bounding each code stays with a person, and the model's coding is checked against those definitions.
- Coded counts measure how many customers hold a view. They count respondents who volunteered a theme in an open answer, which is not the same as the share of customers who would agree if asked directly.
