The Code Frames feature allows a user to create, manage, and apply their own structured code frames — Nets, Codes, and Descriptions — to guide the AI-powered analysis.
By default, Canvs relies on AI-generated code frames that are automatically created and carried forward from wave to wave. Code Frames give users who need greater methodological control, standardized taxonomies, or consistency across questions, waves, or datasets the ability to define that structure themselves.
Key Benefits
• Build it your way - create a code frame directly in the dashboard, or upload a spreadsheet of nets, codes, and descriptions to get started faster
• Reuse across datasets - a code frame built once can be applied to new dataset uploads or waves to existing datasets, enabling clean, apples-to-apples comparisons
• Description-driven accuracy - code descriptions actively guide the AI, not just document it; clear descriptions improve how consistently and accurately verbatims are assigned
• Rigid vs. Flexible control - choose how strictly your framework is enforced
Managing Code Frames
Code Frames are managed from the dashboard landing page, found in the top-right corner of Canvs next to Import Data. You will find both Code Frames and Rules here since both features represent user-defined logic that influences analysis.
Clicking into Code Frames opens the management page, where every code frame in the account is listed with an Active toggle, Nets and Codes counts, and Date Created / creator. This page also toggles between the Code Frames and Rules tabs — Rules is an existing feature that now shares this same experience.
From this page, users can:
• Create Code Frames
• Edit Code Frames
• Duplicate Code Frames
• Export Code Frames
• Delete Code Frames
Each Code Frame consists of:
• Nets
• Codes (maximum 200)
• Code descriptions
Creating a Code Frame
Clicking Create lets users either manually enter their code frame or upload one they already have, both paths live behind the same button.
Upload Validation
During upload, the system will:
• Skip duplicate Nets/Codes
• Skip Codes containing invalid characters
• Skip Codes with descriptions longer than 800 characters
• Reject uploads containing more than 200 Codes
• Display skipped Codes so users can review them before completing the import
NOTE: Only the following are supported characters for nets and codes:
Letters (A-Z, a-z)
Numbers (0-9)
Spaces
Parentheses ( )
Ampersand &
Forward slash /
Apostrophe '
Hyphen -
Managing Existing Code Frames
The “…” menu on each Code Frame row provides Duplicate, Export, and Delete, rounding out management alongside the Active toggle.
Applying Code Frames to Your Dataset
During the Edit Details step of dataset upload, Code Frames is surfaced under AI Summary Customization alongside a Rules toggle. Turning on the Code Frames toggle lets users select from their already-created Code Frames to apply to that new upload.
On the pop up modal, you can select your code frame for each individual question in the imported dataset.
Users can:
• Apply a Code Frame to individual questions
• Apply the same Code Frame to multiple questions
• Leave questions without a Code Frame (default platform behavior)
Note: A dataset may use only one of the following analysis inputs at a time: General Instructions, Rules, or Code Frames.
Coding Behavior: Rigid vs. Flexible
Each Code Frame has a coding mode that determines how the AI uses the supplied taxonomy.
Rigid Code Frames
When a Code Frame is marked Rigid, the AI will only assign the Nets and Codes contained within that Code Frame. No new Codes or themes will be created, even if new concepts appear in the data. It is on the user to add a code to their Code Frame themselves if they want a new theme tracked.
Benefits: high consistency, repeatable coding across waves, strong methodological control.
Tradeoffs: reduced flexibility, and coverage may decrease if the Code Frame does not adequately represent the data.
Important: The quality of analysis is directly tied to the quality of the supplied Code Frame. If important concepts are missing, responses may remain uncoded or be assigned less accurately.
Flexible Code Frames
When a Code Frame is not marked Rigid, the AI uses the supplied Code Frame as a starting point while remaining free to create additional Codes for emerging themes. The user's Code Frame guides analysis and is applied first, and emerging themes continue to be identified automatically.
Benefits: better coverage, ability to detect new themes, more adaptive analysis.
Tradeoffs: less consistency across waves, and reduced methodological control.
Writing Effective Code Descriptions
Code descriptions are used to guide the AI when assigning Codes. Best practices:
Use one or two concise sentences
Clearly describe the intended meaning of the Code
Avoid descriptions that are overly broad or excessively detailed
Poorly written descriptions may reduce coding accuracy.
Access by User Role
Admins and Managers
Can create, edit, and delete Code Frames.
Basic Users
Cannot manage Code Frames — the Code Frames button is disabled.
Common Workflows
No Code Frame Applied (Default)
If a dataset is uploaded without selecting a Code Frame, upload proceeds as it does today: Canvs generates the AI Code Frame, and that AI Code Frame is carried forward across future waves. No change to existing behavior.
Applying a Code Frame During Upload
When a user applies a Code Frame during upload:
• Questions with a Code Frame use the selected Code Frame as the basis for analysis, following either Rigid or Flexible behavior
• Questions without a Code Frame continue using the standard AI-generated Code Frame
This supports hybrid datasets where only selected questions use a user-defined taxonomy.
Introducing a Code Frame Mid-Study
If a study is already in progress and a user introduces a new Code Frame, the existing AI-generated Code Frame is replaced by the user-selected Code Frame for future processing. Future waves use the user-provided Code Frame as the new analytical baseline.
Important considerations: introducing a new Code Frame partway through a study creates a break in continuity.
Historical comparisons may be affected
Earlier AI-generated structures are not preserved
Coding consistency should not be expected across the point where the Code Frame changes
Frequently Asked Questions
Does Code Frame control apply to Full Reprocessing and AI Codes?
Not currently. This use case is still being evaluated for a future release. Currently code frames can only be applied to net new datasets or waves of data.
Does a Code Frame apply to AI Boosting?
Yes — AI Boosting fully respects the Code Frame. Code Frame control is available for new datasets and new waves added to existing datasets, and Boosting on those will always follow the framework the user has defined.
Can I change my Code Frame at any time?
Yes. However, Code Frame adjustments will not retroactively apply to datasets that have already been processed. Changing the Code Frame can also affect the consistency of summarization in subsequent waves for tracker studies, so changes should be made deliberately once a study is underway.
What is the difference between Rigid and Flexible?
With Rigid mode turned on, Canvs AI sticks strictly to the defined framework, classifying every verbatim using only the codes and nets the user has set, even if new patterns emerge in the data. With Rigid mode left off (Flexible mode), Canvs AI still uses the framework as its foundation, but can introduce new codes and nets when it detects themes significant or prevalent enough in the data to warrant them. Rigid favors strict consistency; Flexible favors adaptive discovery.
Canvs AI • Questions? Contact your Customer Success Manager.