Concept & Message Testing
Follow these guides if you are working with stimulus for ad, concept or product testing. These types of projects can be challenging for AI analysis as more basic tools will struggle to disambiguate concepts. This guide will explain best practices and techniques to do this with CoLoop.
Concept testing projects are particularly challenging for AI tools as they often struggle to disambiguate similar concepts. CoLoop features a specialised concept testing mode that can be enabled during project setup.
What does concept testing mode do?
Concept testing mode enables you to enter a list of concepts, brand names or any other key phrase during the project setup. The AI will automatically label text segments where they appear and even track when the conversation changes from discussing one to another.
- These will appear in the transcripts and evidence panel
- Concepts entered here will be passed into the Key Phrases section when transcribing any audio or video files
Study Design
If you’re planning to use CoLoop for analysing research material contain stimulus it’s important to plan ahead.
Choose concept names that will stand out
- Use concept names that will stand out:
- Good: Alpha, Beta, Gamma
- Bad: Concept I, 1, A
- Generic sounding terms are difficult to discern and will not be well picked up
Reference the concepts clearly in your interviews
- Be consistent with terminology when referring to concepts
- Introduce concepts clearly in the interview and make sure these come through in the transcript
Enabling Concept Testing Mode
- Click “enable concept testing” when creating a new project
- Enter a list of concept or brand names that you would like to distinguish between
- Click save and upload your research material
Editing Concepts
As mentioned above concepts will only apply to files uploaded after the concept was applied
- Concepts can be edited from the transcript view
- Click on the edit button and choose a concept from the list or create one
- Concept labels do not have to be perfectly accurate as they are considered along with the context
What if my concepts aren’t labelled correctly?
- In some cases you may notice the AI doesn’t do a perfect job of labelling concepts
- This can occur where concepts in the transcript don’t match up with the labels e.g. “C three” instead of “C3”
- This can be improved by applying corrections to transcripts before labelling the speakers:
- Select a mention of a concept in the transcript and click correct transcript
- Click show all instances
- Replace all mentions with the correct spelling
How do I use concept filters?
- Concept filters can be applied in the AI Chat and Analysis Grid by pressing the # key
- Concept filters improve accuracy by actively filtering sections of your research material to focus on specific labelled sections:
In the Analysis Grid
In the AI Chat
Was this page helpful?