Be as close to right as possible when reporting information. You can avoid many errors in qualitative research by leaving out bias, responses to poorly constructed questions or other incorrect information to ensure that you do not misrepresent any reporting data.
Identify any information that is, or hints at being, subjective. If it is, it could be a source for error or bias, or worse yet, both later on. Screen as early in the interview or survey process as is possible.
Eliminate identified information that is, or could be, biased or subjective. You must eliminate anything with the possibility for being taken as a personal interpretation, exclusive of other interpretations. This is called "objectification" of the inputs.
Acknowledge and work around the presence of errors by controlling for them. Errors often need to be, and can be, controlled in lieu of their being completely eliminated later. For example, if a portion of your population might be biased, identify that bias up front, code it and leave it out later.
Separate fact from belief, and then separate belief from error as soon after the interview as is possible. When reporting interview data or survey response data, for example, if they do not concur with your own beliefs, make sure you do not report them as your own. If such non-truths are included, code and place them squarely where they belong.
Locate a data cleaning software system. Software systems exist for correcting input errors using user-defined "integrity constraints." StreamClean is one system that, once a declarative language is defined, will correct input data errors.
When using a data cleaning software system, be sure to provide examples of correct data with descriptions and clearly defined values.
When an invalid data message appears, read it carefully and make any corrections or eliminate the data. The system will automatically modify the data with the correction.