To save labor some vendors have begun using technologies such as OCR and ICR for data entry automation. These technologies are extremely effective with read rates exceeding 98% for a clean OCR scan line containing a font such as OCRA or OCRB, reading the MICR information on the bottom of a check or reading a bar code. Under ideal circumstances (clean high quality type, standard fonts, no background graphics or extraneous marks, identical documents or forms, etc.) automated technologies have the potential to be useful for converting typed data. The technologies are less beneficial when applied to printed handwritten data and have very limited usefulness when trying to read cursive handwritten data. Most OCR or ICR engines look for a break between one character and the next that may or may not be present with printed handwriting and is definitely not present with cursive handwriting. The OCR and ICR engines also require the user to set a "confidence level" for the probability the image has been interpreted correctly. Set the confidence level too high and the benefits of using automation are defeated as most of the information ends up being entered by a data entry operator. Alternatively, if the confidence level is set too low the incidence of "false positives" increases. False positives (the OCR engine incorrectly believes it has read the image) are a problem because the read may be assumed correct and consequently is not reviewed by a human operator. The one exception in using automation to read handwritten information is when there are an extremely finite number of possible responses (such as "yes" or "no").