Motivation:
My motivation for trying WatchOCR is that NeatReceipts as a company blows, the software interface is crap, and the SQL Database they force you to install is bloated, slow to start, and took 4 hrs with tech support to install. NeatReceipts also only reports the aggregate information from the receipt (no individual breakdowns of line items). The UI really sucks; I have never worked with an management system that required you to jockey so many pulldowns and screen refreshes. Furthermore, when you request the software to export your data, they export your data to an encrypted zip that only they have the key to.

Experience:
I just tried OCRing some scanned CVS receipts. These scans were obtained using NeatReceipts software, so the images have already been cleaned up, properly rotated etc. Cuneiform completely messed up the scan that NeatReceipts was able to do flawlessly. WatchOCR has a LONG way to go in terms of image quality. Far faster to install with a VM though. WatchOCR also will not extract the receipt information, but having the text of the receipt OCRed is the first step.

Hope:
WatchOCR uses an improved algorithm. Somebody else jumps in and creates a module for extracting receipt data (perhaps using the spatial information from the PDF?)