Efficient paper feedback surveys
October 11, 2013 8:04 AM   Subscribe

I want to give paper surveys to a number of people and then use a computer to compile the responses. My thought is to use a camera or scanner to enter the evaluations into the computer, and then run some sort of OCR (which i know nothing about) or other program to highlight which boxes have X's in them or which items are circled, and then throw all that into an excel spreadsheet for further analysis. From a technology standpoint, (a) is this possible? (b) is this possible to set up in a week? (C) is this possible in such a way that tech-illiterate people can do it swiftly once it's been set up?
posted by rebent to Technology (7 answers total) 1 user marked this as a favorite
 
Here you go. Google "ocr survey scanning" for more options.
posted by beagle at 8:27 AM on October 11, 2013


If these are bianary surveys, or multiple choice, then what you want is Scantron.
posted by Ruthless Bunny at 8:40 AM on October 11, 2013


Can you say more about how many surveys you will be collecting? For small numbers (like, less than 1,000), the set-up cost for processing the OCR results or setting up scantron may not be worth it. You might be better off paying someone to do data entry.
posted by OrangeDisk at 10:28 AM on October 11, 2013


Possible? Yes, but you're going to need to do a fair amount of QA, especially for open text fields.

It really depends on the scale. How many surveys? How many questions? Are there open text fields? How will you handle superfluous comments that people write on the form? If you set this up as a single data entry screen on Zoomerang, you could could probably manually enter the data faster than setting up reliable OCR.
posted by 26.2 at 10:37 AM on October 11, 2013


Response by poster: Excellent questions, thank you. Im basically looking at a conference-like situation where multiple people (50-100) are in multiple sessions (1-5). The unique challenge with this is the conference leadership would like to collect feedback from each session and make changes to future sessions based on it. Hand-keying data would be a good idea based on the size, but the time constraint made me wonder if a technological solution would be better.

I think that Scantron would be the perfect solution to the technological constraints issue, but was hoping for more of a open-source, low-cost solution. Also, i think people would dislike it.
posted by rebent at 11:03 AM on October 11, 2013


Check out Captricity. Their service processes scanned paper forms by splitting them into separate fields and using algorithms and crowdsourced human workers to process them. This allows them to provide accurate handwriting recognition for open text fields. Their quickstart guide outlines the process in more detail.
posted by Jotnbeo at 11:06 AM on October 11, 2013


I think the services that folks have linked are the way to go. As a data point, I worked at a company that did something like this. We were using extremely fancy six-figure software with super-precise $30k scanners, and we still had to do endless tweaking of parameters to get it picking up the data. Once that was done, each scan would be visually checked by an operator and then that operator's work spot-checked by another operator. The end result is probably 2 or 3 times faster than a 40wpm data entry clerk.

At your volume, with free software and a basic scanner, I think you could easily spend months getting it set up and still have unsatisfactory results. I would either hand-key or contract it out.
posted by pocams at 8:41 PM on October 11, 2013 [1 favorite]


« Older Answer this question, and I may send you a stupid...   |   Can I use a keyboard with my first-gen kindle... Newer »
This thread is closed to new comments.