In development

A huge progressive organization I‘m familiar with obviously does an online survey like it was just invented. I was to ensure a high response rate by positive user interaction, so I may be biased when I say that part went right – 2,500 completed surveys in 6 weeks. Analyzing more than 60,000 records dumped in an excel spreadsheet, here are a bunch of lessons learned about doing an online survey from Pakistan:

1. Don’t expect your developer to do anything more than code your survey. If you’re thinking about having the information neatly handed over to you in a report (or at least having categorized responses), then you should have listed it in your initial requirements. Because chances are he (in some cases, she) didn’t know you were actually planning on analyzing, he/ she probably thought it was just another one of those mindless instructions that loses its intended purpose once it’s trickled down the various layers of management.
2. Try your best to leave out the ‘Other’ option. This is usually seen as a wonderful opportunity for the majority of Pakistanis to express their creativity, bad English, limited intelligence in understanding how surveys work and aversion towards reading through the list of options and clicking the option they’ve so elaborately misspelled. Incase you need to have it in there, try to make sure the response is standardized in some way or you’ll find a minimum 449 different responses instead of the 6 you expect.
3. Make sure you questions are extremely simple to understand. Some examples of answers to the question ‘Which new services would you like to see offered in Pakistan?’ are “I like to see big services” and “I can’t understand what it means”. Maybe I should have been curt and simply asked “Which new services you want?”

However, I did have a good working relationship with the developer and pivot tables can make up for system generated responses, so it wasn’t entirely unworkable. However, I think developers should be more business-minded and take the time to think about what end-users want, even if they don’t specify it, instead of being fed instructions like a robot. Maybe I’m wrong, but I sure need a better approach next time around.