How to estimate project costs - Mail (Postal) survey?

Question

A brief background on myself: I am a database developer that does freelance work for multiple companies. I generally manage data for these companies for large projects -- primarily surveys in mining regions in Africa and Asia.

Recently, I was asked if I could develop some surveys for a company. The surveys would primarily be done face-to-face but one survey will need to be either mailed out to area residents (mail list) or done by phone (phone list).

I've never been responsible for estimating the costs of a mail survey before (the survey will be done in Canada). The cost to mail each survey will be $0.92 and printing costs are estimated at $0.25. I also said that a $500.00 fund should be set aside for costs such as packaging the mail, delivering them to the post office, and scanning the returned mail.

My question is, does anyone with experience doing mail-out surveys think I have forgotten anything? I simply multiplied the mail & printing costs by the number of surveys (1,900) and added the $500.00 estimate. The $0.92 includes return postage.

Should I plan to mail out a follow-up letter for each recipient that doesn't respond? This would probably help with our response rates which are important to improve the validity of the survey. I would like to get about 400 surveys returned and I'm not sure if sending out 1,900 immediately would be too much (or too little), I came up with this number through my research on survey response rates.

Please don't feel like you have to go into a ton of detail on this. I just wanted to make sure that I'm roughly on the right track. I understand the $500 for costs might be low and should probably be doubled - the scanning should only be ~500 surveys max as response rates likely wont be high.

Thanks so much for your input

Joe

Answer 1

This is something I've done myself...

First, make sure you've got "letter stuffing" as part of the costs. Some places are happy to print, but won't stuff envelopes. I managed around 4/minute at peak stuffing rate.

Second, allow tranches of surveys to be sent. Relieving the pressure on your print/send service will make things cheaper and more accurate, as well as allowing you to cut short if you reach the required response count from a given area. Split your tranches across all areas to keep a good sample.

For example: Group recipients into areas. Assign numbers 1-5 at random within the groups. These are your tranches. You might hit the response level required by tranche 3, so you can cancel tranche 5 (4 will likely be in production already). If you think you won't hit the response count, you can bake a reminder into the sequence.

Third, interpreting scanned responses still takes time. Will you be using data entry people or software? Budget accordingly to allow for exceptions in software. Is the response staying in-country? Do you have people there to do the scanning? Have you allowed for taxes, etc?