statistical bias

16 July 2001

what’s the point in a survey that collects no useful data whatsoever? on friday night we ate at chiquito’s, a mexican restaurant conviniently located next to ikea where we’d been buying furniture that came in bigger boxes than my car. on our table was a survey to fill in, in return for our comments we’d get entered into a draw to win a free meal. not being in the habit of turning down a chance to win things we filled in the survey. presumably the restaurant makes this offer of free grub in return for opinions in order to obtain some useful information about what their customers think of the ‘dining experience’.

there was no way to respond to this survey that would let us tell them what we really thought. the four choices of response for each item on the list were ‘mexcellent’ (yes, really), great, good and ‘not hot’. that doesn’t give you a lot of leeway for saying that something was average or could’ve been better or that whilst the meal wasn’t expensive it wasn’t cheap either. i could complain about the categories too (under ‘service’ what designation would you use to say that the bar service was abysmally slow, though the table waitress was pleasant but the food arrived at such a speed that you didn’t get a chance to breathe between courses?) but that would be a parenthetical rant. what mainly annoys me is that i suspect they take these surveys and go off and brag to the industry that ‘x% of our customers think our service is good’ and suchlike. when many of those customers will have filled in a box closer to the bottom of the scale than the top and expected it to mean ‘not as good as we’d hoped’ like it usually does.

we rated it pretty much ‘not hot’ on statistical principle, i still hope we win a free meal though!