Barry Ritholtz has a good article in Bloomberg about retail spending over the Black Friday weekend.
In short: “data” that is derived from surveys is typically (always?) crap. Other data (such as data derived from actual spending) is probably better. Even if other data is better than crap, however, it still might be crap.
Just because you can measure something – whether through a survey or by looking at receipts or whatever – doesn’t mean that what you obtain is useful. Having a number, and possibly also having a model into which to inject that number, is not necessarily useful.
And a side note, if you find yourself saying something like, “yeah, but it’s all we’ve got,” well, that’s a crappy argument. I, for one, would rather have nothing in my hand than crap. Something is not always better than nothing.