Apparently they don't estimate, they inflate how many results there are... For example, I never seen a case when their "estimate" was lower then the actual number of results.
An upper bound is a valid estimation. "Inflating" assumes that they have a good approximation, and that they arbitrarily increase it. I don't think that they do that.
The choice to take an upper bound (which will systematically err on the high side) and call it an "estimate" (which is not expected to have large, systematic error) inflates the numbers.
Most people don't expect upper bounds, though, when talking about estimations. I for one would appreciate a less-than or less-than-or-equal sign when this is truly an upper bound, as it gives me a better idea of what to expect of the actual number of results in relation to the estimation.
> Most people don't expect upper bounds, though, when talking about estimations.
What do you mean? If all I know about a number X is that it is an estimate for Y, I would expect X to be an upper bound for Y 50% of the time and a lower bound 50% of the time.
