Should I use TREND or LINEST to forecast sales data?

At their core, TREND() and LINEST() are the same function -- they both perform a linear regression on the input data.

The LINEST() function performs the regression, returns the slope(s) and intercept for the linear regression, and can also return additional statistics (standard error for each slope, standard error for y, etc.) http://office.microsoft.com/en-us/ex...005209155.aspx

The TREND() function performs the same regression, computes a slope and intercept (but does not return these to the spreadsheet), then, using the new_x parameter, calculates the y value (in statistics text, this is often referred to as y-hat) corresponding to new_x. http://office.microsoft.com/en-us/ma...081.aspx?CTT=1

In my own spreadsheets, I generally prefer to use the LINEST() function, because, most of the time, I want to actually see the calculated slope and intercept. I can then calculate as many y-hat's as I have new_x's if I want. If I have a lot of new_x's/y-hat combinations to calculate, I expect there might be a slight performance advantage to using LINEST() over TREND(), because Excel only needs to perform the regression once (for the single LINEST function) rather than performing the regression multiple times (once for each instance of the TREND() function).

In the end, the most important consideration for whether to use LINEST() or TREND() is whether or not you need to actually see the slope and intercept computed for the regression. If you need/want to see the slope/intercept values, use LINEST(). If you have absolutely no need to see slope/intercept, the TREND() is a suitable function.