How to predict product sales? I have a set of data that represents the number of products sold per day during three months and I need to predict the sales for the next two weeks. How could I do this? This is a sample of the data I have. I tried to do this with Box-Jenkins, but it didn`t work.
7 6 5 7 9 10 9 7 4 5 4 6 7 10 7 5 6 3 6 8 7 5 4 2 3 5 6 7 9 12 13 10 ...

 A: You submitted 32 values . Is that all you have for 3 months of data ? With 32 values this is what I obtained via Box-Jenkins . Here is the actual/fit/forecast using monte carlo re-sampling of model errors without future pulses  with this equation  .
Now enabling the possibility of future anomalies/pulses we get 
Perhaps your software or you didn't identify a model that included possible anomalies ( period 14 & 18 ) ... always a distinct possibility.
I used AUTOBOX which automatically identified this model. The acf/pacf of the original series is fairly clear as to a possible model . . The two "spikes" in the partial along with the "decay" in the acf suggest an ar(2) model . The initial model that was identified is here  suggesting 2 possible anomalies ( period 14 and 18 ). When two indicator (pulse) series were added this was the final model .
The 14 forecasts are presented graphically here  and tabularly here  . Note that two sets of forecasts are presented. The first 3 columns represent a monte carlo simulation (re-sampling the model residuals allowing for future anomalies) and the standard +- symmetrical limits assuming normality of the model residuals and no future pulses.
Here is the histogram of the model residuals ...    visually suggesting significant deviation from normality.
I am quite shocked that a trivial example like this caused eviews to fail.
