Question
If you convert the frequency of a forecast time series in a Basket, the "last historical point" header field is also converted.
Answer
Did you convert the time series to a lower frequency? If so, the "last historical point" has also been converted. This metadata attribute indicates the last range of complete historical data, at the output frequency, not the native (original, unconverted) frequency of the series. Depending on the point in the calendar year, the results can look odd, but are correct and meaningful.
Examples
FET.US is a quarterly forecast time series. In Basket output, at its native frequency, it shows a last historical point of "3/30/12" meaning 2012Q1. But when converted to an annual frequency, it shows "12/31/11," which you should read not as 2011Q4, but rather, 2011. That is, the series contains complete historical data for year 2011, but 2012 is still partial.
FTRACFSM.US is a monthly forecast series. At its native frequency, its last historical point shows as "5/31/2012" meaning May 2012; at quarterly, it's "3/31/2012" meaning 2012Q1 (because 2012Q2 is partial), and at annual, "12/31/2011" meaning 2011 (because 2012 is partial).
Commentary
Depending on the module of Data Buffet, you may see this metadata attribute labeled as "last historical point," "historical end date," or "LASTHIST."
In Microsoft Excel-formatted basket output, header date fields are always displayed with daily precision, even when expressing monthly, quarterly, or annual periods. You can alter the cell formatting afterward. You can also convert your Basket output to a Power Tools workbook, to which you can apply cell formatting that is persistent even when you refresh data cells.
We integrate historical data ("actuals") into our forecast series on a delay. For example, if the driver is a monthly series and the forecast is quarterly, we wait until all three months constituting the quarter have been reported; then we wait further until our regular forecast production cycle. Hence, the historical portion of the time series is updated four times a year, although the forecast model is run 12 times.