Definition of Terms: 
Gross-Gross :: The gross amount The Orchard receives, without regard for supply chain distribution
Gross-Net   :: The gross amount The Orchard receives, net of supply chain. 
GPU         :: Gross Per Unit. The gross-gross amount per paid unit of stream or download 
Group       ::  ....

Notes: 
* All amounts are in USD unless otherwise stated. Currency conversion is handled by Orchard Engineering.  See below for conversion information. 
* In all data calculations involving monthly statistics, for iTunes the "month" in analytics is adjusted to iTunes unique 5-4-4 accounting.  eg. Transactions that occurred on March 2nd, 2014 are tagged as "February 2014"


Forecast Overview

Three Parts: 
(1) Calculate GPU
(2) Create BI Tables
(3) Run Forecast


GPU Calculation
---------------
* Aggregate from OA Accounting, sales and gross
  BY:  Month, Store, TransactionType
  ** Any reference to "by group" means by the above stated breakdowns
* GPU = Total Gross / Total Units  by group
* We disregard any groups wih $0 gross.  (Generally, this is just one group, "itunes - Upgraded Download Albums - June 2013")

This gives us the GPU until the last accounting date. 
Append one new month after accounting date.  (Eg, currently, as of 2014-10-15, accounting goes until 2014-07-01, so we append 2014-08-01)
(This is DT.storeavg2)

For each group, create two forecats: GPU
(1) A simple 3-month MA 
(2) An ARIMA model with seasonal component
These are each created iteratively, starting from the most historic date to the most recent

Append future dates to GPU table.  Push the forecasts values down one row, to line up with the nex month
Calculate the simple error for each model
Use the last 4 months to measure error


.......

(2) Currency Conversion
(3) Forecast


The amount of revenue per bucket is found by selecting the dollars with the corresponding tag(s)

Old Method: 
Split each Dollar into percentages

New Method: 
Each Dollar is “tagged”, much like our content is tagged.


GPU: 
Looking back over 24 months, 
from sales
calculate gpu by month, store, transac
forecast said GPU

from analytics
- take into acount itunes544

.......

forecast

from bi.accounting take  gross, units
from bi.analytics  take  gross, units, paidunits, freeunits, gest, gest_net_of_SC, gest_net_of_SC_no_rd

## RECALL:  gest is the following
  ----------------------------------------------------

    gest_no_rd = GG.gpu_using * AA.paidunits
    gest       = if   (is_royaltydollar_zero) GG.gpu_using * AA.paidunits   
                 else  royaltydollar_times_paidunits

    SET gest = royaltydollar_times_paidunits
    WHERE (NOT is_royaltydollar_zero)  
    
  --    FILTER BY paidunits_expected_ratio != 1
  --    THEN MULTIPLY gest := gest * paidunits_expected_ratio
    UPDATE bi.aggregated_analytics
    SET gest       = gest * paidunits_expected_ratio
      , gest_no_rd = gest_no_rd * paidunits_expected_ratio
    WHERE (paidunits_expected_ratio != 1) 
    ;
  ----------------------------------------------------

    tmp_DT.split_percs.acc    <- DT.acc[, list(OA_gross_forDSOM = sum(gross)), keyby=kCols.splitgroup]
    tmp_DT.split_percs.anal   <- DT.anal[date %ni% unique(tmp_DT.split_percs.acc$date), list(OA_gross_forDSOM = sum(gest, na.rm=TRUE)), keyby=kCols.splitgroup]
    DT.split_percs.unexpanded <- rbind(tmp_DT.split_percs.acc, tmp_DT.split_percs.anal) [date >= minDateToKeep]

Calculate the split percentages
  OA_gross_forDSOM :=  sum(gross) when DS is in DT.acc, otherwise summ(gest) when DS is in DT.anal
  
DT.split_percs.unexpanded







=========================================================

## Only include stores whose last date in GL is the lastDateClosed (ie, stores we're still working with), and ignore dates beyond lastDateClosed (they will be NA)
tmp_DT.GL_gross_forDS <- DT.merged[(date <= lastDateClosed) & (GL_lastDate_forS == lastDateClosed), list(GL_gross_forDS = GL_gross_forDS[[1]]), keyby=kCols.datestore]

##  --------------------------------------------------  ##
##            RUN FORECAST FUNCTION
##  --------------------------------------------------  ##
cat("Beginning forecast ...    ")
DT.GL_forecasted <- forecastByDateStore(tmp_DT.GL_gross_forDS, datesToForecast=dates_not_in_GL, byCols=kCols.datestore)
cat(" done.\n")
##  --------------------------------------------------  ##


DT.forecasts <- DT.merged[!is.na(forecasted_value), .SD, .SDcols=unique(c(kCols.smry, forecastColsToBringIn, "margin", "OA_percof_DS_forDSOM"))]

------------------

[DEFINE: Group]
The forecast value is computed on a store-by-store basis. [How Calculated?]  
Each Group''s value is computed by multiplying the store value by the group''s respecitve precentage [How Calculated?]


DT.ancillary_with_forecasts <- rbind(DT.ancillary_from_rev_summary, DT.forecasts_manual_from_rev_summary, use.names=TRUE, fill=TRUE)


DT.forecast.full is the rbind of DT.forecasts and DT.ancillary_with_forecasts

We clean up the buckets for DT.forecast.full

Company Overview Forecast is ...