This article describes the difference between click date- and transaction date-based reports and why the model uses data by click_date.
*When you report conversions by transaction date, the data includes transactions whose transaction date occurred during the specified time period. This is the default setting in basic reports and advanced trend reports, and it will show you how much revenue was earned within the specified time period.
*When you report conversions by click date, the data includes transactions that have resulted from a click that occurred during the specified time period. When a portfolio has significant delay between clicks and transactions, this type of report is useful for computing the historical revenue per click for the portfolio, which will give you an idea of what revenue behaviors to expect over time and report procedure.
When building model by using transaction_date, we bid only on those keywords whose transactions are supposed to occur on the day (today) and whose clicks have occurred earlier (and we have spent earlier).
The current approach of building model by click_date means, we will bid on all the keywords whose transactions may occur in future (for these clicks that we are paying to SE for the day).
i.e., when using model by click_date, the click, spend and revenue would be attributed to the same day.
If the transaction_date is used, then the click and spend could be attributed to one day and the revenue to another date; which by logic in terms of modelling would be incorrect and will not give accurate bid points.