I wrote a script that cleans up OFX transaction files downloaded from banks so that the name and memo fields are more readable once imported into gnucash. This works great, I now get very clean records that include only the information I need for each transaction. The problem is, even after the clean up, gnucash's bayesian matching makes a lot of mistakes in matching transactions to accounts, and frequently "forgets" matchings it knew in the past.
I was wondering if there is anything else I can do to the OFX files in my cleanup script to give gnucash better hints as to what account a given transaction should go. I would like to bypass the bayesian matching stuff completely, and instead insert an explicit account name for each transaction in the OFX file. Looking through the OFX spec I could not find any fields for this. Is this possible at all?
If I could do that, then my script can assign default accounts for the common transactions that show up over and over again, like groceries, gas, etc. Then I'll have only the exceptions to handle manually.