We and our connection partners made an update to the categorisation system, which should provide more accurate auto-assigned categories and tags on US and Canadian connections. The improved system is already live, so any incoming transactions are already using it. The default systematization is now more appropriate for personal finance applications like ours, even before being adapted to the default set of categories we use in Toshl. The earlier system used a broader and slightly dated venue classification, more often used by credit card companies.
This of course applies to new transactions, where the applied category and tags are based primarily on the transaction description that’s provided by the bank. On top of that, Toshl automatically learns from your corrections, so entries with similar descriptions will use your preferred categorisation.
In some cases it’s also able to determine the geographic coordinates based on the bank data and put the transaction on your map automatically, although this is still quite rare with most transactions.
If you’re very strict when it comes to your own accounting and would prefer to manually assign every category and tag, you can do that too. Just open the bank connections screen and tap Edit on an individual connection. You can set the categorization option to OFF there, so all new entries come in as “unsorted”. The same can be applied to transfer matching. Keeping the categorization ON however, can also help with transfer matching, as detecting an expense fits in the “Transfer” category is one of the factor that helps with detecting them. The system is then more inclined to look for the income part of the transfer and match it into a transfer.
Smaller improvements were also made in the way pending transaction categorisation is handled on all automatic bank connections, providing for more flexibility when the bank changes the transaction description considerably between authorising and fully settling the transactions.
Even when there are no new systemic changes such as the ones mentioned here, the categorization is constantly improving due to machine learning systems which learn from anonymized categorization data.