[moneydance] getting categories from previous transactions

[Sean Reilly]

On 4/12/07, Jeff Macdonald <macfisherman+md at gmail.com> wrote:
> I thought 2007 would of solved this, but categories matching from previous
> transactions are not matched at all. For example:
>
> SHAW'S SUPERMA 01/16 #000988600
>
> I of course put personal:groceries. Another shopping trip and
>
> SHAW'S SUPERMA 03/21 #000111211
>
> I once again have to put personal:groceries
>
> this gets old pretty fast.
>
> I suppose there are some businesses with that have numbers in them, but I
> imagine the majority of entries use words or have at least a word or two, so
> why not have MoneyDance simply scrub numbers and create indexes based on
> alpha words. Something like this: (code in perl)

Unfortunately many of the transactions are best matched by the numbers
(like a store number, etc) so it requires a combination of the two.
If you update to the "preview release" of Moneydance (install the
Updater extension, then change the settings by clicking the extensions
toolbar icon) it will be more aggressive when matching online
transactions with previously downloaded transactions in order to
figure out the default category.

The new matching algorithm gives each similar transaction a score based on:
 1) exact (but case insensitive) match
 2) substring match (if over a certain length)
 3) match starting from the beginning (would match the transactions
you provided)
 4) match starting from the end (would match transactions with things
like a 'store number' on the end)

The parameters for this scoring need to be dialed back a bit since
some people are getting incorrect matches in the current preview
release, even when there is only a slight similarity.  This will be
improved in the next update to the preview version.

Thanks,
Sean