Parsing Internet Banking data from multiple banks, cards.. ideas?
Posted on 2012-03-13
("The Question" is at the end)
I use the term "banks" to include all sources such as credit cards and building societies.
I am a practicing accountant (which is relevant in this question as it provides context)
Clients (if I am lucky) send me CSV files of bank data, transaction lists, the same sort of thing cloud accounting systems import automatically, at least some do.
Cloud systems tend to use the user's private login details to log in and "scrape" the data or download files for import. These services cost from say 10 pounds/dollars/euros per month upwards, a bit less for a few in their minimum guise.
There is also a service not fully rolled out yet that direct links to banks and downloads data daily, this uses separate login and does not entail use of private login details. This costs about 10 pounds/dollars/euros per month (per client). Data is delivered in a homogenous layout which would solve this Question. Problem is not enough banks are covered, data cannot be historic (that's a killer) and costs make too large a hole in the fee.
The last para above illustrates my question. It's the layouts. Excel has a number of ways to parse data and does it very well.
So does the Mac assist with the ability to create "hot" folders that for example when I drop in a dozen csv files, will concatenate them into one. On my PC that's easily done with a batch file copy *.csv big.fil (or type > etc)
So now we have these files which are invariably delivered upside down (i.e. newest transactions first, which is a real pita because the files arrive oldest files first (filenames)) so we end up with the initial sheet having blocks of data oldest first but the data within newest first. Grr! IF the client has not been VERY careful to commence each download immediately after the last one, this causes time consuming checks because the balances month to do not tally.
The solution involves sorting the transactions into proper ascending order, while maintaining then original (but reversed) order, then adding a calculated balance column, and comparing with reported balance (usually in the csv, not always) and highlighting discrepancies, while allowing for more than one, at this point I fix the first and see what is next, a clever program could take the error, adjust the calculated balances and find following errors too. That would be cool.. actually the calc balance column could do that. I'm just so pissed off I don't bother.
The above is Problem 1.
Next problem is this, I have say 7 banks' csv files, some with many (that means 1,000 or so) transactions and most with a few, credit cards, savings accounts and the like.
These files are delivered with different column headings (field names) and there lies my problem.
Sure I can manually import and muck about with import headings and then spreadsheet table headings etc, which indeed I do and have done for some time - there has been a process where few then some and now nearly all can/will deliver in this form, some much better than others. One uses pdf.. (expletive deleted).
I suppose what I am looking for, though I am open to suggestions here, is how to do a sort of "save settings" for each bank.
Which I just realised is exactly what one can do. However, it doesn't solve my problem, which is the order of the fields/headings. These differ in order (order of headings/columns/fields across the page) and content (balances, amounts, descriptions, references) from bank to bank.
Powerpivot springs to mind. PC Only mind, won't work on my mac, though MS Query will.
So that's Problem 2
Having acquired the data in the right columns order, which at this point would still need problem 1 to be addressed) there is then the issue that banks insist on concatenating all sorts of, dare I say rubbish into what would be a perfectly good payee name. Things like Tesco transaction numbers, such that instead of all Tesco entries being the same (great! I could code that) they are all unique. Bear in mind the payee name is random, we do not have a nice list of payees with which to identify.
I have noticed that judicious use of =mid(etc) can work reasonably effectively because the banks I suspect have fixed field lengths, but as I have found this cannot be guaranteed effective in the application, some entries fail. What I do there is insert a helper column and insert the mid's number of chars from which to start, ref that in the mid formula and manually over ride when one item fails.
Parsing what has already been parsed.
And so that's problem 3
I am NOT asking for applications to be written to solve this.
I am asking what features and perhaps techniques you think I could explore.
For example external data into Pivot Tables, Powerpivot or MS Query, all of which can be saved and I am familiar with, wherein lies part of the problem. However, I need a system not just one method. A means of managing and re-using so if I have done the work fo banks 1 2 & 3 I can save it such that I do not have to remember anything other than say which method I use, so PP or MSQ or whatever.
I do not want to use any non Excel third party applications (I am thinking of parsers here.... maybe I shouldn't be so quick to say that actually).