In online
One of the most promising areas in which work can be cut down in with business accounting is with bank feeds. These are hookups to your bank account transactions that automatically get pulled into your accounting software. Most software can use the descriptions associated with the bank transactions to guess at what the transaction might be and enter it for the user. For example, if the description on a credit card transaction came through as “STARBUCKS STORE #2348” it can recognize that this transaction is for Starbucks and categorize it as a meals and entertainment expense. Starbucks does sell other merchandise, but it’s a good bet that meals and entertainment is the proper categorization.
However, this simplified rule, based on using the description to parse out the vendor name and thus account categorization, doesn’t always work. Take an item was purchased at Staples for example, which could be a computer (asset) or some office supplies (expense).
So, transactions entered purely using bank feeds are not going to be completely accurate. That’s why another source of information – receipts – need to also be used in conjunction with bank feeds (not to mention the fact that all purchases need to be backed up by source documentation).
Working with receipts alongside bank feeds seems simple enough, but what if you have to categorize 100 purchase transactions in a single month. Every time you want to enter/verify a transaction found in the bank feed, you have to wade through 100 receipts to find its match. In a case like this, even rule based auto-categorization of bank feed transactions ends up not saving you much time when you have to keep on finding the right receipt for each of the purchases. This is where it would be a big help if those receipts could be uploaded to the accounting software and automatically matched to bank feed transactions.
A lot of accounting software do offer file attachments, but users have to manually attach documents (whether they be pictures of receipts or
- The $ amount.
- The date.
If OCR could extract that data, the chance of matching up a receipt to a transaction pulled from a credit card’s bank feed would be near 100%. Even simply knowing the final $ amount would result in matching accuracy above 90%.
We all know that OCR isn’t perfect. Perhaps it only gets the amount right 80% of the time. Still, with that level of accuracy, out of 100 transactions it would be able to match up roughly 80 of 100 transactions, leaving the user to only match up 20 manually. However, for those 20 unmatched receipts, if the user then verified the OCR, making sure the $ amount and dates were correct, you’d probably get close to 100% accuracy in matching receipts to transactions from the bank feed.
Automatically matching receipts to bank feed transactions would automate all the things that computers are better than humans at. What would then be left for the human to do is to verify entries. The user would see that Staples bill and that the $1,000 was automatically entered as office supplies when it should have been entered as a computer asset. With the receipt automatically attached to the transaction, this would be an easy thing to check and correct, with no need to wade through 100 receipts.
So, here’s how the idealized workflow would look like.
- Bank feed is pulled automatically from the bank
- The accounting software categorize the transactions based on rules formed from past entries of the user, an anonymous database of all users, and custom rules created by the user.
- Receipts are snapped by a camera, scanned in, or emailed to the accounting software.
- OCR is used to extract the amount and date from receipts.
- Using the amount and date information extracted from the receipts, they are matched up to transactions found in the bank feed. Because dates of transactions can vary by a few day in certain cases (like credit card purchases), exact date matches aren’t necessary.
- A user sees the majority of their transactions automatically categorized for them, matched up to receipts. Their job is to go through the transactions and verify that they are correctly entered.
Wave is probably the software that has come closest to doing something like this, as they do an OCR on receipts. However, matching those receipts to transactions from the bank feed has not been done.
Services like Receipt Bank or Shoeboxed do a good job of categorizing transactions based off of the receipt. The problem lies with the fact that they export their data to accounting software as expense entries, as opposed to data that can be combined with bank feeds. This causes a lot of mess.
We already have services that can do a decent job of extracting data off of receipts and we also have online accounting software that uses rules to automatically categorize a decent amount of transactions. If the bank feed transactions and receipts could be automatically matched up, this would save a lot of time as well as improve the accuracy of accounting records. So, which one of you online accounting software companies are going to do this for me?