What reconciling actually means
Reconciling means taking two lists that ought to tell the same story, putting them side by side, and finding the points where they don't match. It's the business software and the warehouse spreadsheet, the courier's portal and the shipments on record, the list of invoices issued and the list of payments received. The subject changes, the work is always the same.
The heavy part isn't understanding the difference once it is in front of you, that takes a moment, because you know your own company. The heavy part is getting there, scrolling through hundreds of good lines to find the few that are wrong.
Why it goes badly by hand, and nobody is to blame
Whoever does this by hand runs their eyes down two screens, or exports two sheets and lines them up, and the further they go the more the mind does what it always does, it gets used to it. At line forty you are accurate. At line four hundred you read what you expect to read instead of what is written there.
Line-by-line comparison is exactly the kind of work a machine is better at than anyone, not because it is clever, but because it looks at line four hundred with the same attention as the first one, every time, including Friday at six.
The first step is deciding what ties the two lines together
Before you compare anything you have to know which line on this side matches which line on that side, and this is the decision everything else rests on. You need a piece of data that means the same thing in both systems, the document number, the item code, the VAT number, the date together with the amount.
This is almost always where the real problem shows up, that the two programs call the same thing by two names, your item code on one side and the supplier's code on the other. It isn't an obstacle, it's a conversion table, written once and then left there doing its job, but it has to be written first and written by someone who knows those codes.
The differences aren't all the same
When the comparison runs, what comes out sorts itself into four families, and each one asks for something different
- it's on this side and not on that one, so something was never recorded
- it's in both, but with a different number, a price, a quantity, an amount
- it's there twice on one side only, the classic duplicate
- it's in both, it adds up, and nobody needs to look at it
A made-up sum, just to show the proportions
Say that every month you have four hundred lines to compare between two programs, and that twelve of them are wrong
By hand you pay for the time to look at all four hundred, because there is no other way to find the twelve. With the machine doing the comparison, nobody looks at the three hundred and eighty-eight that match, and what reaches you is a list of twelve lines with what one program says next to what the other says
The time doesn't disappear, it moves. From half a day of scrolling to ten minutes of real decisions. And the decisions are the part where you are worth something, the scrolling is not
The automation doesn't decide who is right
Better to be clear on this, because it is where most of the hot air gets sold. A machine that compares two sets of records can tell you that a line doesn't match. It can't tell you why, and above all it can't tell you which of the two programs is wrong.
Behind a difference there can be a delivery note that was never booked in, a return recorded on one side only, a discount agreed verbally, a typo from three weeks ago. These are things you work out by knowing the company, and knowing the company is your job.
What an automation built properly can do is prepare the list with the context attached, the two values side by side, the document, the date, who entered it if the system records that, so that you look at it and know straight away where to go.
The check becomes something that just runs
There is an effect you only see after a while, and it counts for more than the time saved. A check that costs half a day gets done once a month, if you are lucky, and in the good years you skip a round or two.
A check that costs ten minutes you can run every week, or every night, and it changes the very kind of problem you find in front of you. A difference found the next day is a phone call. The same difference found in March is an investigation.
Where to start
To tell whether your cross-check can be automated, very little is needed. What the two sources are, how the data comes out of each one, which piece of data ties the lines together, and how many lines there are a month.
If the two sources are two different programs that don't talk to each other, that is the normal situation, and the screens are there anyway. This is why automating the business software you already use stays the road even when neither of them offers an integration. If instead, after the comparison, you want the data to actually move from one program to the other, that is a different job, and it is covered by the guide on two business systems that have to know the same things