The quiet anatomy of a bank rec: where 20 to 50 hours a month go

Business day three of the close, 7:40 PM. The ledger fills one screen, the bank portal the other, and a controller tabs between them, ticking wires off against journal entries in a highlighter grid rebuilt this morning because last month's version broke. Nothing on either screen is hard. All of it is slow.

And the slow part is measurable. In Ledge's 2025 close benchmarks, drawn from 100 finance professionals at companies from 51 employees to over 10,000, half of teams take six or more business days to close the books, and only 18 percent land the three-day close the software vendors keep pitching. Cash reconciliation alone runs 20 to 50 hours a month across three to five systems. The spreadsheet is not incidental: 94 percent of teams run the close through Excel, and half of them name it as a key reason the close is slow. APQC's benchmarking across 2,300 organizations puts the median monthly close at 6.4 calendar days, with the top quartile at 4.8 or less and the bottom quartile past ten.

If the fix looks like building something in-house, price the second year before you start

Where the 20 to 50 hours actually go

Watch a reconciliation week closely and the hours sort into three buckets. The first is collection: logging into bank portals, exporting statements, pasting them next to the ledger. The second is matching: the same fifty rules a person carries in their head, applied line by line. A wire from the same counterparty, same reference format, lands every Tuesday; a processor settles net of fees; a batch deposit splits across three ledger entries. None of this is judgment. It is memory, applied slowly. The third bucket is the real work: the twenty lines that do not match, each one a small investigation. When we map a reconciliation week, nearly all of it sits in the first two buckets, and the actual investigations get squeezed into whatever is left, which is exactly backwards.

What bank reconciliation automation actually does

Bank reconciliation automation connects the bank feed and the ledger in one place, then applies matching rules to clear the lines a person would have cleared without thinking. Every transaction the rules recognize is matched and logged with its rule attached, so the audit trail gets stronger, not weaker. Everything the rules do not recognize lands in an exception queue with an owner and an age. The people who spent their week collecting and ticking now spend it on the exception queue, which is the only part of the job that ever needed them. The sign-off stays human: automation prepares the reconciliation, a controller approves it. Teams that automate the matching layer stop measuring the close in screens open and start measuring it in exceptions cleared, and the 20 to 50 monthly hours compress toward the handful the exceptions genuinely require.

The order we would run it

1. Write down the rules you already use. Have the person who does the rec dictate their matching logic for one hour. If it can be said out loud, it can be coded. The list is the spec, and it is free.
2. Get the feeds into one place. Direct bank connections beat statement exports. Every portal login you eliminate is collection time gone for good.
3. Automate the top ten rules first. The ten most frequent match patterns usually cover the large majority of volume. Run them in parallel with the manual process for one cycle and compare.
4. Build the exception queue before you widen the rules. An unmatched line needs an owner, a reason code, and an age. This is where the close actually gets faster, because aged exceptions are what blow up day six.
5. Add learning last. Tools that suggest matches from history are useful once the rules and the queue exist, and premature machine matching on a messy ledger just automates the mess.

The honest caveat: the match rates quoted in demos are measured on clean histories. Your first cycle will match less than the demo did, because your data carries years of workarounds. That is not a failure of the tool; it is the mess becoming visible. The same pattern holds in accounts payable, where the payback hides in the invoices that do not match, and in the wider back office, where a Chicago bank's cost mostly sits in handoffs nobody priced.

First Monday, we would do exactly one thing: sit with whoever runs the rec and write the fifty rules down. No software decision, no vendor calls. The list tells you what a machine can take off their hands, and it usually tells you by lunchtime.