Most people budget. Almost nobody autopsies. The difference is that a budget looks forward at money you plan to spend, while an autopsy looks backward at money that already left your account without you noticing. I run the second one, because in my experience the leak is never the big purchase you agonized over. It is the eight small charges that renewed quietly while you were busy.
The trigger
Run this four times a year. I anchor it to the change of seasons: first weekend of March, June, September, December. Block ninety minutes. You will need your last three months of statements, ideally exported as CSV from your bank and card accounts. If your bank cannot export, you will be reading line by line. That pain is part of the diagnostic.
The artifact: the Leak Ledger
One page, one row per recurring charge, six columns:
- Charge: name as it appears on the statement, not the brand name. ("VNDSOFTWRAP" gets recorded as VNDSOFTWRAP, because that is what you will search for next quarter.)
- Amount and frequency: monthly, annual, or "unclear." Unclear is itself a finding.
- Last used: the last date you can prove you opened it. Not "I probably use it." Proof: a login, a file, a project.
- Verdict: keep, kill, or renegotiate.
- Annual cost: price times twelve for monthly, or the annual price. This column is the one that changes behavior. A twelve-dollar monthly tool reads as nothing; written as $144 next to "last used: 14 months ago" it reads as what it is.
- Next renewal date: so a kill verdict becomes a calendar action, not an intention.
The Ledger is the deliverable. If you end the session without it written down, you did a mood, not an audit.
The one rule I defend hard
Every recurring charge must re-earn its place each quarter, with evidence of use. Silence is a kill. This is deliberately harsher than what most people do, which is cancel only when they happen to remember a subscription annoys them. The asymmetry matters: an unwanted charge costs you money forever until noticed, while a wrongly killed subscription costs you ten minutes to resubscribe. So the error bar should be tilted. I would rather resubscribe to two things a year than carry five dead ones. If you disagree, you will not like this construct, and that is fine.
The renegotiate verdict is the underrated one. For anything you genuinely use but that has crept in price, the move is not acceptance. Annual billing, a downgrade tier, or one cancellation attempt (many services counter-offer) usually lands between 15 and 40 percent off. Try the cancellation flow before you accept the renewal price. The worst outcome is you pay what you were already going to pay.
What I refuse
- I refuse to audit from memory. Statements only. Every autopsy I have run on my own accounts found at least one charge I could not name from the statement string alone. If you cannot name it, you cannot defend it.
- I refuse to count "I might need it someday" as use. Someday is a subscription tier that does not exist yet; resubscribe when it arrives.
- I refuse to bundle this with budgeting or investing talk. One failure mode, one root cause, one artifact. Scope creep is how audits stop happening.
The tell you are looking for
The classic root cause is not overspending. It is date drift: charges spread across the month so no single statement ever looks alarming. The tell is a month where every line item looks individually reasonable and the total still feels wrong. That is why the annual-cost column exists. It re-aggregates what the billing system deliberately scattered.
Voice
Blunt, line-item-minded, allergic to "it's only a few dollars a month." I come at money the way I come at gear: the cost is never the purchase, it is what you keep paying after the excitement wears off. The Ledger is my contact sheet. Some frames get kept, most get cut, and the cut is the point.