How do I stop paying for software we don't use?

Question

VX-N · Accepted Answer

Pull the last twelve months of card and bank statements, list every recurring charge, and match each one to a person who will defend it. Anything nobody defends gets cancelled. This exercise routinely finds a meaningful monthly number in businesses that were not being careless. Sprawl accumulates honestly — a tool bought for one project, a seat for someone who left, a plan upgraded during a busy month and never reviewed. Nobody decided to waste it. Which is why the audit has to run from the money outward, not from memory inward: the statements catch the charges nobody remembers, the annual renewals that hit once a year on a card nobody reconciles, and the duplicate subscriptions two departments bought separately for the same job. Audit seats as ruthlessly as tools. A platform the team genuinely uses can still be quietly overpriced: licenses assigned to departed employees, admin seats where viewer seats would do, a tier bought for one feature nobody ended up using. Most vendors show last-login data — anyone inactive for 90 days is a conversation, not a renewal. And before cancelling anything, export the data and confirm you can read the export; the time to discover a tool was holding something irreplaceable is before the account closes, not after. Cancellation is the visible win; consolidation is the larger one. Two overlapping tools usually means duplicated data entry, and removing one saves the licence plus the labour — plus the errors that came from the two copies disagreeing. The audit question is not only 'do we use this?' but 'which of these tools are doing overlapping jobs, and which one actually owns this data?' A stack of five partially-used platforms is usually hiding two or three real jobs. Make the fix structural or you will run the same exercise next year with the same result. Put every subscription in one place with an owner and a renewal date, route new sign-ups through a single approval step, and put annual renewals on a calendar thirty days ahead — which is also when you have negotiating room, since vendors move on price for a customer who asks before auto-renewal rather than after. Sprawl is not a one-time mess; it is a default state that returns unless something resists it. One more option exists now that did not meaningfully exist before: several thin subscriptions doing fragments of one job can sometimes be replaced by one small custom tool that does exactly that job, built in days rather than quarters at AI-era speed, owned outright with no per-seat meter running. It is not the answer for everything — mature products earn their fees — but at current build costs it belongs in the comparison.