When a company asks us to look at technology spend, the expectation is usually that we will find one large mistake. Occasionally we do. Far more often the recoverable money is spread across seven categories, none of which is individually embarrassing, and all of which have the same root cause: nobody owns the aggregate.
Here is where it hides, and the question that surfaces each one fastest.
1. Licences assigned to people who left
The question: Reconcile your active licence assignments against your current payroll.
Not your seat count — your assignments. In a company with any turnover, licences detach from leavers in the identity system but stay attached in the billing system, because those are two different processes owned by two different people.
The gap is rarely dramatic. It is routinely five to fifteen per cent of the per-seat bill, and it compounds: every year the licence count ratchets up at renewal because the count is taken from the invoice rather than from the directory.
2. Editions bought for capabilities nobody uses
The question: For each premium tier you pay for, name the feature that justified the upgrade and show me its usage.
Premium editions are usually bought for one specific capability, often during a project. The project ends. The edition stays. Five years later nobody remembers which feature drove the decision, and the answer to “can we downgrade?” is a shrug, so the premium renews.
The usage data almost always exists in the admin console. It is just that pulling it is nobody’s job.
3. Overlapping tools bought by different departments
The question: List every tool that stores a file, hosts a video call, or tracks a task.
Departmental purchasing is efficient right up until the point where the company is paying for three project trackers, two file-sharing platforms and a video tool nobody has opened since the platform they already own added the same feature for free.
This is the category with the largest gap between what finance believes is being spent and what is actually being spent, because the charges arrive on expense cards rather than as vendor invoices.
4. Cloud commitments made against a forecast that did not happen
The question: Show me committed spend versus actual consumption, by month, for the last twelve months.
Reserved capacity and committed-use discounts are genuinely good deals when the forecast is right. When it is not, you are paying for a discount on capacity you are not using — which is worse than paying list price for what you do use.
The second half of this question matters just as much: what is running that nobody has looked at? Non-production environments left on overnight, storage from a migration completed two years ago, snapshots with no retention policy.
5. Telecom and connectivity inventory nobody has reconciled
The question: Match every circuit and every mobile line on your invoices to a location or a person.
This is the least glamorous item on the list and frequently the fastest payback. Circuits to closed sites, backup lines for systems that were decommissioned, mobile plans for devices that were returned. Telecom billing is complex enough that these survive audits for years.
6. Auto-renewals that nobody negotiated
The question: List your top twenty contracts by value, with their renewal dates and notice periods.
Most mid-market companies cannot produce this list in under a week. In the absence of it, contracts renew on their original terms, at their built-in uplifts, without anyone testing the market.
The savings here are not from switching vendors. They come from arriving at the negotiation with utilisation data, a credible alternative, and enough notice period left to be taken seriously.
7. Manual effort that has been normalised
The question: Which reports does someone build by hand every month, and how long does it take?
This is the category that never appears in a technology budget, because it appears in payroll. Someone spends three days a month assembling a report from four systems. Someone else re-keys orders between a portal and the ERP. It got that way because integrating the systems was always next quarter’s project.
The cost is real, it is large, and it is entirely invisible to a spend analysis that only looks at invoices.
The pattern underneath
None of these is a failure of competence. Each is the residue of a decision that was correct at the time, made by someone who has since moved on, in a system that has no mechanism for revisiting it.
That is why an outside review finds them and an internal one usually does not. It is not that the outsider is smarter. It is that reviewing the aggregate is the outsider’s entire job for four weeks, and it is nobody’s job the rest of the time.