Enterprise software ROI leaks in 3 places: 56% value gap, 46% idle seats

01Build: large projects deliver 56% less value than planned

The most cited measurement of IT project outcomes is still McKinsey’s work with the BT Centre for Major Programme Management at the University of Oxford, which examined more than 5,400 IT projects. On average, large IT projects, defined as those with an initial price tag above US$15 million, ran 45% over budget and 7% over time while delivering 56% less value than predicted. Seventeen per cent went so badly that they could threaten the existence of the company.¹

That is the original’s first practice, “identify and define business goals”, made testable. Before the purchase order, write down three numbers the software must move (days to close the books, tickets per agent, hours to onboard a joiner), the current value of each, and the date you will measure again. A goal without a baseline cannot be missed, which is why it so often is.

02Use: 46 of every 100 licences sit idle

Zylo’s 2026 SaaS Management Index, drawn from the portfolios Zylo manages for its customers, finds that the average organisation uses 54% of its SaaS licences and wastes US$19.8 million a year on unused licences alone. The waste scales with size: US$3.8 million at companies of up to 500 staff, US$9.5 million at 501 to 2,500, US$29.8 million at 2,501 to 10,000 and US$80.6 million above 10,000.²

The firm that measures the disease sells the cure, so read the figure as reported, not counted. But the direction matches what every admin console shows when someone finally looks: seats assigned to leavers, licences bought for a project that ended, premium tiers given to people who use the basic features. The original’s “ensure user adoption” is half of the fix. The other half is removing the seat when adoption does not happen.

03Renew: prices rose whether or not the seats were used

Vertice’s SaaS inflation tracker, built on the spend it processes for customers, recorded a record 16.4% rise in SaaS prices in June 2026, easing to 12.9% in July and 12.7% in August.³ The best documented single case is Microsoft’s. On 16 February 2026 it announced list-price changes effective 1 July 2026: Microsoft 365 E3 with Teams from US$36 to US$39 a user a month, E5 from US$57 to US$60, Office 365 E3 from US$23 to US$26, Business Basic from US$6 to US$7 and Business Standard from US$12.50 to US$14. Business Premium stayed at US$22, and the frontline F1 and F3 seats rose the most.⁴

The pattern is a business model, not an accident. A flat Business Premium beside a 16.7% rise on Business Basic and 12% on Business Standard makes the security bundle look cheap at the moment of renewal. That may be the right move for you; it should be a decision, not a default. The original’s practices stopped at go-live. Renewal is where the third leak sits, and it has a date you already know.

04One 500-seat estate, worked in rupees

Put the three leaks on one bill. Take 500 users on Microsoft 365 E3 with Teams for a year, at US list and the ₹83 to the dollar TechBag’s buying guides use.

The order of the bars is the order of the work. Negotiating a few points off the rise is worth less than finding the idle seats, and both are worth less than the value question in section one if the software is not moving the number it was bought for. The original’s “customise” advice deserves a caution here too: every customisation is a cost at each upgrade and an argument against switching at renewal, so customise only where the baseline number demands it.

The original’s last practice, “measure performance”, becomes a calendar. For every contract above a few lakh a year:

1. At signature, record the renewal date, the notice period and the price-increase clause in one register, next to the three baseline numbers from section one. A renewal that auto-extends because nobody saw the notice date is the most expensive line in the register.
2. Ninety days before renewal, pull last-login data and the tier each user actually needs. Remove idle seats, move light users to a lighter SKU where the licence terms allow it, and ask the vendor for the new price in writing.
3. Sixty days before, re-measure the three numbers. If the software has not moved them, the conversation is about whether to renew at all, not about the discount.
4. At renewal, sign for the seats you counted, not the seats you had, and put the data-handling terms in the same order form.

None of this needs a new tool for an estate of a handful of large contracts. A SaaS management platform starts to pay when the application count runs into the hundreds and nobody can say who bought what; below that, an owner per contract and a shared register do the job.