Software portfolios rarely become complicated because somebody deliberately designed them that way. They accumulate. A team buys a specialist tool for one need, another department solves a similar problem elsewhere, a platform is replaced but not fully retired, and a few years later nobody can explain which applications are essential, duplicated or barely used.
Zylo's 2026 SaaS Management Index reports that business units control most SaaS spend and that unused licence capacity remains material across the organisations in its dataset. The exact percentages will vary by company, but the operating lesson is broader: decentralised buying and weak application ownership make waste difficult to see until renewal time.
The safest software to remove is not the tool with the lowest login count. It is the tool whose business capability, data, integrations and control responsibilities can be retired or moved deliberately.
Start with capability, not vendor
Begin by asking what business capability each application provides. Does it manage customer records, approvals, project delivery, documents, reporting, billing, service tickets or employee workflows? Two products can look different while doing substantially the same job. Conversely, a lightly used application may carry one small but critical control that no other system currently provides.
Create an inventory that links each application to an owner, users, annual cost, renewal date, core capability, integrations, important data, security or compliance obligations, and known workarounds. Without that map, rationalisation becomes a procurement opinion rather than a systems decision.
Use six removal tests
- Business value: the capability is no longer needed or is duplicated elsewhere.
- Usage: sustained usage is low relative to the licences or capacity being paid for.
- Fit: teams work around the system because it no longer matches the process.
- Architecture: another strategic platform can perform the function with fewer handoffs or integrations.
- Risk: the product is unsupported, poorly governed or creates unnecessary data exposure.
- Economics: the total cost of keeping it exceeds the cost and risk of migration or retirement.
No single test should decide the outcome. A low-cost app that creates duplicate customer data can be more expensive operationally than its invoice suggests. A high-cost platform may still be justified if several revenue-critical workflows depend on it.
Do not confuse low adoption with low value
Low usage can mean the software is unnecessary. It can also mean the implementation was poor, training was weak, the workflow was never redesigned, or users were never given a reason to change. Removing the tool without diagnosing the cause can simply move the same problem into another application.
This is why rationalisation should distinguish between remove, right-size, reconfigure, integrate and replace. Those are different interventions with different costs.
Design the exit before cancelling
Before termination, identify what must happen to records, reports, automations, integrations, user access, historical evidence and downstream processes. Decide which data must be migrated, which can be archived, and which can be deleted under the organisation's retention rules. Replace integrations before disconnecting the source. Give users a clear destination for the work they used to perform in the retiring system.
Then time the exit around contract terms and operational risk. Renewal dates are useful forcing functions, but they should not be the first time the business asks whether the application still belongs in the architecture.
What better looks like
A healthier application portfolio is smaller only when smaller is better. It has explicit owners, known capabilities, understood data flows, documented integration dependencies and regular review points. New tools enter through a clear decision process, and old tools leave through a controlled retirement process.
If the organisation cannot produce a reliable application inventory or explain why several overlapping tools exist, the first step is not cancelling licences. It is establishing the current state.