Application obsolescence produces no incident on the day it begins. The application delivers the same service as the day before, while a vendor closes a version, a library stops being maintained or the last person who knows a module changes jobs.
It can nevertheless be recognised by observable signs, on five layers: the technical platform, dependencies, skills, documentation and the ability to evolve. Each sign may reveal a distinct risk, of security, continuity or immobility, and calls for a response of its own, from a version upgrade to deeper modernisation. The diagnosis is made application by application, on the evidence.
An old application is not necessarily obsolete
Age says nothing on its own: our entry on technical debt points out that an old application, tested and understood, is not necessarily indebted. It treats obsolescence as a form of debt, born of the ageing of components.
Here we extend the term to what ages around the code: knowledge of the application and the means of making it evolve. The diagnosis therefore looks for dated facts that anyone can check: a support phase, a version number, a number of people, the date of a document, an observed lead time.
The technical platform and its support phases
The platform brings together the operating system, the database, the application server, the language and its runtime environment. When the vendor of a component publishes a lifecycle policy, the expression “end of support” may cover several deadlines in it, as two examples show.
Microsoft’s Fixed Lifecycle Policy distinguishes Mainstream Support, then, for some products, Extended Support, during which security updates continue while the vendor no longer accepts requests for changes. After the end of support, security updates go through a separate programme (Extended Security Update). Red Hat describes, for Red Hat Enterprise Linux, three phases: Full Support, Maintenance Support, then the Extended Life Phase, during which neither bug fixes nor security fixes are provided, unless an extension offering is subscribed separately. These two pages were read on 12 November 2026.
The sign to look for is therefore the phase each component is in, with the date of the next one. A component that has left the last covered phase may reveal an exposure: a published vulnerability may no longer be fixed in it. For the products within its scope, the Cyber Resilience Act provides for a support period borne by the manufacturer, whose content and application timetable our entry sets out.
Dependencies that age out of sight
An application embeds libraries and components that its team did not write and that no support contract necessarily covers. The signs are read in the list of these components: a project that no longer publishes releases, a lag of several major versions, a published vulnerability with no fix in the branch used. Third-party component analysis, described in our article on application security, identifies published vulnerabilities.
Other dependencies are external to the code: a partner’s interface whose version is announced as retired, an abandoned protocol or format, a browser or workstation required in an old version.
A version lag may reveal a workstream that is growing. Each skipped major version may add its breaking changes to those of the next one, and a routine update may turn into a project.
The knowledge that leaves with people
This layer escapes technical inventories. It is established by questioning the team, module by module: how many people know how to modify this module without help, and how many know how to bring it back into service after an outage? When maintenance is entrusted to a service provider, under third-party application maintenance, the same question is put to them, together with that of the contract’s expiry date.
Documentation is judged on two dates: that of its last update and that of the last release. A gap of several years suggests that the code has become the only reliable description of the business rules. The absence of any record of past choices, of the kind kept in an architecture decision record, is another sign: everyone observes what the application does, without anyone being able to say why.
What these signs may reveal is a continuity risk. The application may become impossible to modify before it is exposed to a vulnerability, and any later transformation will need this knowledge.
The ability to evolve, read from recent requests
The last layer is read in the history of business requests. How long separated a simple request from its production release? Which requests were refused or postponed for a technical reason? What workarounds have grown up around the application, a spreadsheet kept in parallel or re-entry from one screen to another?
Delivery gives other signs: few production releases, manual and dreaded, tests replayed by hand, an interface that the application cannot expose to another system. Our article on measuring technical debt gives the corresponding indicators and their limits.
This is the layer that a business department perceives, the one where obsolescence translates into unmet requests.
The response that application obsolescence calls for, depending on the layer affected
The responses are ranked by increasing effort.
- A platform at end of support calls for a version upgrade or a change of platform. In the meantime, keeping it is decided as a dated derogation, accompanied by protective measures such as isolating the component on the network.
- Lagging dependencies call for regular, tool-supported updating, and the replacement of abandoned libraries.
- Knowledge that is being lost calls for a transfer, documentation rebuilt from the code and tests that record the existing behaviour. This work comes before any transformation.
- A lost ability to evolve calls for a transformation of the application itself. Our article on legacy application modernisation compares the options, and the one on mainframe migration focuses on applications carried by a large system.
When several layers are affected on a little-used application, retirement is discussed before any expenditure. The choice of a strategy and the conduct of the workstream fall under application modernisation.
The survey joins the application register or the IT landscape map, where it brings to light the critical processes that rely on a component out of support: a finding that falls under IT architecture. This survey ages too: it carries its date and is redone at a fixed interval, since a support phase, a headcount or a contract can change without the application having moved.