Each development team chooses languages, software frameworks, databases and tools, and these choices commit those who will operate and maintain its applications after it. A tech radar, or technology radar, makes the choices of an entire organisation visible: each technology examined occupies a quadrant, which classifies it by nature, and a ring, which states the confidence granted to it.
We propose building it in the following order: set the scope and the quadrants, write down the meaning of the rings, state who proposes and who decides, publish each entry with its reason, revise on a fixed date. A radar is neither an inventory of the information system nor a catalogue of standards. Well maintained, it changes the way projects justify their architecture choices.
Public radars, maintained by companies and by a foundation
Several companies publish their radar, and an open source software foundation publishes one of a different kind.
Zalando presents its own (opensource.zalando.com/tech-radar) as a list of technologies accompanied by an assessment result, the assignment to a ring, and makes the tool that generates its visualisation available as open source, under the MIT licence (github.com/zalando/tech-radar). Backstage, an open framework for developer portals, has a Tech Radar plugin based on this code; its repository states that it is used at Spotify to visualise the official guidelines of several areas of development.
The CNCF (Cloud Native Computing Foundation), for its part, publishes reports entitled “CNCF Technology Radar”, built on a survey of developers who assess the maturity, usefulness and reliability of cloud native tools. This radar summarises the opinions of a community; a company’s radar records its own decisions.
Setting the scope and quadrants, then the meaning of the rings
The scope is decided first: the whole IT department, or one domain (data, business applications, infrastructure). It includes the technologies whose choice commits construction or operations: languages, frameworks, databases, platforms, tools, and techniques such as a way of splitting services or of testing.
The quadrants are taken from an existing radar or adapted: Zalando’s tool lets you name the four quadrants freely, and Zalando names its own Languages, Infrastructure, Datastores and Data Management. We advise not spending time on them, since they serve only for reading.
The rings, on the other hand, require a written definition, because a project will deduce from them what it is allowed to do. The table sets out the usual terms, a gloss proposed by us, the meaning given to them by Zalando’s public radar, and the consequence we propose drawing from them for a project.
| Ring (usual term) | Gloss | Meaning on Zalando’s radar | Proposed consequence for a project |
|---|---|---|---|
| Adopt | Adopt | High confidence, established use in production, broad use recommended | Default choice, with no particular justification |
| Trial | Try | Success observed on real projects, slightly higher risk | Use possible, with feedback to share |
| Assess | Evaluate | Clear potential value, worth research and prototyping | Bounded prototype, outside production |
| Hold | Suspend | Not recommended for new projects, may be kept in existing ones | Reasoned exemption for a new use, exit plan for existing ones |
Who proposes an entry, who settles its ring
Proposals are best coming from all teams, because it is they who have experience of the technologies. On Zalando’s radar, open to contributions from all engineering teams, the assignment to a ring results from change proposals discussed then voted on, led by its Principal Engineers.
We propose a short proposal file: the problem addressed, the experience gained and on which project, the solutions ruled out, the cost of an exit. The architecture body settles the ring and designates a person who writes the text of the entry. Where a project has recorded the decision that led to trying the technology in an architecture decision record, the entry refers to it instead of rewriting it.
Publishing each entry with its reason and revising on a fixed date
Each published entry carries a short text: why it is in this ring, since when, what has changed since the previous edition. It is this text that a project reads, more than the drawing. Zalando’s tool produces a static page and marks new entries and those moved up or down; the Backstage plugin places the radar in the portal that teams already consult.
Revision follows a regular pace, set by the body that maintains the radar, and is also triggered by events: end of support announced by a vendor, serious vulnerability, unfavourable feedback. A radar that guides projects can keep its entries stable from one edition to the next, provided each edition is dated and the previous ones are kept.
An entry that moves into Hold calls for a survey of the applications that use it. It is one of the signs described in our article on application obsolescence, and the resulting exit cost falls under technical debt.
The limits of the radar and its effect on architecture decisions
A radar does not list everything that runs: in the method proposed here, it retains the technologies on which a decision has to be made or recalled. The inventory of applications, flows and components falls under IT landscape mapping.
Nor does it replace a catalogue of standards. A standard is mandatory; a ring states a confidence and leaves room for the project’s judgement. The signposted and tooled path an organisation offers for a common need, a golden path, can rely on the radar without being confused with it. The target architecture and multi-year orientations belong to enterprise architecture, which a framework such as TOGAF helps to conduct, and to the IT master plan.
Its effect can be read in architecture files. A project that retains a technology from the Adopt ring has nothing to demonstrate; one that retains a technology from Hold requests an exemption, with a deadline; one that wants to try a technology from Assess plans a bounded prototype. The debate about the tool can then leave the end-of-design review and take place beforehand, calmly. A first radar can start from the technologies cited in the architecture files of recent projects: our page on IT architecture describes the governance in which these exemptions are examined.