Your specialist group gets the same questions every month. Can we use this new queue? Is that framework still supported? Why did platform reject the database another team is already running in production? You answer each one in a thread, the answer is never written down, and six months later a different team asks again. An internal technology radar is meant to fix this, and many organizations start one. Most end up as a pretty diagram that nobody checks before choosing a tool. This post shows you how to run one that teams use when they decide things: rings that say what a team is allowed to do, a short write-up for every entry, a fixed review rhythm, and clear evidence for moving an entry from one ring to another.
Where the format comes from
The format comes from the Thoughtworks Technology Radar, a snapshot published twice a year by the company's Technology Advisory Board. Entries ("blips") are sorted into four quadrants: techniques, platforms, tools, and languages and frameworks. They also sit in rings: Adopt, Trial, Assess, and a fourth ring for things to avoid, called Hold for most of the radar's history and renamed Caution in 2026. Thoughtworks openly encourages organizations to build their own radars and publishes a free tool for drawing one.
Two details of how Thoughtworks runs it matter more than the diagram. First, blips are nominated by practitioners and argued over before anything goes on the radar. Second, according to the radar FAQ, a blip by default appears for only one edition unless it moves rings. Neal Ford and Ken Mugrage have made the same point on the Thoughtworks Technology Podcast: building the radar matters as much as the finished picture. Internal radars that copy the picture without the process are the ones that die.
Why internal radars stop being used
The failure modes are predictable:
- Rings that describe opinions, not permissions. "Trial" means something different to every reader, so teams can't tell whether they need to ask anyone.
- Blips with no reasoning. A dot labelled "Hold" with no explanation reads as a ruling handed down from above. Teams don't argue with it; they ignore it.
- One person's taste. If the radar reflects what the principal engineer likes, it loses credibility the first time a team ships something good that sits in Hold.
- No expiry. Entries pile up, half are out of date, and a radar that's wrong about one thing isn't trusted about anything.
- Nobody checks it at the moment of decision. If the radar isn't linked from design review templates and the platform portal, people never open it.
Define rings as decision rules
The most useful change you can make is to say what each ring lets a team do, and who has to be involved. Here's an example to adapt:
| Ring | What it means here | What a team may do | Who must be involved |
|---|---|---|---|
| Adopt | Our default for this need. Supported by the platform, documented, with an owner. | Use it for new work without asking. | Nobody; follow the golden path docs. |
| Trial | Proven elsewhere, promising here, not yet supported by default. | Use it on a real project that can absorb some risk, after telling the radar group. | The trial team commits to writing up what it learned. |
| Assess | Worth understanding. Not ready for production here. | Spikes, prototypes, reading, conference talks. No production use. | Nobody, but share findings. |
| Hold | Don't start new work with it. | Keep existing use running; plan migration where the entry says so. | New use needs an exception from the standard's owner. |
Two rules make the table work. Adopt has to come with support. If something is in Adopt, a team using it should find templates, a runbook and someone to ask. Otherwise you are just endorsing it. Hold is about new work, not existing systems. Moving something to Hold shouldn't force a migration on every team overnight. If you do want existing use to end, say so explicitly in the write-up, with a date and a migration path. For teams that genuinely need to start something new on a Hold entry, use your normal exception process; exceptions to engineering standards that expire describes one.
A one-page write-up for every blip
No entry goes on the radar without a short write-up. This is what turns a dot into advice:
Blip: <name> Quadrant: <techniques / platforms / tools / languages & frameworks>
Ring: <Adopt / Trial / Assess / Hold> Previous ring: <ring or "new">
Owner: <a named person who answers questions about this entry>
Last reviewed: <date> Next review: <date>
What it is for
The need it serves, in one or two sentences.
Why this ring
The evidence: where we've used it, what happened,
what we compared it with.
Where it fits / where it doesn't
The contexts it suits and the ones it doesn't.
What would move it
The evidence that would promote or demote it.
Links
Docs, trial write-ups, the decision record, the golden path.
The two fields people skip are the most important. Where it doesn't fit stops an Adopt entry being read as "use it for everything". What would move it tells a team that disagrees how to change the outcome: bring evidence, not volume.
Run it on a fixed rhythm
A radar needs a cadence people can plan around. Twice a year suits most organizations, and it matches the Thoughtworks rhythm.
- Open nominations four weeks before the review. Any engineer can nominate a blip or propose a move, using the write-up template. A nomination without evidence is still welcome for Assess; promotions need evidence.
- Use a small, mixed group to decide. Five to eight people: specialists from security, platform and data, plus senior engineers from product teams who will live with the decisions. Rotate some seats every edition so the radar doesn't harden into one group's taste.
- Hold one working session, half a day. Spend the time on contested entries and ring moves, not on things everyone agrees on. Record the reasoning, including the dissent; the write-up should say "we considered X and chose not to" where that's true.
- Re-justify or retire. Borrow the spirit of the Thoughtworks one-edition rule: every entry must be reaffirmed by its owner each edition. An entry with no owner willing to reaffirm it drops off the radar or goes to Hold, which forces the stale ones out.
- Publish with a change log. Teams care most about what moved. Lead with "new this edition" and "moved rings", each with a one-line reason.
Between editions, allow one fast path: a demotion to Hold for security reasons can happen at any time, with the write-up updated the same day.
Evidence for moving rings
Arguments about rings get easier when the bar for each move is written down in advance. A hypothetical set of criteria:
| Move | Evidence required |
|---|---|
| New → Assess | Someone can say what problem it solves for us, and has read beyond the vendor's material. |
| Assess → Trial | A spike or prototype on our systems, a named team willing to run it on real work, and an exit plan if it fails. |
| Trial → Adopt | At least two teams running it in production for a quarter, a support owner, docs and a runbook, and known costs. |
| Any → Hold | A concrete problem: security, cost, an abandoned upstream, a better Adopt option, or failed trials. Plus guidance for existing users. |
The Trial-to-Adopt bar is deliberately the highest. Geoffrey Moore's Crossing the Chasm describes how early adopters tolerate rough edges that the early majority won't. Inside a company the same gap applies: the team that championed a trial will forgive things that the tenth team to adopt it won't. Requiring a second production team before Adopt tests whether the technology works for teams who didn't pick it.
A worked example
Say a team trialled a managed workflow engine for long-running order processes. After a quarter in production, their write-up reports that retries and timeouts were much simpler than with the hand-built queue consumers they replaced, but that local development was awkward and the cost model was hard to predict. Only one team uses it.
It stays in Trial. The write-up records what was learned, lists "a second team in production and a cost estimate per thousand workflows" under What would move it, and links the trial report. A second team now has a reason to try it and knows exactly what to report. Nobody has to win an argument in a meeting.
Make it part of the moment of decision
A radar is only consulted if it sits where decisions happen:
- Design review templates ask: "Which radar entries does this use? Any in Trial or Hold?"
- The platform portal links each golden path to its Adopt entry, and each Adopt entry back to the golden path.
- Exception requests for Hold entries reference the blip's write-up.
- New-joiner onboarding includes a walk through the current edition.
Be careful what you measure. Counting blips or tracking "percentage of services on Adopt technologies" invites gaming, a familiar Goodhart's law trap. Better signals: how often the radar is linked from design documents, how many nominations come from outside the specialist group, and whether the same question still turns up in your support channel every month.
Your first step
Before drawing anything, collect the last ten technology questions your group answered in chat or in design reviews. For each, write the answer as a blip using the template above, with a ring, an owner and a reason. If you can't agree on the ring, that's your first review session agenda. Ten honest write-ups are more useful than a hundred unexplained dots, and they give you something to link from the next design review. Running technology evaluation and adoption this way is a core part of innovation leadership in the PTMS study guide.