Pick a tracker for a six-person team by reading feature matrices and you will end up with Jira, because Jira has every feature. That is the trap. For a team that cuts a release every week, the tracker's job is narrow: keep the next five days legible and stay out of the way. Most of what shows up in a comparison grid — portfolio roadmaps, capacity planning, custom workflow schemes, approval gates — is dead weight at that size. Worse, it is dead weight that somebody on the team has to maintain forever.
The useful question is not "which tool does more?" It is "which tool costs the least attention per week?" That reframing changes the answer for most small teams.
What a weekly cycle actually demands
Before comparing products, write down what one week of work needs from software. For a team shipping every Friday, the list is short:
Issue creation fast enough that you do it mid-conversation. If filing a bug takes a minute and four required fields, people stop filing bugs. They go to Slack instead, and the tracker becomes a partial record of reality — which is worse than no record, because you start trusting it.
A cycle object that closes and reopens itself. Weekly cadence means 52 cycle boundaries a year. If closing a sprint and starting the next one is a manual ceremony, someone spends an hour a month clicking through it, and it gets skipped the week you actually needed it.
A link between a merged PR and a closed issue that nobody has to click. Branch-name conventions that auto-move issues to done are not a luxury. They are the difference between a board that reflects the repo and a board that lies by Wednesday.
Zero standing admin. No permission schemes, no workflow editors, no field configuration screens. Any surface that can be configured will eventually be configured badly by whoever had a bad afternoon.
Everything else — story points, epics containing epics, a QA handoff status, custom fields for the customer-facing severity — is optional at this size and usually net negative. Each one is a decision you re-litigate every quarter.
Linear, Jira, and Height side by side
Linear is the one that assumes your process. Cycles, triage, projects, and a single global workflow come preconfigured, and the customization you get is deliberately shallow. For a team of three to twenty-five engineers with a normal software process, that constraint is the feature — there is nothing to tune, so nobody tunes it. The cost shows up when your process genuinely is unusual: a hardware dependency, a regulated release sign-off, a support queue that needs its own states. Linear will make you approximate.
Jira is the opposite bet. Nothing is assumed, everything is configurable, and the ceiling is high enough that companies with thousands of engineers run on it. The free tier covers small teams, which makes the initial economics attractive. The real price is measured in configuration drift: six months in, you have three issue types nobody agreed on, a workflow with a status that only one person understands, and a board filtered by a JQL query somebody wrote in a hurry. Small teams that do well on Jira are the ones with the discipline to leave the defaults alone — and if you have that discipline, you probably did not need the configurability.
Height positions itself around autonomous project management: the pitch is that an AI layer handles the chores humans skip, such as deduplicating issues, updating statuses from activity, and keeping the backlog from rotting. Judge that on your own workload rather than on the marketing, because the value depends entirely on whether your team's specific chore is one the automation actually covers. The structural tradeoff is clearer: Height's table-and-view model is more flexible than Linear's and far lighter than Jira's, with a correspondingly smaller integration ecosystem.
All three vendors have changed plan structure and per-seat pricing more than once, and annual-versus-monthly gaps are significant. Treat any price you read in an article — including this one — as a starting point, not a quote. Pull the current numbers from each vendor's pricing page for your exact seat count, and check whether the tier you need for SSO or audit logs is two steps up rather than one.
Choosing in an afternoon
Three rules cover most small teams:
Default to Linear if you are under roughly 25 engineers, greenfield, and your process is ordinary. You will spend zero hours on configuration, the cycle mechanics match a weekly cadence without setup, and the git integration means the board stays honest without anyone maintaining it.
Choose Jira if you already live in the Atlassian estate, or if non-engineering stakeholders need to file and track work. Compliance requirements, a support org that needs Jira Service Management, or an existing Confluence corpus all tip the decision. The migration cost of leaving later is real, so this is a choice worth making deliberately rather than by inertia.
Try Height if your pain is backlog maintenance rather than execution. Teams that ship fine but drown in stale issues, duplicates, and untriaged inbound are the ones its automation targets. Run a two-week trial against your actual backlog, not a demo project.
One more option deserves mention: for teams of three or four shipping weekly, a database in a general-purpose workspace tool is often enough, and it consolidates specs, meeting notes, and the issue list in one place. That stops working somewhere around six people or the first time you want PR-to-issue automation — but plenty of teams reach for a dedicated tracker a year before they need one.
The cost nobody budgets for
Whatever you pick, the expensive part is not the subscription. It is the second migration.
Every tracker exports issues. Almost none of them export the things that matter: comment threads with the reasoning behind a decision, the link between an issue and the PR that closed it, attachments, and stable issue IDs that appear in a thousand commit messages and Slack links. After a migration, PROJ-412 in a two-year-old commit points nowhere.
So make the decision once, on the assumption that you will keep it for three years, and weight it accordingly. Being slightly wrong about which tool fits your process today costs less than being right twice.
Originally published at pickuma.com. Subscribe to the RSS or follow @pickuma.bsky.social for new reviews.
Top comments (0)