Agile & Project Management

Project Management Tools: Choosing and Using Them Well

Beginner9 min readUpdated

A tracking tool records decisions; it does not make them, and most tool complaints are really process complaints. This guide compares the main options by the situations they fit, explains when a board, a backlog or a Gantt view is the right view, warns about workflow customisation, and covers pipeline integration, useful reports and data hygiene.

What you will learn

  • Separate decisions a tool should record from decisions a team must make
  • Compare Jira, Azure DevOps, GitLab, Trello and Redmine by the situations each suits
  • Choose between board, backlog, timeline and roadmap views for a given question
  • Avoid the workflow and field customisation traps that make tools unusable
  • Connect tracking to the delivery pipeline and build reports people act on

Every team that has ever complained about its tracking tool was, in most cases, complaining about its process. The tool made something visible — a queue of half-finished work, an unclear approval step, a backlog nobody orders — and the visibility was uncomfortable. Choosing a different product does not resolve that, which is why tool migrations so often deliver a year of effort and the same problems in new colours.

That said, the choice does matter at the margins, because tools have opinions. Some assume a single team with a board; some assume a portfolio with dependencies and dates; some assume the work is inseparable from the code. This guide is about matching the opinion to your situation, then configuring the thing so it helps rather than obstructs.

What a tool should and should not decide

A tracking tool is a shared record. Its legitimate jobs are keeping one authoritative list of the work, showing the current state honestly, preserving the history of what was decided and when, connecting work to the code and deployments that implemented it, and answering questions about flow that nobody can answer from memory.

What it must not do is decide the process. If the workflow has seven states because the tool template had seven states, the team has adopted a stranger's process by accident. If items cannot move to "in progress" without an estimate because a field is mandatory, an estimation policy has been set by configuration rather than by agreement. The order of operations is always the same: agree how you want to work, then configure the tool to reflect it, then revisit both.

The corollary is that a tool cannot fix ownership. A backlog is only ordered if somebody orders it, whatever the software allows — see product owner and backlog management for what that involves.

The main options, by fit

Feature lists are close to useless for choosing, because the mainstream products have converged. What differs is the situation each one is comfortable in.

ToolComfortable whenUncomfortable whenNotes
JiraSeveral teams, formal workflows, portfolio-level rollups, an ecosystem of integrationsA small team wants something simple; nobody owns the configurationThe default in mid-size and large organisations; power comes with an administration burden
Azure DevOpsThe organisation is already on Microsoft tooling and wants boards, repositories, pipelines and test management in one productYou want best-of-breed in each area rather than one integrated suiteStrong out-of-the-box link between work items, branches and builds
GitLabDevelopment-centric teams who want issues, code, review and CI in a single applicationNon-technical stakeholders are the primary usersIssue boards, epics and milestones are lighter than Jira but tie tightly to merge requests
TrelloA small team, a simple flow, low ceremony, work that fits on cardsYou need hierarchy, dependencies, reporting or many teamsExcellent at what it does; people outgrow it and then fight it
RedmineSelf-hosting is required, budget is zero, or data must stay on premisesYou want a polished modern interface without effortOpen source, long-lived, includes Gantt and time tracking; needs an administrator

Two honest observations. First, most teams could work well in any of these; the difference in outcomes between two well-configured tools is far smaller than the difference between a well-configured and a badly configured instance of the same one. Second, the cost that dominates is not the licence but the administration and the migration, so switching for a feature is almost never justified while switching because nobody maintains the current instance often is.

Dedicated scheduling products remain a separate category. Where a delivery genuinely needs a dependency network, a resource histogram and a critical path, a tracking tool's timeline view is a bar chart rather than a schedule model, and the distinction matters for the reasons set out in schedule management.

Views, and which question each answers

Most tools offer the same handful of views, and confusion about their purpose causes more trouble than any missing feature.

A board answers "what is the state of work in progress right now". Columns are states, cards are items, and the value comes from limiting how many cards may sit in each column — a work-in-progress limit is what turns a board from a status display into a flow control. A board with no limits and forty cards in progress is a wall of intentions.

A backlog answers "what should we do next". It is an ordered list, and the ordering is the content; a backlog displayed as a flat unordered pile of hundreds of tickets is not a backlog.

A timeline or Gantt view answers "when will things land relative to each other". It suits fixed external dates, hardware lead times and multi-team dependencies, and it becomes misleading when it is drawn by hand and does not recompute when a bar moves.

A roadmap view answers "what are we aiming for over the next few horizons" and works best when it shows goals rather than dated features.

A report or dashboard answers "how are we doing", which is only a useful question if the answer changes a decision.

Pick the view per question rather than per person's preference, and be suspicious of any request for a view whose purpose cannot be stated as a question.

Customisation traps

Configuration is where tools go wrong, and the failure modes are consistent enough to list.

  • Workflow sprawl. Ten states, six transitions each, three of them requiring approval. Every extra state is a place for work to sit and a decision someone must make. Three to five states cover most teams; add one only when a real handoff exists and its absence has caused a problem.
  • Mandatory field creep. Each field seemed reasonable to the person who asked for it. Collectively they make creating an item a five-minute chore, so people paste placeholder text into fields they do not understand and the data becomes worse than none.
  • Per-team divergence. Eight teams, eight different workflows and issue types, and no possibility of a meaningful cross-team view. Agree a common core and allow local additions.
  • Reporting-driven configuration. Fields that exist only to populate a management chart. The team gets no value from maintaining them, so they do not maintain them accurately.
  • Automation that surprises people. Rules that reassign, transition or close items invisibly. Automation is excellent for mechanical steps and corrosive when the board changes for reasons nobody can explain.
  • Permission lockdown. Configurations where developers cannot edit their own items and must ask an administrator. Every such restriction moves work out of the tool and into chat, where there is no record at all.

The general principle: the cost of configuration is paid on every single interaction, forever, while the benefit is usually a report someone requested once.

Connecting to the delivery pipeline

The most valuable integration is the cheapest: put the item key in the branch name and the commit message, and the tool links the work to the code automatically. From there, the build system can report which commits and therefore which items are in a given deployment, and the item can show whether it reached staging or production. The branching workflow guide covers the naming conventions that make this reliable.

Three further connections repay the effort. Automatic transitions on merge — moving an item to "in review" or "done" when its merge request is approved — remove a status update nobody wants to do manually. A link from a deployment back to its items produces release notes for free. And connecting incident tickets to the change that caused them builds the data needed to measure change failure rate honestly, which is where tracking meets the delivery pipeline as a feedback loop rather than an administration system.

Resist the temptation to automate transitions that represent a human judgement. "Done" usually means someone verified something; a rule that sets it on merge quietly redefines the team's definition of done.

Reports worth having, and data hygiene

Every tool ships more charts than anyone needs. A small set actually informs decisions.

ReportQuestion it answersWatch out for
Cumulative flow diagramWhere is work piling up, and is work in progress growing?Needs consistent states to be readable
Cycle time distributionHow long do items take, and how variable is that?An average hides the tail; look at the spread
Throughput per periodHow many items finish, and is it stable?Only comparable if items are split consistently
Ageing work in progressWhich items have been open unusually long?The single most actionable daily report
Burn-up with scope lineAre we progressing, and is scope growing?Burn-down hides scope growth; burn-up shows it
Deployment and failure dataDoes finished work actually reach production?Requires the pipeline integration above

Notice that velocity charts are absent. They are fine as a private input for a team, as discussed in the Scrum framework guide, and harmful the moment they appear on a dashboard that compares teams.

Every one of these depends on data hygiene, which is unglamorous and decisive. States must be updated when they change, not reconstructed on the Friday before a report. Items must be closed when they are done, and stale items must be deleted rather than left to accumulate — a backlog with three years of untouched tickets is a backlog nobody trusts. Duplicates must be merged. And the estimate or size recorded should be the one the team actually agreed, never one edited afterwards to make a chart look better.

Selecting and rolling out

  1. Write down the questions you need answered

    Not features — questions. "Which items are blocked", "when will this release land", "how much of our capacity goes to defects". A tool that cannot answer your questions is disqualified regardless of its feature list.
  2. Model your actual process first

    Draw the states work really passes through today, including the informal ones. If the process is unclear, fix that before shopping; a tool will encode whatever confusion you bring to it.
  3. Trial with one real team on real work

    Two to four weeks, live work, not a sandbox. Evaluate speed of everyday actions, quality of the board, and how much administration it demanded.
  4. Configure minimally

    Default workflow, few fields, few issue types. Add configuration only in response to a demonstrated problem, and record why each addition exists so it can be removed later.
  5. Decide who owns the configuration

    A named administrator with time allocated. Instances without an owner accumulate custom fields and abandoned projects until nobody trusts the data.
  6. Migrate deliberately, not completely

    Move open work and recent history; archive the rest read-only. Full-fidelity migration of years of closed tickets costs far more than it returns.
  7. Review after a quarter

    Which fields are empty, which states are never used, which reports nobody opens. Deleting configuration is as valuable as adding it and happens far less often.

Key takeaways

  • A tool records decisions and makes state visible; it cannot supply ownership, ordering or a process the team has not agreed.
  • The mainstream products have converged on features, so choose by the situation each is comfortable in and by who will administer it, not by comparison tables.
  • Match the view to the question: boards for work in progress with limits, backlogs for ordering, timelines for external dates, roadmaps for goals.
  • Configuration is paid on every interaction — keep workflows to three to five states, resist mandatory fields, and never configure for a report the team gains nothing from.
  • Link items to branches, commits and deployments; that one integration produces release notes, traceability and honest delivery metrics almost for free.
  • Reports are only useful with clean data, so close, merge and delete relentlessly, and keep velocity out of any dashboard that compares teams.

Sources and further reading

Frequently asked questions

Is a physical board still better than a digital one?
For a colocated team, often yes. A wall is always visible, costs nothing to change, invites people to stand around it and cannot be filtered into a view that hides bad news. Digital tools win when anyone is remote, when history and reporting matter, and when the work needs to link to commits and pipelines. Many teams run both, with the wall as the working surface and the tool as the record.
How much of the tool should we customise?
Start with the defaults and change something only when a specific problem has occurred twice. Most heavily customised installations were configured in the first month by people guessing at their own process, and the resulting workflows outlive the reasons for them by years. A ten-state workflow with mandatory fields collects data nobody reads and slows every update.
Should tasks be tracked separately from user stories?
Only if the team gets value from it. Sub-tasks help when several people work one item and need to divide it visibly, and they become pure overhead when a developer creates four sub-tasks after the fact to satisfy a report. The unit that must be tracked accurately is the one the product owner orders and the team finishes; anything below that is the team's own business.
Can one tool serve both an agile team and executive reporting?
Usually, but not with the same views. Teams need a board and a backlog; executives need outcomes, milestones and forecasts. The failure mode is forcing the team to maintain fields whose only purpose is a management report — which corrupts the data, because people fill required fields with whatever passes validation. Derive management views from data the team maintains for its own reasons.