Kanban in a small team usually dies the same way: someone sets up a board, lays the cards out neatly, and two weeks later the “In progress” column holds 40 tasks, half of them from last month, and everyone is back to assigning work in the chat.
That isn’t a problem with kanban. It’s a problem with a board built as a display case rather than a working tool. Here’s how to assemble, in a single working day, a board that survives week three.
What kanban actually is
Kanban isn’t “a board with columns”. It’s three rules:
- Visualise the work. Everything the team is doing is on one board. Not “mostly” — everything. If a task exists only in someone’s head or in a chat thread, then as far as kanban is concerned it doesn’t exist.
- Limit work in progress (WIP). Each stage can hold only a limited number of tasks at once. This is the central rule, and it’s the one skipped in 90% of rollouts.
- Manage flow. Watch not who is busy, but where tasks get stuck — and fix that specific place.
Drop rule two and you have an ordinary to-do list spread across three columns. It won’t shorten how long tasks take, and it won’t reveal bottlenecks.
Why kanban suits small teams better than Scrum
Scrum needs sprints, estimates, retros, a scrum master and a stable team composition. In a team of six where half the people are on client revisions and the other half are firefighting, two-week planning breaks on day one.
Kanban requires neither sprints nor story points. It works with a continuous, unpredictable flow of work: agencies, studios, support, a small product, field services. You can start tomorrow with no team training.
The honest downside: kanban gives you no ready answer to “what will be finished by the 20th”. If your work is project-based with hard release dates, you’ll need to supplement the board with at least a simple milestone plan.
The one-day plan
Step 1. Gather all the work in one place (60–90 minutes)
Don’t design columns before you’ve seen the real volume. Collect everything currently in flight into a single list: from chats, email, spreadsheets, sticky notes and memory. Bring the team in — a manager working alone will recall about sixty per cent of it.
This step almost always produces the first unpleasant surprise: there are three times more open tasks than anyone believed. That, not laziness, is the main reason deadlines slip — attention is spread too thin.
If most of your tasks are born in conversations, this step can be partly automated: the bot recognises tasks in chats and offers them for confirmation, so you don’t have to comb through the backlog of discussions by hand.
Step 2. Draw columns that match your real process (30 minutes)
Don’t copy someone else’s board. The rule is simple: a column is a stage where a task can wait. If waiting never happens at that stage, it doesn’t need its own column.
A universal minimum for a team under 15:
- Backlog — accepted, not started.
- In progress — someone is working on it right now.
- In review — done, waiting on a client, a manager or a colleague.
- Done — closed.
Extend it only when a stage genuinely hurts. The usual worthwhile additions: “Waiting on client” (if revisions regularly hang for weeks for external reasons) and “Ready to deliver” (if results accumulate and ship in batches).
The opposite danger is a nine-column board. Every extra column is one more place to move a card to instead of finishing it.
Step 3. Set WIP limits (15 minutes)
The most important step, and the most skipped. A starting formula: the limit on “In progress” = number of doers + 1. Five people, six tasks in progress, no more. Not “six on average” — strictly: nothing new enters until something leaves.
The first few days this will be irritating. That’s the point: the limit makes visible what used to be smeared across the week. When a seventh task can’t be taken, the team is forced to finish what’s open, and the manager is forced to prioritise explicitly instead of quietly dumping everything at once.
The side effect this is all for: the time an individual task takes goes down. Five tasks run in parallel finish later than the same five run in sequence, simply because context switching eats the difference.
Step 4. Agree what “done” means (20 minutes)
Half of all stuck cards are stuck not because the work isn’t finished, but because nobody can tell whether it is. Write one sentence per column stating the condition under which a card moves on.
For example: “In review → Done: the client has confirmed in writing and the files are in the project folder.” Without that, “Done” degrades into “I think I finished”.
Step 5. Fill the board and agree the rules (40 minutes)
Move over everything you gathered in step 1. At the same time introduce three rules, without which the board will fall apart:
- No task enters “In progress” without an owner and a date. Otherwise “someone will do it” means nobody will.
- Tasks are assigned on the board only. An instruction in a chat isn’t an assignment. This is the hardest rule and the most important one.
- The doer moves the card, not the manager. Otherwise the board becomes a report for the boss, and the team stops trusting it.
Step 6. Add one five-minute ritual (10 minutes)
A daily stand-up of 5–10 minutes — but not “what did you do yesterday”. Read the board right to left and discuss only what’s stuck: what hasn’t moved in two days, what’s blocking it, who needs help. Everything else is visible on the board and doesn’t need saying aloud.
Without this ritual the board turns into a dump within two weeks, simply because nobody ever discusses the stuck cards.
Five mistakes that kill a board
1. A board built for reporting. If the only person who needs the board is the manager, the team will maintain it as a formality. The value has to land with the doer: it’s clear what to do next, and urgent work doesn’t keep landing on top of urgent work.
2. No WIP limits. The most common one. Without a limit, kanban is indistinguishable from a to-do list and delivers none of its main effect.
3. Running the board and the chat in parallel. Two sources of truth always resolve the same way: the chat is more convenient, and within a month the board is empty. Either everything is on the board, or don’t start. The only thing that rescues this is automation — when an instruction in a conversation becomes a task by itself instead of needing manual transfer.
4. Too many columns and fields. Mandatory fields like “task type”, “estimate” or “epic” only raise the entry barrier at the start. Title, owner, date is enough.
5. An eternal backlog. A backlog where everything lands and nothing is ever removed stops being read. Once a month, close whatever hasn’t moved in 60 days: if nobody has needed it for two months, nobody needs it.
How to tell in two weeks whether it’s working
Three metrics you can count by hand:
- Lead time — the average number of days from a task appearing to “Done”. The headline metric. It should be falling.
- Number of tasks in progress — is it staying inside the limit, or is the limit constantly broken? Constant breaches mean either the limit is unrealistic or more work is being pushed in than the team can absorb.
- Share of overdue tasks — is it going down.
If your tool counts this for you, no manual tally is needed: in Gosudarynya deadlines and per-person workload are visible in analytics, and board, list and table are three views of the same tasks, switched with one click.
Plus one non-numeric sign that matters more than any metric: after two weeks the team discusses work while looking at the board, rather than recounting it from memory. If they’re recounting from memory, the board hasn’t stuck.
What comes next
Once the flow is stable, it’s worth adding depth: swimlanes to separate urgent from planned work, card-age limits, separate boards per project. But all of that comes after the basics work. Rolling out “grown-up” kanban straight away in a team of six is a reliable way to bury it.
If you’re still deciding what to run the board on, the criteria are in how to choose a task tracker, and a comparison of specific services is in our review of Russian task trackers.
Bottom line
Kanban in a day is realistic as long as you don’t try to build the perfect process. Gather all the work, make four columns, set a WIP limit, agree on “done”, and ban assigning tasks in the chat. The board will tell you the rest within two weeks.
If your team lives in chats and half the work is born there, take a look at Gosudarynya: the bot pulls assignments out of Telegram and MAX conversations, and the board, CRM and finance are already in the box. 7 days free, no card.