Every team has one. Somewhere in the way your work flows – from idea to review to sign-off to shipped – there is a single step that is slower than all the others. It sets the pace for everything. You can hire faster designers, write cleaner briefs, and answer emails at midnight, but if the real hold-up is that one person does all the approvals and they are swamped, none of it moves the finish line. That slowest step is your bottleneck, and until you find it, you are usually optimising the wrong thing.
The frustrating part is that bottlenecks hide in plain sight. The team feels slow everywhere, so it is tempting to push harder across the board. But a chain only moves as fast as its tightest link, and effort spent anywhere except that link is effort wasted. The good news: a kanban board is one of the simplest tools ever invented for finding the tight link. It practically points at it for you.
What a bottleneck actually is
A bottleneck – the “constraint” if you want the formal term – is the step in your workflow with the least capacity relative to the demand flowing into it. Think of a wide highway that narrows to a single lane for one stretch. It does not matter how many lanes come before or after; the whole road’s throughput is decided by that one narrow section. Cars back up behind it, and the road ahead of it runs empty.
Your work behaves exactly the same way. If your “review” step can only clear four items a day but six keep arriving, two pile up every single day. The queue in front of review grows, the columns after review go quiet, and the team’s overall output gets capped at four – no matter how much everyone upstream is producing. More input just makes the pile bigger, not the output larger.
The board shows you where it lives
Here is why kanban is so good at this. When you lay your work out as columns – say backlog, in progress, review, done – and give each card a place, the bottleneck reveals itself as a visible traffic jam. Look for the column where cards stack up on the incoming side while the column right after it sits half-empty. That gap is the tell. Work is arriving faster than that step can process it, and everything downstream is starving.
Watch it for a few days rather than judging on a single snapshot. A true bottleneck is persistent: the same column is crowded morning after morning, and cards visibly age there before they move. A one-off pile-up might just be a busy Tuesday. A constraint is the column that is always the deepest. If you already track how long cards take to move, you will often see it in the numbers too – the wait time in front of one step dwarfs all the others, which is exactly the kind of stall that drags down your flow efficiency.
Confirm it before you fix it
Before you rush to solve, make sure you have found the real constraint and not a symptom. A common trap is confusing a bottleneck with a blockage. A blocked card is stuck waiting on something external – an answer from a client, a missing asset – and it can clog a column temporarily without that step being your true limit. A bottleneck is different: the step is working flat out and still cannot keep up. One is waiting; the other is overloaded.
The quick test is to ask what happens if you pour more work in. Add capacity upstream for a week. If the pile in front of one particular column just grows while the finish rate stays flat, you have confirmed the constraint. That column is the governor on your whole engine, and it is the only place where an improvement will actually raise your output.
Fix one: stop feeding the jam
The instinct when you find a bottleneck is to try to speed it up. That is step two. Step one is cheaper and faster: stop overloading it. Piling more onto an already-maxed step does not make it go faster – it just grows the queue, ages the cards, and adds the overhead of managing a bigger pile. Slowing the flow into the constraint feels counterintuitive, but it steadies everything.
This is precisely what work-in-progress limits are for. Put a cap on the columns feeding your bottleneck so upstream work cannot flood it. When the limit is hit, people upstream do not start yet another item – they help clear the constraint instead. The board stops hiding the problem behind a mountain of started-but-unfinished work, and the whole team’s attention naturally points at the one place that matters.
Fix two: give the constraint more room
Once you have stopped drowning the bottleneck, you can widen it. There are usually three levers, in rough order of effort. First, protect it: shield the constrained step from interruptions and low-value tasks so its scarce capacity goes only to the work that needs it. If approvals are the choke point, do not also make that person chase invoices during approval time.
Second, offload it: pull any part of the step that does not truly require the constrained person or tool and hand it to someone with spare capacity – prep, formatting, first-pass checks. Second-and-a-half, subordinate to it: have the rest of the team pace themselves to the constraint and pitch in to relieve it rather than racing ahead and building inventory the bottleneck will never reach. Only when those are exhausted do you reach the third and most expensive lever: add real capacity – another reviewer, a better tool, more hours. Do it in that order and you often solve the problem without spending a dollar.
Watch the bottleneck move
Here is the twist nobody warns you about: fixing a bottleneck does not eliminate bottlenecks, it relocates them. The moment you widen your review step, some other column becomes the new slowest link, and the traffic jam quietly shifts down the board. This is normal and healthy – it means you are genuinely raising your capacity – but it also means the work is never quite done.
So make finding the constraint a habit, not a one-time rescue. Once a week, stand in front of the board and ask a single question: where is the pile today? Then aim your team’s improvement effort at that one spot and nowhere else. That is the whole discipline – find the slowest step, relieve it, and repeat as it moves. A team that keeps doing this steadily gets faster without anyone working longer, and a simple kanban board is usually all you need to see exactly where to push next.