Capacity and Flow
How work actually flows through organisations and why capacity discipline is the master lever.
The capacity illusion
When teams are stretched, the conversation defaults to headcount.
“We need more engineers.” “Sales needs two more reps.” “We’re underwater - can we get budget for contractors?”
The instinct is understandable. People are visibly busy. Backlogs are growing. Delivery is slipping. Surely the answer is more capacity?
Usually, it isn’t.
The real constraint isn’t capacity. It’s flow. How work moves through the system, how much is in flight at once, and how cleanly requests arrive.
Most organisations are running far below their theoretical capacity - not because people aren’t working hard enough, but because the system is fighting them.
Too much work in progress creates queue effects that slow everything down. A bias toward starting new work instead of finishing existing work fragments attention. Hidden variety in how work arrives consumes time that should be spent executing.
These aren’t execution problems. They’re system problems. And adding headcount often makes them worse, not better.
This essay lays out the capacity disciplines that actually work. Not wishful thinking about “doing more with less,” but practical levers that let teams operate closer to their actual capacity.
The stakes are higher than most leaders realise. Flow isn’t just about delivery speed. It’s about learning speed. Teams that finish work quickly learn faster. They get feedback sooner. They compound improvements.
Teams that fragment across too much work-in-progress learn slowly. Feedback arrives late. Improvements don’t compound. The organisation drifts.
Capacity discipline is the master lever. Get it right and you don’t just deliver faster. You learn faster, adapt faster, and build resilience into the system.
The hidden queue
In knowledge work, queues don’t sit on factory floors. They pile up invisibly - in inboxes, contracts awaiting review, features half-built.
Because they’re unseen, managers underestimate them. A request lands, a project starts, a colleague asks for help. Saying yes feels incremental. Surely one more won’t make much difference?
But it does.
Each extra item lengthens the wait for everything else. Delivery slows, stress builds, and costs ripple out to customers. The “cheap” yes isn’t cheap at all.
This isn’t just a hunch. Little’s Law makes the relationship clear.
Lead time = Work in Process รท Throughput
Think of a coffee shop. If one barista serves 30 drinks an hour and there are 15 drinks in the queue, the average wait is 30 minutes. Add five more orders without extra capacity and the average wait jumps to 40 minutes.
The extra yes doesn’t just delay those five customers. It slows everyone.
Knowledge work is no different. Each new commitment stretches every other delivery already in progress.
The costs layer up.
At the individual level, people work longer hours and carry hidden stress. At the organisational level, teams chase local priorities and firefight instead of improving flow. At the business level, customers wait longer, sales cycles drag, reputation erodes.
These costs rarely trace back to the yes that triggered them. They surface later, detached, as if from nowhere.
Perfect WIP control is rare. Priorities shift, resources are constrained, politics intrude. But discipline starts with a habit: before saying yes, ask what it does to the queue.
Is this the piece of work worth slowing everything else for?
Making WIP visible changes the conversation. Contracts in review, invoices in flight, features in development - suddenly the hidden queue is no longer invisible. The maths meets lived reality.
Most teams, when they make WIP visible, discover they’re carrying two to three times what they can process efficiently. The backlog isn’t because they’re understaffed. It’s because they keep saying yes.
The instinct is to work harder. The system solution is to say no more often.
This isn’t about laziness or lowering standards. It’s about recognising that every yes has a queue cost, and that cost is often higher than the value of the work itself.
At heart, this is about leadership discipline. A generous yes feels easy in the moment. The harder move - and often the wiser one - is knowing when that yes is too expensive.
The starting bias
Leaders optimise for starting. New ideas feel like progress.
Sharing the next initiative, launching the next project, kicking off the next piece of work - it looks like momentum. It feels decisive. It’s visible.
But the bottleneck isn’t generating ideas. It’s finishing them.
Teams fragment when work arrives faster than they can process it. The maths is simple: start ten things and finish two, you learn less than starting three and finishing three.
But starting feels productive. Finishing is slower, less visible, harder to showcase.
This creates a pathology. Leaders share the next idea whilst the last few are still working through design, delivery, or debate. From the outside it looks creative. Inside it’s idea diarrhoea - a well-intentioned flood that leaves everyone sprinting, no one finishing.
Every team has a limit to how much new work it can absorb. Push past it and momentum doesn’t build - it fragments. Half-finished projects, blurred priorities, rising friction.
Manufacturing learned this long ago. Feed material into a line faster than it can be processed and the line slows. Knowledge work is the same, only the raw material changed - ideas, requests, initiatives. Arrive too quickly and the system clogs.
The fix isn’t complicated, but it requires discipline.
First, separate capture from release. Write ideas down so you don’t lose them, but don’t immediately put them into flight. Capturing an idea doesn’t mean starting it. Create a holding space - a backlog, a roadmap, a parking lot. Ideas stay there until there’s capacity to execute.
Second, limit work-in-progress explicitly. Set a number - three active initiatives, five projects in delivery, whatever matches team capacity - and enforce it. When something new needs to start, something else needs to pause or finish.
This feels constraining. It is meant to. The constraint is what creates flow.
Third, measure completions, not starts. Track what finishes, not what begins. Completion rate is the signal. Starts are just motion.
Most teams, when they start tracking this, discover their start-to-finish ratio is embarrassingly high. They’re launching five things for every one they complete. That’s not execution. It’s churn.
Count what you’ve started in the last quarter versus what you’ve finished. If the ratio is worse than 2:1, you’re overloading. Pick one active initiative to pause. Don’t start the next thing until something finishes. Track cycle time - how long from start to done.
Most teams see it drop immediately. Not because they work harder, but because the system stops fighting them.
Real speed comes from finishing, not starting.
Variety as time cost
Even with WIP discipline and a bias toward finishing, teams can still move slowly if they’re handling too much variety.
When teams are stretched, the conversation is usually about output or resources. Work faster, add headcount. But capacity has a third variable: time. And variety creates hidden time costs that quietly erode it.
Processing ten different formats for the same request type consumes time. Context switching, different monitoring points, aggregation overhead, setup costs for each type.
In knowledge work, this is easy to miss. We see “fifty tasks” not “fifty tasks across fifteen different processing patterns.” But that variety consumes time.
A finance team receives expense reports in ten different formats - PDFs, spreadsheets, emails, photos of receipts. Each format demands different processing. Different validation rules, different error handling, different approval paths.
Processing slows not because the team lacks skill, but because the system is absorbing more variety than it can handle efficiently.
The time costs aren’t obvious. Different inputs mean different monitoring points (where do I check for new requests?), different information digestion (how do I extract what I need?), different aggregation logic (how do I combine these?), different prioritisation rules (which matters more?).
In manufacturing, handling three different raw materials makes these costs visible. In knowledge work, they’re hidden in what looks like “just doing the work.”
Before reaching for more capacity, it’s worth asking: what variety is the team handling, and what time is it consuming?
There are three moves worth making.
First, cut variety at the entry point. Most variety isn’t valuable - it’s just different ways of asking for the same thing. Give people one way to submit a request type, not ten.
Don’t start work until you have everything you need upfront. Handle genuine edge cases separately rather than forcing your main process to flex for everything. Limit how much work is in flight at once.
Second, handle unavoidable variety better. Some variation you can’t eliminate, but you can process it more efficiently.
Automate the repetitive validation steps. Write down what decisions people can make themselves without asking you. Use the same process structure even when the content changes - same steps, different inputs. Don’t let variations pile up at bottleneck resources.
Third, keep information clean. Variety degrades information quality. Requests arrive incomplete, handoffs lose context, priorities stay ambiguous.
Use a one-page handoff format that’s the same every time. Check that key fields match across systems before work starts. Set clear triggers for when something needs escalation - specific risk levels, specific age thresholds, specific buffer levels.
Information only matters if it changes what you do.
Look at where work enters your team. Count how many different formats or paths exist for the same type of request. If it’s more than three, variety might be consuming more time than you realise.
Pick one request type. Define what must be present before work starts. Standardise the entry format. Limit work-in-progress to one less than your current average. Track how long processing takes before and after.
Most teams find they have more capacity than they thought - not because people work harder, but because the system stops fighting them.
Variety isn’t the only answer, but it’s worth checking before you add headcount.
System-level discipline
These three levers work together.
WIP discipline stops you from overloading the system. Finishing bias ensures work actually completes. Variety reduction removes hidden time costs.
But the real power comes from treating them as a system, not isolated fixes.
WIP limits only work if you’re finishing things. Otherwise you hit the limit and stall. Finishing bias only works if you’re not drowning in variety - otherwise every completion takes so long that throughput stays low. Variety reduction only works if you’re not saying yes to everything - otherwise new variety keeps arriving faster than you can clean it up.
Together, they create a reinforcing loop.
Lower WIP means faster cycle times. Faster cycle times mean quicker feedback. Quicker feedback means better learning. Better learning means smarter choices about what to start next.
Finishing bias means completed work generates value. Value creates confidence. Confidence attracts resources. Resources let you say no to low-value work and yes to high-value work.
Variety reduction means cleaner processes. Cleaner processes mean less wasted time. Less wasted time means more throughput. More throughput means shorter queues. Shorter queues mean happier teams and customers.
The opposite is also true. High WIP, starting bias, and unchecked variety create a vicious loop.
Long cycle times delay feedback. Delayed feedback slows learning. Slow learning leads to poor choices about what to start. Poor choices create more WIP.
Unfinished work generates no value. No value creates pressure. Pressure leads to starting more things to show momentum. More starts fragment attention further.
Variety breeds more variety. Unchecked formats multiply. Processes flex to handle edge cases. Flex becomes the norm. Time costs compound.
The system either reinforces discipline or erodes it. There’s no steady state.
Most organisations drift toward erosion. WIP creeps up, finishing slows down, variety multiplies. It happens gradually, then suddenly you’re in permanent firefighting mode.
The fix requires conscious, recurring discipline.
Make WIP visible. Count how many active projects, open requests, or in-flight deliveries you’re carrying. Set a limit based on capacity, not wishful thinking. Enforce it.
Track completions, not starts. Every week, ask: what finished? If the answer is “nothing” more than once, you have a finishing problem.
Audit variety quarterly. Where does work enter the system? How many different formats exist for the same request type? Pick one to standardise each quarter.
This isn’t glamorous work. It doesn’t show up in strategy decks or board presentations. But it’s the difference between organisations that scale smoothly and organisations that add headcount but stay slow.
Flow is a system property, not an individual heroics property. You can’t willpower your way to better flow. You have to design it.
Flow as competitive advantage
The organisations with the best flow don’t just deliver faster. They learn faster.
Fast cycle times mean quick feedback loops. Quick feedback loops mean rapid iteration. Rapid iteration means you discover what works before competitors do.
This compounds. Each cycle teaches you something. The next cycle starts with better information. Over time, the gap widens.
Slow organisations don’t learn slowly because they’re less smart. They learn slowly because their system prevents fast feedback. Work takes months to complete. By the time feedback arrives, context has shifted. Learning doesn’t compound - it dissipates.
Flow also builds resilience. Teams with low WIP can absorb shocks. A priority shifts? They’re not paralysed - they finish what’s in flight and pivot. An urgent request arrives? They have headroom to respond without derailing everything else.
Teams with high WIP are fragile. Every shock cascades. Shifting priorities means abandonedwork. Urgent requests mean everything else slips further. The system is always one surprise away from breakdown.
This matters more than most leaders realise. Markets shift, customer needs evolve, competitors move. The organisations that survive aren’t the ones with the best plan. They’re the ones that can adapt fastest when the plan breaks.
Flow creates that adaptability. Not through heroics, but through discipline.
The irony is that most organisations do the opposite. When pressure builds, they abandon discipline. They say yes to everything, start more projects, tolerate more variety. They think they’re being responsive. Actually, they’re destroying their ability to respond.
Real responsiveness comes from discipline. Low WIP, finishing bias, clean variety. These aren’t constraints that slow you down. They’re the foundations that let you move fast.
The best organisations understand this. They treat flow as a strategic asset, not an operational detail.
They make WIP visible at the executive level. They celebrate completions, not starts. They invest in standardising entry points and reducing variety. They build systems that protect flow, even under pressure.
This isn’t about working less or lowering ambition. It’s about recognising that capacity without flow is waste.
You can have the most talented team, the biggest budget, the best strategy - and still move slowly if your system is fighting you.
Or you can have modest resources and modest ambitions, but move fast because your system enables flow.
Over time, the second group wins. They learn faster, adapt faster, and build resilience that carries them through shocks the first group can’t survive.
Flow is the master lever. Get it right and everything else becomes easier.
The practice
This isn’t theory. It’s practice.
Every week, ask three questions:
How much work is in progress? Count active projects, open requests, in-flight deliveries. If it’s more than your team can process in a month, you have too much WIP. Pick something to pause or kill.
What finished this week? Not what got started, what got done. If the answer is “nothing” more than once a month, you have a finishing problem. Stop starting new work until you clear the backlog.
How many different ways does work arrive? Count formats, entry points, process variations for the same request type. If it’s more than three, variety is costing you time. Pick one to standardise this quarter.
Do this every week. Not as a report, as a discipline. These questions surface problems before they metastasise.
When pressure builds, the instinct is to abandon discipline. Resist it. Pressure is when discipline matters most.
Say no to new work. Finish what’s in flight. Standardise the chaos. Protect the flow.
The organisations that do this well don’t look special from the outside. They’re not the ones with the flashiest launches or the most ambitious roadmaps.
But watch them over time. They ship consistently. They adapt quickly. They handle shocks without breaking. They learn faster than their competitors.
That’s not luck. That’s flow discipline, compounded over years.
Build the discipline. The rest follows.
This essay synthesises ideas from:
See also: Reading Guide for the complete collection of Field Notes.