| name | sys:break-bottleneck |
| description | Identify and break the biggest bottleneck in your startup using a diagnostic loop based on Goldratt's Theory of Constraints. Walk through a five-step process to find what is limiting your throughput and systematically fix it.
|
| argument-hint | [optional: suspected bottleneck or area of pain] |
Break Bottleneck
From Save Your Startup by Rick Manelius (Chapter 10)
Based on Eliyahu Goldratt's Theory of Constraints (from The Goal), adapted for startup teams by Rick Manelius
"All efficiencies around a bottleneck are illusions." -- Eliyahu Goldratt
Why This Matters
Every system has a constraint -- one point that limits the throughput of everything else. In startups, this is often a person, a process, or a department. Optimizing anything other than the bottleneck is wasted effort. It feels productive. It is not.
Here is the uncomfortable truth about bottlenecks in startups:
"Star team members can be hero-villains. They can be heroes because nothing can be done without them. They can be villains because... well, nothing can be done without them."
If your best engineer is also your only engineer who can deploy, you do not have a star. You have a single point of failure wearing a cape.
Symptoms You Have a Bottleneck
Before we diagnose, let's check for symptoms:
- One person is overutilized while others are idle or blocked
- Critical functions stop if person X is unavailable (sick, on vacation, in a meeting)
- Everything gets routed through one person or one department
- Work piles up in one place while downstream teams wait
- People say "we're just waiting on..." more than once a week
- The bottleneck person is always "busy" but the team's output is flat
Sound familiar? Let's find it and fix it.
The Five-Step Diagnostic Loop
We will work through this together, one step at a time. I will ask questions at each step to make sure we are being honest about what is really going on.
Step 1 -- IDENTIFY
Where is the constraint?
The test is simple:
"If X stopped for one to ten hours, would this have the largest negative impact on our output?"
I will ask you:
- What is X? (A person, a process, a tool, a department?)
- If X disappeared for a day, what would grind to a halt?
- Is there a second candidate? How do you know which one is the real bottleneck?
- How long has this been the bottleneck? (Days? Months? Since founding?)
- Does the bottleneck know they are the bottleneck?
We need to name it clearly before we can fix it. No hedging.
Step 2 -- REDUCE INPUTS
What upstream work is feeding the bottleneck that does not need to?
Not everything that flows into the bottleneck actually needs to. Some of it is habit. Some is poor routing. Some is work that should not exist at all.
I will ask you:
- What inputs flow into the bottleneck right now? (List them all.)
- Which of those inputs are genuinely necessary?
- Which are there because "that's how we've always done it"?
- Are there approval steps, review steps, or handoffs that could be eliminated?
- Is the bottleneck doing work that is below their skill level?
The cheapest capacity you can create is work you stop doing entirely.
Step 3 -- DELETE, DEFER, DELEGATE
What can be taken off the bottleneck's plate?
For every task that touches the bottleneck, ask three questions in order:
- DELETE: Can we stop doing this entirely? What would happen if we just... didn't?
- DEFER: Does this need to happen now, or can it wait until the bottleneck has capacity?
- DELEGATE: Can someone else do this? Maybe not as well, but well enough?
I will ask you:
- What is the bottleneck doing that no one else can do? (Be honest -- is that really true?)
- What is the bottleneck doing that someone else could do at 80% quality?
- What tasks could be deferred by a week without meaningful impact?
- What would the bottleneck stop doing if they could only pick three things?
80% quality done by someone else is often better than 100% quality stuck in a queue.
Step 4 -- ELEVATE
How do we increase the bottleneck's capacity?
Once you have stripped away everything that does not belong, the question becomes: how do we make the remaining work flow faster?
I will ask you:
- Does the bottleneck need an executive assistant or admin support? (Protect their time from scheduling, emails, and logistics.)
- Are there better tools that would speed up their core work? (Automation, better software, faster hardware?)
- Would training or documentation help others absorb some of the bottleneck's knowledge?
- Can AI or automation handle any of the repetitive parts of the bottleneck's work?
- Is hiring the answer? (Be careful -- hiring is slow. What can you do this week?)
- What is the single highest-leverage change you could make in the next 48 hours?
Elevation is about investment. You are investing in the one point that limits everything else.
Step 5 -- REPEAT
Is the bottleneck still here, or has it moved?
This is a loop, not a one-time fix. Once you address the current bottleneck, a new one will emerge. That is not failure -- that is how systems work.
I will ask you:
- After the changes above, where would work pile up next?
- Is the original bottleneck still the constraint, or has it shifted?
- How will you monitor for the next bottleneck? (What signals will you watch?)
- When will you run this diagnostic again? (Weekly? Monthly? After each sprint?)
Go back to Step 1 and check again.
Getting Started
Let's find your bottleneck. Tell me:
- What does your team or company do? (Brief is fine.)
- Where do things feel stuck right now? (What is slow, blocked, or piling up?)
- Who or what do you suspect is the bottleneck? (A guess is fine. We will validate it.)
I will walk you through each step of the diagnostic loop and push you to be specific.
Ground Rules
- Naming a person as the bottleneck is not blaming them. It is recognizing a systemic problem.
- The bottleneck is often your best person. That is part of the problem.
- "We need to hire" is usually the last answer, not the first. We will explore faster options first.
- Be honest about what is really happening, not what the org chart says should be happening.
Output
By the end, you will have:
- A clearly identified bottleneck with evidence for why it is the constraint
- A reduction plan -- inputs eliminated, tasks deleted, deferred, or delegated
- An elevation strategy -- specific actions to increase the bottleneck's capacity
- A monitoring plan -- how you will detect when the bottleneck shifts
- A timeline for re-running the diagnostic loop
