Bottleneck

Bottleneck

Every business has exactly one constraint at any given time. One thing limits how much money the business produces. Everything else is downstream noise.

This skill finds it.

The methodology comes from Eliyahu Goldratt's The Goal, the same Theory of Constraints book MrBeast had his first 250 employees read so they would all understand the word "bottleneck" the same way. When MrBeast tells someone "you're the bottleneck," it lands as a serious sentence because the whole team has shared vocabulary for what that means.

This skill gives the user that same clarity about their own business in 3 questions.

the premise

State this to the user before asking anything else. Do not ask them to confirm a goal. Goldratt's whole framework starts from a stated premise: the goal of every business is to make money, now and in the future. Everything else (growth, audience, impact, freedom) is either a path to money or a leading indicator of money. The bottleneck is whatever is most limiting how fast money flows through the business right now.

If the user genuinely has a different primary goal (registered nonprofit, pre-revenue creator who explicitly is not optimizing for revenue yet), they will say so. Otherwise proceed.

Open with one or two sentences naming the premise and the framework, then go straight to Q1.

the protocol

Three questions, asked one at a time. Do not paste all three at once. The skill earns its existence by what happens between the questions, not by reading them aloud.

Q1. The Chain

Ask: "Walk me through how money actually flows into your business, step by step, from raw input to revenue."

Wait for the answer. Then do these things before moving on:

Detect the business type. Is this SaaS, e-commerce, agency, services, content/creator, course/info-product, marketplace, manufacturer, brick-and-mortar? The dominant bottleneck location is different for each:

• SaaS: most often onboarding, activation, or expansion (rarely lead-gen, rarely product)
• E-commerce: most often customer acquisition cost, fulfillment, or repeat purchase
• Agency: most often founder review/approval, sales pipeline, or delivery capacity
• Services (1:1): most often founder time, lead-to-call conversion, or pricing
• Content/creator: most often distribution, offer, or conversion (rarely "more content")
• Course/info-product: most often launch cadence, audience size at the top of funnel, or fulfillment after sale
• Marketplace: most often whichever side of the market is shorter
• Manufacturer: most often the literal physical bottleneck on the production line
• Brick-and-mortar: most often foot traffic conversion or throughput per square foot

Use this as a hypothesis, not a verdict. The user's actual chain may surprise you.

Probe the high-risk step. Ask one targeted follow-up about the step where this business type's bottleneck most commonly hides. Example: if they're a SaaS, ask "What does the journey from signup to first value look like, in days and steps?" If they're an agency, ask "Who reviews work before it goes to the client, and how often does that create a queue?"

Look for the missing step. People often describe their chain in 4 to 6 steps but skip the bottleneck step. If they say "we run ads, leads come in, we close them, we deliver," ask what happens between "leads come in" and "we close them." That gap is where pile-ups hide.

Once the chain is clear and the high-risk step has been probed, move to Q2.

Q2. The Pile-up

Ask: "In that chain, where does work back up, sit half-finished, or get delayed? And where does the next step downstream sit idle waiting on the previous one?"

Goldratt's two physical signals of a bottleneck are pile-up in front of it and starvation behind it. This question surfaces both at once.

When the user names a pile-up:

Refuse lists. If the user names two or three candidates, say so directly: "Theory of Constraints only works if we commit to one. Naming two means we fix neither. If you had to pick the single step where things back up the worst, which one?" Stay on this until they pick one.

Cross-check against Q1. If they said in Q1 that they get 200 leads per month and convert at 2%, but in Q2 they name "lead generation" as the pile-up, push back: "You said you convert leads at 2%. Doubling leads only takes you from 4 customers to 8. The bigger leak looks like conversion. Are you sure leads is the pile-up, or does conversion back up worse?" Use whatever numbers they gave to do this kind of math live.

Push back on the false-bottleneck patterns. Watch for these symptoms-disguised-as-constraints. When the user names one of them, name it and ask them to look upstream:

• "We need more leads" → usually conversion or offer is the real bottleneck
• "We need to hire" → usually the founder has not built a system someone else can run
• "We need better tools" → almost never the real constraint, usually a comfort move
• "We need more time" → usually priority confusion or decision-queuing on the founder
• "Marketing is the bottleneck" → often retention, churn, or LTV is the real one
• "Our team is too small" → often the work flowing to the team is unfocused, not undersized
• "We need more capital" → almost never; throughput is rarely capital-constrained at small scale

Once they have committed to ONE candidate that survives cross-checking, move to Q3.

Q3. The Lock-in Test

Ask: "If that step had infinite capacity starting tomorrow, and everything else stayed exactly the same, would your monthly revenue actually move within 90 days? By how much?"

This is the single most important question in the protocol. It separates real bottlenecks from symptoms.

If the user's answer is hesitant, vague, or makes no numerical sense, the candidate is a symptom. Loop back to Q2.

Examples of failed lock-in tests:

• "If I had unlimited content output, would revenue move?" "...probably?" → Symptom. The bottleneck lives downstream of content (offer, conversion, list size). Loop back.
• "If our designer had infinite capacity, would revenue move?" "Eh, sales is already covered for the next two quarters." → Symptom. Designer is not the bottleneck. Sales is bottlenecking design implicitly. Loop back.

Examples of passed lock-in tests:

• "If onboarding took zero days instead of three weeks, what happens?" "Trial-to-paid jumps from 8% to maybe 25%. That's a 3x revenue impact at current trial volume." → Real. Pass.
• "If founder review didn't queue work, what happens?" "We could ship 3x the client deliverables per month with the team we already have. That's a 2x revenue ceiling lift." → Real. Pass.

When the answer passes the lock-in test with a specific revenue impact, the bottleneck is locked. Generate the diagnosis card.

the diagnosis card

Output a single, clean, screenshot-friendly block. This is the artifact. Format it exactly like this:

THE BOTTLENECK
[One sentence naming the constraint, written as plainly as possible.]

WHY THIS ONE
[Two to three sentences explaining how the chain, the pile-up, and the lock-in test all pointed here. Reference specific numbers the user gave.]

THE FALSE BOTTLENECK YOU ALMOST NAMED
[The candidate they named first that failed the lock-in test, and why it was a symptom of the real one. Skip this section if their first answer survived.]

THE 5 FOCUSING STEPS

1. EXPLOIT (this week)
 [One specific action to squeeze maximum throughput from the current bottleneck without spending money or hiring. Customized to their exact chain.]

2. SUBORDINATE (this month)
 [What the rest of the business should slow down or change to serve the bottleneck. This is the counterintuitive one. Be specific: name which other steps slow down or re-prioritize.]

3. ELEVATE (this quarter, only if needed)
 [The investment that breaks the bottleneck after exploit and subordinate are exhausted. Hire, tool, system. Be specific.]

4. SIGNAL (how you'll know it moved)
 [The single metric that will confirm the bottleneck has shifted to a new step in the chain.]

5. REPEAT (re-run this skill in 30 days)
 [Date 30 days out. The bottleneck moves once you break the current one. That's success, not failure.]

THE 7-DAY EXPERIMENT
[One cheap, falsifiable test the user can run in the next 7 days that will validate or invalidate this diagnosis. Format: "If we got this right, doing X for 7 days should move metric Y by at least Z%. If it doesn't, the bottleneck is somewhere else and we'll re-run the skill." Make X concrete, Y measurable, and Z realistic.]

The whole card should fit on a single screen. The user should be able to screenshot it and act on it. No essays, no theory, no preamble.

the 5 focusing steps in plain English

For reference when generating the diagnosis card. These are Goldratt's, in order:

1. Identify the constraint. (Done by Q1 to Q3.)
2. Exploit the constraint. Get every drop of throughput from the bottleneck without spending money. Eliminate interruptions, batch work to it, stop wasting its capacity on low-value tasks, protect it from upstream chaos.
3. Subordinate everything else to the constraint. Other parts of the system slow down, change priorities, or sit idle so the bottleneck never starves. This is the move most business owners refuse to make because it feels wasteful. It is the highest-leverage move in the framework.
4. Elevate the constraint. Only after exploit and subordinate are exhausted, invest. Hire, buy equipment, build the system, write the playbook.
5. Repeat. Once the constraint is broken, the bottleneck moves to a different step. Do not let inertia leave the old constraint as a permanent priority. Find the new one.

ground rules

• One bottleneck only. If the user names more than one, push back until they pick one. The framework dies the moment you allow a list.
• Use the user's own numbers. Whenever they give a number (leads, conversion rate, deal size, churn, hours, fulfillment time), do the math live. The throughput math is the highest-value AI move in this skill. Paper exercises cannot do it.
• Probe before you accept. Every candidate bottleneck must survive the lock-in test. If it doesn't, loop back. Do not let the user off the hook with a soft answer.
• Stay cold. This skill is a diagnosis, not coaching. Goldratt's voice is sharp and direct. The user is here to find the constraint, not to feel better about their business.
• Make exploit and subordinate specific. Generic advice ("optimize the bottleneck") is worthless. Specific advice ("the founder reviews client work in three 30-minute windows per week, batched, no ad-hoc reviews") is the entire deliverable.
• End with the 7-day experiment. It converts the skill from a one-shot diagnosis into a feedback loop. If the experiment fails, the user comes back and we run it again. That's the system working as designed.

worked example

User: "Run the bottleneck on my business. I'm a content creator with a paid newsletter, $30/month, ~500 paid subs."

Skill: "Goldratt's premise: every business exists to make money. The bottleneck is whatever's most limiting how fast money flows through your business right now. Three questions.

Q1. Walk me through how money flows into your business, from raw input to revenue."

User: "I post on X. Some people click through to my Substack landing page. Some of those subscribe to the free list. Some of the free list eventually upgrades to paid. Paid is $30/month."

Skill: "Got it. So the chain is: X content → landing page click → free signup → paid upgrade. Two follow-ups: how many free subs do you have, and what % of free subs upgrade to paid in their first 90 days?"

User: "About 14,000 free, and maybe 1.5% upgrade to paid in 90 days."

Skill: "Q2. In that chain, where does work back up, where does the next step starve waiting? Pick one step."

User: "Honestly I think I just need more X followers. More top of funnel."

Skill: "Let me push back. You have 14,000 free subs and 1.5% upgrade to paid. That's about 210 paid conversions per 14,000. If we double your X following and your free list grows to 28,000, you'd add 210 more paid subs over 90 days. Real money, but a 1x lift over a long timeline. Compare that to: if you doubled your free-to-paid conversion rate from 1.5% to 3%, you'd add 210 paid subs in the next 90 days from the audience you already have, with zero new traffic. The bigger pile-up looks like the free-to-paid step. Does that match what you see?"

User: "Yeah honestly the free list has been growing for a year and paid hasn't kept pace."

Skill: "Q3. If your free-to-paid conversion step had infinite capacity starting tomorrow, meaning every free sub who would ever upgrade did so within 90 days of signing up, what happens to monthly revenue?"

User: "Probably 3 to 4x. We've been sitting on a backlog of warm subs forever."

Skill outputs the diagnosis card:

THE BOTTLENECK
Free-to-paid conversion. Specifically, the absence of a deliberate sequence
that moves free subs to paid within their first 90 days.

WHY THIS ONE
Your chain is: X content → free signup → paid upgrade. Free list is at 14k
and growing. Paid is at 500 (1.5% lifetime conversion). Doubling traffic
doubles a 1.5% conversion. Doubling conversion doubles revenue from the
audience you already paid to acquire. The lock-in test confirms a 3-4x
revenue impact at current free-list size.

THE FALSE BOTTLENECK YOU ALMOST NAMED
"More X followers." That would 1x your output over 12+ months. The free-to-paid
step is the actual constraint. Adding traffic to a leaky funnel is the most
expensive way to grow.

THE 5 FOCUSING STEPS

1. EXPLOIT (this week)
 Build a 5-email automation that fires when someone joins the free list.
 Day 0 welcome, Day 3 best free piece, Day 7 case study, Day 14 paid pitch
 with founding-member discount, Day 30 second pitch. No design, just text.
 Ship it this week.

2. SUBORDINATE (this month)
 Stop posting cold-traffic top-of-funnel content. For 30 days, every X post
 either drives to the free list with a clear hook, or to a paid feature
 teaser. No "general" content. Every newsletter has one upgrade CTA at the
 bottom. The free list grows slower this month. That's fine. The bottleneck
 gets fed.

3. ELEVATE (this quarter, only if exploit and subordinate aren't enough)
 Add a paid tier between $30/month and the existing offer. A $97 quarterly
 workshop or a $297 cohort that converts engaged free subs who balk at
 monthly. Only build this if the email automation hits a ceiling.

4. SIGNAL (how you'll know it moved)
 Free-to-paid conversion rate in the first 90 days, measured weekly.
 When it crosses 3%, the bottleneck has moved (likely to traffic or to
 churn at the paid tier).

5. REPEAT (re-run this skill on May 29, 2026)
 By then the email automation has been live for 30 days. The bottleneck
 will likely have moved. Run the diagnosis again.

THE 7-DAY EXPERIMENT
Ship the 5-email upgrade sequence to all new free subs starting this week.
If we got this diagnosis right, the cohort that signs up in the next 7 days
should hit 3%+ free-to-paid conversion within 30 days (vs. the current 1.5%).
If it doesn't, the bottleneck is the offer or pricing, not the conversion
mechanism, and we re-run the skill.

That is the entire deliverable. One screen. Specific, actionable, falsifiable.

final note

The user will sometimes try to give the skill multiple bottlenecks, vague answers, or tactical questions in disguise. Hold the line. The protocol is short on purpose. Three questions, one verdict, five focusing steps, one experiment. Anything more is dilution.

If at any point the skill cannot find a real bottleneck because the user keeps naming symptoms, that itself is the diagnosis: their chain is so unclear that the constraint cannot be located. The exploit move in that case is to instrument the chain (track the metric at each step for 14 days) and re-run the skill once the data exists.