The System for Scaling a Business

Scaling a business is hard. Period. But I'm one of those people who think anything can be learned: with time, with mistakes and, above all, with the right theory.

Getting good at it is worth a lot. It lets you turn an idea into a business that grows without relying on luck, on a good month, or on the founder (you) giving up sleep.

Want to know how it's done? I'm no expert, but today I want to walk you through the system that has convinced me the most so far.

This whole article is my reading of Alex Hormozi's method, which has older roots in Goldratt's Theory of Constraints. It is not an official framework reproduced word for word.

The most interesting part is that this isn't about trying the latest trendy strategy. It's about looking at the business calmly, stage by stage, number by number, until you find the exact spot where it gets stuck.

The main idea is very simple. Almost stupid:

At any given moment, exactly one thing limits how much your business produces.

Work on that thing and the whole system moves. Work on anything else and you're just spinning your wheels.

In reality, several things usually hold a business back, but we call "the constraint" the one where an improvement would pay off the most. And there is only one of those.

Any improvement not made at the constraint is an illusion.

For me, the biggest appeal of this approach is that it turns something that used to feel like a craft into a system you can measure and repeat.

It's a way to get right what matters now, this week, with these customers and this team, and to ignore the rest without feeling guilty.

Note: if you invest by looking at stocks as pieces of real businesses, this post may be very interesting for you too.

What scaling really means

Scaling doesn't just mean growing revenue.

A business scales when it can serve more customers and make more profit without the founder's involvement (yours), the costs, the complexity, and the complaints growing at the same rate.

Picture a business that doubles its revenue but triples its headcount, wrecks its margins, increases churn, and needs the founder to work twice as hard. It has grown, sure, but it hasn't scaled.

That's why revenue on its own is a bad north star. Before looking for the constraint, define your final outcome:

Final outcome = the money the business actually makes per period

How you measure it depends on your business model. It could be gross profit, cash generated, or profitable retained customers. What you shouldn't do is measure activity: calls made, posts published, hours worked, things that go up while the business stays exactly where it was.

Growing your customer count is only one option, not a requirement. If you raise prices and accept that more people will say no, you can end up with fewer customers and a better business.

Whatever number you choose, it only grows if four things happen:

Break any of the four and the "growth" is worthless.

Acquisition, conversion, retention, and capacity metrics are not the goal. They help you diagnose what is holding back the outcome you chose.

Hormozi boils all growth down to three levers:

1. Get more customers.
2. Make each customer worth more.
3. Keep customers longer.

Everything else should push one of those three numbers or reduce the risk that they drop. If you can't say which lever an activity feeds, it's probably set dressing and you don't need it.

The operating loop

The whole system fits in this loop.

1. Pick a goal

   One project, one 90-day financial goal, with guardrails.

2. Map the business

   Write down the whole system, from market to referral, with numbers.

3. Find the constraint

   The one stage that limits the final outcome right now.

4. Fix the constraint

   The cheapest intervention that, with enough volume, tests the diagnosis.

5. Scale what works

   Pour resources into the proven winner: more, then better.

6. Standardize it

   Turn the repeated win into a documented process.

7. Delegate it

   Hand over the process, the metric, and the authority to decide.

8. Find the next constraint

   Solving one bottleneck reveals the next one.

9. …and back to the start

Easy to say. Hard to do.

At every step, the temptation is to do what's comfortable instead of what's necessary.

The founder who loves marketing tweaks ads while delivery falls apart. The one who loves product ships features while nobody is selling. The loop exists to put your attention where the business needs it, not where you feel like putting it.

One safeguard before you start: run the loop on one project at a time.

A portfolio of businesses is also a business, and it has a constraint too. It's almost always your attention, your capital, or your ability to recruit operators.

For each cycle, pick one main project that gets most of your attention, one proven project in maintenance mode, and at most one small exploratory bet. It's my own take on the 70/20/10 split of "More, Better, New", which we'll cover later.

You can maintain several projects. You can't transform several at once without chopping your attention into little pieces.

First, know what stage you're at

The right constraint to work on depends on the stage of the business.

The Acquisition.com roadmap splits businesses into 10 stages. Each one builds on the previous one, so the order matters.

1. 01DiscoverIs there a painful problem for a specific buyer?

   Evidence required

   Repeated conversations and observable existing behavior.

   The premature trap

   Building a large product before anyone has asked for it.

2. 02MonetizeWill anyone pay for the outcome?

   Evidence required

   Founder-led paid sales. Money, not compliments.

   The premature trap

   Branding, automation, and hiring before the first sales.

3. 03AcquireCan we produce qualified opportunities repeatedly?

   Evidence required

   One acquisition method that works predictably.

   The premature trap

   Launching five channels at once.

4. 04StabilizeCan we serve customers consistently?

   Evidence required

   Quality holds as volume increases.

   The premature trap

   Buying much more traffic while delivery creaks.

5. 05FocusWhich avatar, offer, and channel deserve concentration?

   Evidence required

   Clear segment and funnel data.

   The premature trap

   Adding more products instead of choosing.

6. 06ProductizeCan delivery occur through a repeatable system?

   Evidence required

   Documented process and a predictable result.

   The premature trap

   Hiring without a process for people to run.

7. 07OptimizeCan we improve margin, conversion, retention, and capacity?

   Evidence required

   A stable baseline and reliable measurements.

   The premature trap

   Constant reinvention of things that already work.

8. 08OrganizeCan functions operate with clear owners?

   Evidence required

   Departmental ownership and accountability.

   The premature trap

   The founder still approving everything.

9. 09SpecializeCan narrower roles outperform generalists?

   Evidence required

   Enough volume to justify specialists.

   The premature trap

   Creating unnecessary layers of management.

10. 10CapitalizeWhere should the next unit of cash and attention go?

    Evidence required

    A predictable operating engine.

    The premature trap

    Diversifying before mastering the core.

There's no point optimizing delivery when nobody is placing orders. Hence the golden rule:

For example:

• Don't automate a sales process that has never closed customers manually.
• Don't optimize ads when you can't keep the customers you already have.
• Don't build management layers for work that doesn't exist yet.

You get the idea. It's obvious, sure, but it's one of those things you forget exactly when it matters most.

Pick a 90-day goal

A constraint only exists relative to a goal. "Find the bottleneck" means nothing until you can answer: the bottleneck for what?

Bad goals can't be measured and have neither a baseline nor guardrails. You can work toward them without ever knowing whether you've hit them.

Examples of bad goals:

They are directions, not destinations.

Good goals follow this structure:

Increase [outcome] from [baseline] to [target] by [date], without worsening [guardrails]

For example:

Increase profitable retained customers from 100 to 150 a month within 90 days, without gross margin falling below 65% or first response time going over four hours.

Guardrails matter as much as the target. Any experiment can improve one number while quietly destroying another. Guardrails are there to stop you from fooling yourself.

Don't optimize KPIs for the sake of it. A vanity metric that isn't the bottleneck goes up without moving the business, and worse, it leaves you feeling productive.

Be specific and accept that you're going to fail sometimes. Screwing up is part of the process, especially at the start, when you still don't know which number moves with which lever.

And one more thing: keep the method out of the goal.

This is a classic mistake.

"Hire three salespeople" is not a goal. It's one possible solution to the problem.

The goal might be "increase completed sales from 40 to 70 a month". Maybe hiring is the right move. Maybe the real problem is lead quality, follow-up, or price.

The moment you write the solution into the goal, you've stopped diagnosing. Watch out for your biases. Look at the numbers the way Dr. House looks at his patients: objective, cold, and slightly insufferable.

Map the business

Almost every business is the same machine in a different costume. We can think of a business as a 12-stage funnel:

1. 01Market
2. 02Offer
3. 03Attention
4. 04Lead
5. 05Sale
6. 06Payment
7. 07Onboarding
8. 08Delivery
9. 09Outcome
10. 10Retention
11. 11Expansion
12. 12Referral

The labels change by model, but the machine underneath is the same:

• Services
  Leadappointmentshow-upsaledeliveryrenewal / referral
• SaaS
  Visitorsign-upactivationpaid conversionretentionexpansion
• E-commerce
  Visitorproduct pagecartcheckoutdeliveryrepeat purchase
• Marketplace
  Supplydemandmatchtransactionfulfillmentrepeat
• Media / affiliate
  Audiencequalified visitclickmerchant conversionrepeat visit
• Education
  Leadenrollmentactivationcompletionoutcomenext offer

For each stage, ideally you want six numbers:

The business's constraint is hiding in one of these six numbers in one of the 12 stages.

That doesn't mean you need all six numbers for every stage before acting. Be careful not to get obsessed with collecting data. We've all been there.

If today you're deciding blind, without knowing how much you sell or how much you produce, getting that basic data may be the first step. But if you can already see where the business gets stuck, measure that stage and start there.

The funnel's output is just the stage results multiplied together:

Final customers = initial opportunities × conversion at each stage

Ten thousand visitors, 5% become leads, 20% book a call, 70% show up, 30% buy: 21 customers.

Every stage multiplies, so a small leak anywhere drags down everything downstream. But here's the trap most people fall into: the percentage that looks worst is not automatically the constraint.

A stage is the constraint only when improving it moves the final outcome.

For example, terrible conversion can sit in front of a delivery process that's already maxed out. Improve it and all you'll get is a longer waiting list.

Find the constraint

To diagnose the business, we can start with a very simple test:

The 2× test

If customer demand doubled tomorrow, could you serve it without hurting margins, quality, or delivery times?

Delivery would breakWe could handle it

This question works as a first capacity check. On its own, it doesn't identify the current constraint.

1. If doubling demand would create operational problems, check how much spare capacity is left. If you serve 20 customers a month and can serve 30, you can't double, but you can still grow 50% before hitting that limit. Expand capacity when current demand or the cycle's goal calls for it.

2. If you could comfortably absorb twice as many customers, capacity seems to have room. Look at demand and conversion, but also at retention and per-customer economics. Having spare capacity doesn't automatically prove that getting more customers is the best intervention.

The most common constraints can be grouped into four areas, split into two halves:

Demand side

• Getting customers
• Selling

Capacity side

• Product delivery
• Internal operations

Founder dependency (your time)

Underneath all four there's usually a fifth constraint: your time.

The test helps point your search in the right direction. The diagnosis gets confirmed with the business's numbers.

Walk the diagnostic tree

The previous test is fine for a quick read. But now let's run a full diagnosis.

The order of the questions isn't random. Each one only matters if all the previous ones hold. There's no point optimizing sales conversations for a product nobody wants, or buying leads for a process that can't close them.

01 / 9

Question 1/9. Is there a proven market?

• ·Does a specific group hit this problem repeatedly?
• ·Are people already spending money or time trying to solve it?
• ·Are they paying you — or just saying the idea sounds interesting?

YesNo

Two things about how to use this tool:

1. Answer with evidence, not faith. "Is there a proven market?" means people have paid, not that your mom liked the idea or that you went viral on Twitter.

2. Resist the urge to jump to the question that matches the work you enjoy. Its value is right there: pointing you toward the work that actually matters.

Confirm it's the real constraint

A weak metric isn't necessarily a constraint. Before you fix something that doesn't matter, run these four checks:

1. The counterfactual test. If this metric improved 50% tomorrow and everything else stayed the same, how much would the final outcome improve? If the answer is "barely at all", it's not the constraint.
2. The queue test. Where do work, demand, inventory, or decisions wait? Persistent queues point to bottlenecks that may be hidden.
3. The throughput test. Which stage currently puts the lowest ceiling on your final outcome?
4. The next-break test. If you removed this constraint, what would break next? A real constraint reveals its successor; if removing it changes nothing downstream, it wasn't important.

The counterfactual test is easier to see with numbers.

Go back to the earlier funnel: 10,000 visitors a month, 5% become leads, 20% book, 70% show up, 30% buy. 21 sales a month.

If delivery can only absorb 15 customers, six are left out. And as long as capacity is the ceiling, improving any conversion rate by 20% adds exactly zero customers, because the extra sales hit the same limit. Raise capacity to 25 and those six show up all at once.

Funnel simulator

Conversion improvementConstraint+0%

21 sales a month

Delivery capacityConstraint15

customers a month

Customers served

15

Capacity is the ceiling. 6 sales are left out.

This is the pattern most operators take years to internalize: the constraint is wherever a 20% improvement actually shows up in the final number. Everywhere else, the results of your effort fade away.

The playbooks

You have the diagnosis. Now for the treatment.

Each playbook covers the symptoms, the tests you should run, and the exit conditions. Those conditions are the evidence that tells you the constraint has moved and it's time to re-diagnose instead of continuing to improve.

If you don't want to learn every fix and only care about improving your business, you can use the test above to go straight to the playbook you need.

Or jump directly to the one you want with these links:

The market constraintBack to the diagnosis

Symptoms: people like the idea but don't pay. Selling requires long explanations. There's no urgency. Every conversation ends in a price objection. Retention is bad because the problem was never important. You can't name exactly who the buyer is.

The fix isn't better marketing. You can't advertise your way out of a problem nobody has. That's why the usual advice is to build painkillers, not vitamins. Demand is stronger when the product takes away a pain.

Diagnostic tests, ordered by strength of evidence:

• Behavioral interviews. Ask about what people have already done, not what they would do: when they last had the problem, what it cost them, what they've tried to fix it, what they pay today, who decides the budget, and what happens if they do nothing. Never trust a "yes" to "would you buy this?". Nobody can predict what they'd buy, and almost everyone is too polite to tell you no. If you want to go deeper, read The Mom Test by Rob Fitzpatrick. It's about exactly this.

• Manual presale. Explain the outcome you'll deliver, ask them to pay, and deliver it by hand. One payment proves more than all the praise, surveys, sign-ups, and waiting lists combined.

• Concierge pilot. Deliver the service by hand before building anything that scales. You'll see how the work really flows, which objections come up, and what the customer truly values, which is almost never what you'd assumed.

  The best example I know is Photo AI, by Pieter Levels (@levelsio). When he launched it in October 2022, nothing was automated: customers paid through a Stripe link, uploaded their photos through a form, and he trained the model and generated the photos himself, by hand, one by one. That's how he served his first few hundred customers, until demand swamped him and he had no choice but to automate it. Today Photo AI makes over $100,000 a month.

Exit condition: repeated paid demand, from a recognizable type of buyer, for a recognizable outcome.

Personally, I think these two details matter:

• "from a recognizable type of buyer"
• "for a recognizable outcome"

And they're easy to skip over.

People paying is a very good sign, but I'm not sure it's enough. You need to understand the buyer and what they're trying to get done. Otherwise you'll be building a product blind, and you won't know what to fix next.

It's a mistake I've made myself, on projects like Roast My Web. Even after making thousands of euros from it, I still didn't fully understand my customer or what they needed. Don't fall into the same trap.

So when do you graduate? In theory, a sensible default is five to ten paying customers sold by the founder. You can raise the bar in proportion to the capital the business will need.

The offer constraintBack to the diagnosis

The offer is not the product.

The offer is the customer, the promised outcome, the mechanism, the price, the payment structure, the proof, the risk reversal, the deadline, and the positioning. In other words, the whole package the buyer says yes or no to.

Hormozi lays this out brilliantly in $100M Offers, and its core is the Value Equation:

Value ∝

Dream outcome × Perceived likelihood÷Time delay × Effort and sacrifice

Value goes up when the promised outcome gets bigger and more believable. It often goes up faster when the customer gets results sooner and with less work.

Founders tend to inflate the numerator (bigger promises) and underestimate the denominator (faster and easier). The truth is that the denominator is usually cheaper to move.

Value =Dream outcome7×Perceived likelihood5÷6Time delay×6Effort & sacrifice=0.97

×0.1×1×10

↑ Right now, the single point that moves value most: add proof, guarantees, or demonstrations (+1 likelihood).

The goal is simple:

Make people an offer so good they would feel stupid saying no.

Building offers is a whole topic in itself. To go deeper, check out the Hormozi book I mentioned above.

Still, here's one important warning:

When you build an offer, write it down explicitly: customer, problem, desired outcome, current alternative, unique mechanism, time to first value, required effort, deliverables, proof, risk reversal, price, payment terms, reason to act now.

Any of these fields you can't fill in clearly is something your customer won't understand either.

Then experiment, but do it properly. Test with the same audience and the same channel, changing only one important variable. You can change the outcome, the price, the guarantee, the time to value, or the proof. But only one thing at a time.

And measure both conversion and gross profit per opportunity. A cheaper offer can convert more prospects and still make less money. That's a loss.

The acquisition constraintBack to the diagnosis

The offer sells when the right people see it. If you're struggling, not enough of the right people are seeing it, or they don't see it often enough, or they see it in the wrong place at the wrong time.

This is $100M Leads territory, and its backbone is the "Core Four": every way of generating demand is either one-to-one or one-to-many, aimed at people who know you or people who don't.

One-to-one

One-to-many

People who know you

Warm outreach

Direct messages, calls, and emails to your existing network.

Organic content

Posts, videos, and articles your audience finds and shares.

People who don't

Cold outreach

Targeted messages to strangers who fit the avatar.

Paid advertising

Buying attention from strangers at a predictable cost.

For an unproven business, the sequence to learn and scale cheaply is:

1. Warm outreach
2. Cold outreach
3. Consistent content
4. Referrals and partnerships
5. Paid ads

The order is mostly about not paying for ads before the offer is proven. That would be burning money. First validate that the offer works organically, then scale through the channels that cost money. The other option, if you have plenty of cash and want to validate the offer fast, is to start with ads straight away.

On top of that comes the Rule of 100, which on its own fixes most acquisition problems. Around a hundred reps of your primary activity, every day. A hundred outreach messages, or a hundred minutes of content work, or enough ad variations to actually learn something.

Adapt the number to the channel. In B2B sales, 15 well-researched outreach attempts beat 100 generic ones. What's non-negotiable is the principle: consistent, sufficient volume. With small numbers you're not relying on the system, but on luck.

Volume negates luck.

When acquisition underperforms, diagnose it as its own mini-funnel:

Weak stage	Likely problem
Low reach	Not enough distribution volume
Reach but no attention	Weak hook or poor targeting
Attention but no response	Weak message or weak offer
Responses but bad leads	Targeting or qualification
Good leads but no appointments	Friction or missing follow-up
Appointments but no-shows	Commitment and reminders

Exit condition: one channel that produces qualified opportunities at acceptable economics, consistently enough to plan around. And only one. Every extra channel before that point splits your volume and slows learning across all of them.

The sales constraintBack to the diagnosis

The opportunities are there, but the money doesn't come in. Leads arrive, calls get booked, someone says "let me think about it", and the month ends exactly where it started.

Before blaming the sales team, map the whole path: lead received, first response, qualification, booking, show-up, discovery, offer presented, objections, decision, payment. And measure conversion and time at each step.

A "sales problem" usually turns out to mean one specific leak, and the most common ones are unglamorous: slow first response, zero follow-up, and friction when booking.

A useful test:

• If the founder or the best salesperson can't close qualified prospects → suspect the market or the offer, not the sales team.
• If the best salesperson closes and the others don't → suspect training, process, or hiring.
• If prospects say yes but never pay → suspect payment friction, urgency, or decision-making authority.
• If the close rate is strong but profit is weak → suspect pricing and packaging.

For the conversation itself, Hormozi's CLOSER framework is a good structure:

• Clarify why they're here
• Label the problem
• Overview what they've tried
• Sell the destination, not the features
• Explain away their concerns
• Reinforce the decision after the purchase

What makes this framework valuable is that it's repeatable. A structured conversation can be measured, compared, and trained; an improvised one can only be admired or regretted.

The improvement loop: record calls, find the exact stage where deals die, define observable behaviors for that stage, rewrite, role-play, test with real prospects, compare by salesperson, segment, and source. Turn what wins into a scorecard.

The delivery constraintBack to the diagnosis

You sell more than you can serve, with customers waiting in line, the team putting out fires, and quality slipping a little more every week.

Congratulations! This is the constraint everyone would love to have. It's also the one that can destroy your reputation, because it hurts existing customers first.

Map delivery like a production line: payment, onboarding, setup, core delivery, quality control, support, completion, renewal. And for each stage, track throughput, safe capacity, cycle time, queue size, error rate, and owner.

The first necessary stage that can't absorb additional volume is your operational constraint.

Then try fixes from cheapest to most expensive. By default, in this order:

1. Eliminate: unnecessary approvals, redundant steps, low-value customization, meetings that don't change decisions.
2. Protect the bottleneck: scarce specialists shouldn't spend their time on admin, scheduling, or work someone else can do.
3. Standardize: templates, checklists, defined inputs, quality criteria, decision rules.
4. Batch: group similar work to kill context-switching costs.
5. Automate: only when the manual process works and the exceptions are understood. Automating a broken process just breaks things faster.
6. Delegate: hand over repeatable work with clear outputs and standards.
7. Hire or add infrastructure: only after confirming that capacity really is the constraint. Notice where hiring sits on this list: seventh, not first.
8. Throttle demand: raise prices, limit availability, prioritize valuable customers, create a waiting list.

If the queue is already hurting current customers, don't wait for step 8. Throttle demand from the start while you work on the rest.

Exit condition: the business can take on significantly more customers without an unacceptable drop in quality, turnaround times, customer outcomes, margin, or team workload.

The retention constraintBack to the diagnosis

A company can look acquisition-constrained when its real problem is that customers disappear too fast. This one is hard to spot at a glance. New customers keep coming in, revenue seems fine, and the marketing team keeps filling the bucket. The problem is that nobody has stopped to look for the holes.

Retention is its own system: purchase, onboarding, activation, first value, habit, outcome, renewal, expansion. And when people leave tells you a lot about why they leave:

When churn happens	Likely problem
Right after purchase	Wrong expectations, buyer's remorse, poor handoff
Before activation	Setup friction, confusing onboarding
After first use	Weak product or poor fit
After getting the outcome	Finite problem, no ongoing value
After several months	Declining value, competition, missing use case
At renewal	Poor communication of value, procurement friction

Measure by cohorts, never by blended averages, and slice by acquisition source and segment, because "our churn is 4%" often breaks down into "customers from channel A stay forever and customers from channel B leave within a month".

Improve from the top down, because each step builds on the previous ones. If your cohorts show that the leak is clearly further down the list, start there:

1. sell to better-fit customers
2. set accurate expectations
3. fix onboarding
4. shorten time to first value
5. make progress visible
6. drive the highest-value behavior
7. step in proactively when usage drops
8. improve the core outcome

And only then add continuity offers, and only where ongoing value truly exists.

The economics constraintBack to the diagnosis

If everything works and the bank account keeps shrinking, month after month, with more customers and more sales than ever, you have an economics problem.

Growing is dangerous when the company doesn't understand its own math. Three formulas cover almost everything:

CAC = total acquisition and sales cost ÷ new paying customers

Contribution profit = revenue − direct delivery costs − variable acquisition costs

Payback period =

CAC÷Monthly gross profit per customer

Calculate fully loaded CAC, not just ad spend. And build plans on realized gross profit by cohort, not on an optimistic lifetime value estimate extrapolated from customers who've been around for three months. Theoretical LTV doesn't pay the bills.

Two different diseases share these symptoms, and they need different medicine.

1. If customers end up profitable but eat too much cash up front, the constraint is payback and working capital. Fix it with upfront payment, annual plans, deposits, onboarding fees, faster activation, earlier upsells.

2. If customers are structurally unprofitable, the constraint is the model itself: price, delivery cost, retention, or customer quality. And no collection schedule fixes that math.

This is where Hormozi's most recent book, $100M Money Models, fits in. It's about designing the offer ladder so that profits stack up.

There are four broad types of offers:

1. attraction offers that make it easy to get in
2. upsells that solve the customer's next problem
3. downsells that rescue the almost-yes
4. continuity that creates recurring value and recurring payment

The test for every added layer: does it solve the customer's next problem or improve their outcome? Adding extra steps that cost money and don't really help the customer is a short-term win that hurts you later.

Exit condition: incremental customers generate attractive contribution profit, payback is fundable, and growth stops creating cash problems.

The founder constraintBack to the diagnosis

The last constraint is you.

The business works, as long as you never stop. Every important decision waits for you. Employees queue up instead of acting. Only you can close the big deals. The company's most important knowledge lives in your head, and performance collapses the week you go on vacation.

The escape sequence is productize, then delegate, in that order:

1. The founder does the work
2. Repeats it
3. Identifies the winning process
4. Documents it
5. Trains someone
6. Measures their results
7. Hands over authority

Don't document every experimental process. Document what has already worked repeatedly.

Define roles by outcomes, not activities.

• "Help with marketing" is not a role.
• "Produce 80 qualified opportunities a month at a fully loaded cost below $100 each" is a role.

Because it has an outcome, a metric, and a boundary of authority.

Then move each person up the delegation ladder:

1. observes
2. executes with instructions
3. executes with review
4. recommends
5. decides within limits
6. owns the outcome

Delegation isn't complete until decision rights move with the work. If everything still needs your approval, you haven't delegated.

And set up the management infrastructure that holds all of this together:

• each manager owns a clear outcome
• a few leading metrics
• the resources to move them
• a weekly review
• and corrective actions when they miss

Never hold someone accountable for an outcome while denying them the authority to produce it.

Pick the smallest intervention that teaches you something

Remember, a diagnosis is not a to-do list. Always keep the constraint separate from the fix.

"Onboarding capacity is capped at 100 customers a month" is a constraint. Cutting onboarding steps, running group onboarding, automating setup, improving training, hiring a specialist, charging for custom onboarding. Those are six interventions competing with each other, and "hire someone" is rarely the cheapest test.

To compare them, use this as a proxy:

Priority =

Expected impact × confidence × speed÷Cash required × effort × complexity

Then run it like a scientific experiment.

Write down the constraint, the evidence, the hypothesis, one primary local metric, the business's final metric, the guardrails, and the test volume. And before you see any results, write down the success condition and the stop condition.

Change one important variable at a time. Run at least one full business cycle. Don't redesign the test halfway through because the first numbers look bad.

Small means cheap and contained, not thin. A test with so few reps that it can neither work nor fail teaches you nothing. The same thing applies here as in customer acquisition: volume negates luck. Decide before you start how many reps the result needs to mean something, and don't quit before you get there.

Reading the result is a skill in itself:

• Local metric up, final outcome up. Diagnosis confirmed. Scale it.
• Local metric up, final outcome flat. It wasn't the constraint, the constraint immediately moved to a later step, or you simply improved a vanity metric. Either way: re-diagnose.
• Final outcome up, a guardrail destroyed. You bought activity, not throughput. Sales going up while refunds double is not growth.
• Nothing changes, low volume. Inconclusive. Don't mistake a tiny sample for a failed strategy. Increase the volume.
• Nothing changes, enough volume. The intervention doesn't work. Kill it, but don't kill the diagnosis along with it. Go back to the list of interventions and try the next one. It's still a useful experiment, because it told you which fix doesn't work.

A failed solution doesn't prove the problem doesn't exist. Say serving each customer takes too long. You introduce response templates, use them for a full cycle with enough volume, and the time doesn't drop. What failed is the template. Slowness is still the constraint.

Only question the diagnosis when an intervention moves the local metric and the final outcome stays flat, or when several different interventions fail with enough volume.

Don't dismiss the problem because one solution fails. And don't dismiss the solution before you've really tested it.

Scale the winner: More, Better, New

When something works, most founders immediately look for the next thing to build. Resist. Boredom is part of growing a business.

That's what More, Better, New is for, the 70/20/10 split from the Acquisition.com lead generation system.

70%More

Repeat what already works, at higher volume.

• ·Raise budget on the profitable campaign
• ·More of the outreach that gets replies
• ·More reps of the proven content format

20%Better

Improve or extend the winner.

• ·Stronger hook, better proof
• ·Faster onboarding, sharper targeting
• ·Adjacent customer segment

10%New

Test something genuinely different.

• ·New channel, product, or pricing model
• ·Expect most of these to fail
• ·The goal is finding the next winner

Do more of what works as long as it stays profitable and you can deliver it well. When adding volume stops being a good option, improve the process or look for another route.

The percentages are a guide for splitting time and money, not an exact quota. These are the signs that it's time to switch buckets:

• From More to Better, when you can no longer add volume safely, each extra unit clearly returns less, or quality starts to suffer.
• From Better to New, when improvements stop moving the outcome, the channel saturates, or depending on it has become the risk.

Expect most new experiments to fail. Their job isn't efficiency, it's finding the next winner before the current one hits its ceiling.

And when an experiment wins repeatedly, it graduates:

New → Better → More → standardize → delegate.

Document the inputs, the behaviors, the outputs, the quality standards, and the common exceptions. Assign an owner. Remove founder approval. A process isn't complete because a document exists. It's complete when another competent person produces the outcome repeatedly without you directing every action.

The operating cadence

Systems decay little by little, with a review skipped here and a metric nobody checks there, until nobody knows what the constraint is.

To prevent that, schedule four kinds of review:

• Daily. The primary actions tied to the current constraint. Outreach attempts, sales conversations, customer interviews, units delivered, churn interventions. Don't confuse activity with outcome, but do measure the activity that produces the outcome.

• Weekly, 30 to 60 minutes. The constraint review. What's the 90-day goal? Current outcome versus target? What's the constraint, and what evidence says so? Did the constraint metric move? Did the final metric move? Did any guardrail get worse? What do we continue, stop, or change, and who owns the next action, by when? Status reports can be written. Meetings exist to produce decisions.

• Monthly. Funnel conversion, cohort retention, CAC, gross margin, contribution profit, payback, capacity, founder dependencies, cash. And one question above all. Is the constraint still the same?

• Quarterly, per project. Recalculate the economics, repeat the 2× test, identify the stage and the constraint, and decide whether to scale, improve, maintain, pause, or kill. Then allocate founder time and capital accordingly.

Managing several projects

The same logic scales to a portfolio of projects. Classify each project honestly:

• Explore: plausible market, no meaningful payment evidence yet. Goal: learn whether the problem and the customer are real.
• Prove: there are customers who have paid, delivery is still manual or inconsistent. Goal: prove repeatable demand, delivery, and value.
• Scale: demand, delivery, retention, and economics hold up. Goal: volume, without breaking economics or quality.
• Maintain: it works, but it's not the highest-return use of your next hour. Goal: preserve performance with minimal involvement.
• Pause or kill: the evidence no longer justifies the investment. Sunk cost is not a reason; the work already invested is lost either way, and keeping a dead project alive charges you twice.

Allocate attention roughly like "More, Better, New": ~70% to proven engines, ~20% to promising projects with evidence, ~10% to genuinely new bets.

When scoring projects against each other, look at pain, willingness to pay, retention, economics, distribution, founder advantage, learning speed, capital efficiency, operational load, and upside; but don't just add up the scores. A single fatal weakness disqualifies a project: no willingness to pay, impossible economics, no legal path, no distribution, or no ability to deliver.

A 9/10 opportunity you can't reach is worth 0.

One warning: don't diversify too early.

Diversifying has a real cost: it splits attention, lowers volume per channel, and slows learning everywhere.

Reduce a concentration risk only when three things are true. The main engine already works. The concentration threatens survival. And you have the management capacity to run the alternative without killing the core.

What to fix, symptom by symptom

The whole system compresses into a chain of "if X, then Y" rules. When you don't know what to do, follow these questions:

• No clear, painful customer problem?

  →Research and solve problems manually.
• People care but don't pay?

  →Fix the market, the outcome, the offer, the proof, or the price.
• You can sell manually but lack opportunities?

  →Fix acquisition.
• Leads but few appointments?

  →Fix response speed, qualification, follow-up, booking.
• Appointments but few sales?

  →Fix the offer, the sales process, the targeting, or trust.
• Sales but delivery breaks?

  →Fix capacity and operations.
• Delivery works but customers don't succeed?

  →Fix product quality and time to value.
• Customers succeed but leave?

  →Fix ongoing value, retention, or customer fit.
• Customers stay but growth eats cash?

  →Fix CAC, payback, collection timing, or the money model.
• Everything works but you're overwhelmed?

  →Standardize, delegate, hand over authority.
• The proven mechanism has unused capacity?

  →Do More.
• More produces diminishing returns?

  →Do Better.
• More and Better are exhausted?

  →Try New.
• An intervention wins repeatedly?

  →Document it and delegate it.
• The constraint disappears?

  →Diagnose the business again.

The biggest mistakes

Apparently, almost every mistake is one of these, and most of them are the same mistake in different costumes (working on a non-constraint):

How do you put it into practice?

None of this helps just by reading it. You don't need to rebuild the business. You need a few hours, a spreadsheet with whatever numbers you have, and a couple of decisions you've been putting off.

1. Write down your final outcome. One financial number per period that fits your model, like gross profit, cash, or profitable retained customers. If you can't calculate it, ask yourself whether you're deciding blind. If you don't know how much you sell, how much delivery costs, or how much you keep, measuring that may come first. If you know it roughly, an estimate is enough to start.
2. Run the 2× test. If twice the demand showed up tomorrow, could you serve it without breaking margin, quality, or timelines? Then compare current demand and your goal against sustainable capacity. Not being able to double doesn't mean you're already at the limit.
3. Name one constraint, with its number. Not a list of problems. One specific stage and the figure that proves it's that one. If you don't have the figure, get the minimum that lets you choose, even if it's a rough estimate done by hand.
4. Pick the cheapest intervention that moves it, and decide when you'll kill it. What you change, which local metric should move, how much volume you need for the result to mean something, and what you'd need to see to declare it dead. If it dies with enough volume, the intervention dies, not the diagnosis.

You'll get it wrong the first time. It doesn't matter. A badly executed cycle teaches more than a well-written plan, and the constraint will move as soon as you start really looking at it. When it moves, go back to step 2.

The real lesson fits in one sentence:

“Work on the one problem that is limiting the whole business right now. When it stops being the constraint, stop optimizing it and find the next one.”

Everything else is commentary.