Chapter 5: Limiting Resources: Money, Time, and Judgment

One or two pieces of paper #

To judge a company's true strategy, there are two pieces of paper to look at. One is the strategic statement (vision, mission, annual theme words), written for others to see; the other is the budget flow (where the money actually flows), written for accountants to see. Two pieces of paper often tell two stories, and only one piece of paper doesn't lie. Don’t listen to what the company says, watch where the money goes. The function of a declaration is to manage other people's judgments of you, and it is only the flow of water that exposes your own judgments. I named these two actions Interference Method (narrowing the other party's options) and Clarity Method (expanding one's own options) respectively. Chapters 8 and 7 will each use a whole chapter to expand, but let's remember the phenomenon itself first.

One founder simply included this rule in his letter to shareholders. In 1997, in Amazon's first shareholder letter after it went public, Jeff Bezos wrote in black and white a sentence that was later quoted for thirty years: "When forced to choose between beautifying GAAP statements and maximizing the present value of future cash flows, we choose cash flow." Pay attention to the structure of this sentence: it is not a performance report, but a publicly committed order of trade-offs, telling everyone in advance what order we will choose when two values ​​conflict. This letter will appear again later, so let’s take it first to set up the question of this chapter.

The question is: money, time, manpower, judgment, there are many kinds of resources in the hands of a company, why should cash flow be singled out and included in the commitment? The answer is that resources are divided into two categories, and most management disasters result from confusing the two categories.

2. limiting resource: shortcomings determine output #

There is a concept in chemistry called limiting reactants: the extent to which a reaction can proceed is not determined by the raw material added the most, but by the one that is used up first. The organization is exactly the same. Resources are divided into two categories: excess resources talk about scale, and scarce resources talk about ROI. Excess resources can be generous, can be wasted, and can be managed according to "the more the merrier"; for the restrictive kind, every destination must answer the same question: What does it get in exchange for?

Three levels, three limiting resources:

• At the company level, limiting resource is money. Manpower can be recruited, execution can be outsourced, but the day the cash flow is cut off, everything stops at the same time. So when looking at a company's strategy, we have to look at the flow of money. Money is its scarcest voting right, and the outcome of the vote is the strategy itself.
• At the individual level, the limiting resource is time. Money can be made again, attention can be restored, and time is the only input that is strictly non-renewable. So look at a person's true priorities, don't look at what he says is important, look at his calendar. Wherever time goes, that is who he is.
• AI-native organization level, limiting resource is a judgment. This is a new addition to this chapter. The derivation has been completed in the first four chapters: execution tends to be free (Chapter 1), and the upper limit of the organization's output after execution is free is determined by the quality of judgment (Chapter 4). When other raw materials are close to unlimited supply, judgment becomes the reactant that is exhausted first. It is scarce, and as Chapter 4 stated, it cannot be purchased directly, but can only be solidified and copied.

The three layers are not a parallel relationship, but a funnel: the company's money hires people's time, people's time produces judgment, and judgment determines the next flow of money. The same question should be asked at every level of the funnel: What is the current ROI of this limiting resource? Next, proceed layer by layer in funnel order, with one sample for each layer.

3. The level of money: companies that write down the order of choice in shareholder letters #

Back to that 1997 letter. It is worth reading as an organizational sample rather than a collection of quotes, because it shows the complete form of "limiting resources configured according to standards".

The title of the letter is the standard title: "It's All About the Long Term." In the year when I wrote this letter, Amazon’s revenue was only $147.8 million, and its employees increased from 158 to 614. It was a small company that could die at any time. The allocation rules set out in the letter are as precise as legal provisions: investment decisions are based on long-term market leadership, not short-term profits and Wall Street's short-term reaction; "We will make bold rather than timid investment decisions - some will succeed, some will not, and we have learned valuable lessons in both cases"; and the cash flow terms quoted earlier.More important than the terms are two mechanical details. First, this letter will be reprinted every year and attached to the back of each new shareholder letter. The 1998 letter explained the reason: the number of shareholders increased from 13,000 to 200,000, and new shareholders needed to know "what kind of company we are." The original words were "We do not claim that this is the correct philosophy, we only claim that this is our philosophy." Please note the position of this action in the framework of this book: the judgment standards are documented, published publicly (to be tested for consistency), and reiterated year by year (to prevent drift). Chapter 4 says that judgments must be written into standards before they can be stored in the organization. This letter is the physical object of the resource allocation standards and will be stored for twenty-four years. Second, the fulfillment of standards can withstand verification: Prime, Marketplace, and AWS are all products of "bold investment" terms. The 2020 shareholder letter speaks for itself: in 1997 these things "didn't even exist". By 2020, AWS’s annualized revenue is $50 billion, more than three hundred times the revenue of the entire company in the year of writing.

This sample answers the core question at the money level: the ROI of money is not to calculate the immediate return one by one (that way there will never be AWS), but to set a sorting standard for money across the account period, and then implement it with consistency for decades**. A single investment is allowed to fail (the terms state "some will not succeed"), and the ranking criteria are not allowed to drift.

4. The first layer of money, the opposite: treating borrowed money as excess resources #

The reverse sample of the same layer is equally astonishing in scale.

At the end of 2013, Evergrande entered the bottled water industry, and the first-year sales target of Evergrande Ice Spring was set at 10 billion yuan. Actual results: 1.09 billion yuan, loss 2.37 billion yuan. This was followed by grain and oil, dairy, football, life insurance, cultural tourism (the cumulative investment in Haihua Island was approximately 160 billion yuan), and automobiles (over 47.4 billion yuan was invested). Each business line has a grand narrative when establishing a project, but the common point is: "What criteria will be used to accept the ROI of this money, when will it be accepted, and what will happen if it does not meet the standards" can not be found in the project basis. Borrowed money (the most typical limiting resource, the one with interest) is treated as excess resources and allocated according to the logic of "the more, the better". Making one more product is not adding a revenue line, but multiplying the complexity: Each new business line not only consumes cash, but also consumes management bandwidth, organizational credit, and error correction windows.

The most anatomical value of this specimen is that it has corrected errors. In September 2016, Evergrande packaged and sold its three major fast-moving consumer goods businesses, grain and oil, dairy, and mineral water, for 2.7 billion yuan. The stop-loss action was clean and tidy, and the first round of contagion ended here. But please compare Klarna's correction in Chapter 2: Klarna changed the standard (replacing the "cost only" evaluation standard), Evergrande cut the business, and the error-producing standard of high leverage and diversification remained intact. So in the same year, after the name was changed to "China Evergrande", the same standard began to produce a second round of greater spread: cultural tourism, health, automobiles, plus a 7 billion yuan floating loss in the Baowan dispute. In the first round, the loss was several billion, and in the second round, the loss was hundreds of billions. Correcting deviations by only cutting business without changing standards is tantamount to applying beauty treatments to the lesions, and beauty treatments will make the next attack more confident: Look, didn't we stop the loss in time last time? The outcome is written in public documents: as of the end of 2022, total liabilities were 2.44 trillion yuan and losses were 812 billion yuan in two years. In January 2024, the Hong Kong High Court issued a liquidation order.

If Amazon and Evergrande are put on the same level, the difference is not in their courage (both companies have made huge bets), but in whether the bets are governed by ranking standards: one has written the standards into shareholder letters and reprinted them for 24 years, and the other has not even included the word "acceptance" in the project approval documents.

5. A layer of time: the true meaning of delayed gratification #

Funnel down to the individual level. The special thing about time, a limited resource, is that its mismatch is almost imperceptible: if you spend the wrong money, you will get a bill, if you spend the wrong time, you won’t even get a receipt.

The sample at this level uses Zhang Yiming, and the material is his first-hand statement in an exclusive interview with Caijing in 2016. He named his most admired trait "delayed gratification", and his explanation is much more precise than the popular version of the word: "The conservative nature is because I believe in delayed gratification. If you think something is good, you might as well delay it later, which will allow you to raise your standards and leave a buffer." The same logic applies to money: "Many companies spend money and then refinance it, but I always set aside enough money."

Translate this passage into the framework of this chapter: Delayed gratification is not a performance of endurance, but an inter-temporal allocation strategy of limiting resources, which is to transfer resources from "immediate satisfaction" to "the side with higher compound interest". There are two benefits of mobile, and he named them both: improved standards (waiting one more step will screen out barely qualified options) and buffer (the redundancy left is the error correction window). He even did double-sided accounting for this strategy, which is the most valuable part of this material: He openly admitted the cost, "If Toutiao can spend a little more money in 2012 or 13, maybe the growth will be faster." Delayed gratification is not a costless virtue, but a calculated trade-off; long-termism, which only emphasizes benefits but ignores costs, is another Interference Method.He also left a fine distinction that can be directly put into the management toolbox: "Because the founder personally needs to succeed, failure to make decisions at the right time is a problem of delayed gratification; because of errors in judgment, it is a skill problem." It is also a matter of missing the opportunity, a kind of greed that is wrong in the time dimension, and a mistake that is based on judgment ability. The attributions are different and the practices are completely different.

6. The first level of judgment: human flywheel and all-employee ROI #

The deepest level of the funnel: Judgment. Its allocation unit is not yuan or hours, but people. Judgment is parasitic on people, and the way organizations pay for judgment is to pay for people. Therefore, "judgment ROI" has been implemented into an executable standard:

**Personal ROI = Attributable value over validation period ÷ Total cost per person over validation period > 1. **

Let me first explain the origin of this formula. It is not derived from management textbooks, but is forced out by budgets. The human budget we can invest every month is limited, and the term "limiting resource" is not abstract here; and as long as there is no way to measure and evaluate each person's output, there is no way to answer "Who should the next budget be allocated to?"; without prioritization, personnel management degenerates into impression points and seniority points. After being stuck here and working backwards, the standard can only look like this: establish an attributable value caliber for each person, and answer the investment and ranking questions with whether the ROI is greater than 1. This is also a recurring pattern throughout the book, which will be formally named in Chapter 7: organizational standards are not promulgated, but are forced out when the methodology reaches a stuck point.

Three definitions turn this formula from a slogan into a tool. Full Cost: Compensation and benefits are just the starting point, plus recruitment, management collaboration, tool resources and opportunity costs, the true price of a person is usually 1.5 to 2 times the salary bill. Value: It can be direct value such as revenue, gross profit, and cost savings, or long-term value such as verified standards, reusable capabilities, and customer assets. But long-term value must define proxy indicators before committing, that is, surrogate signals of progress that can be observed in advance, without allowing for hindsight narratives when results are poor (this prohibition comes from the feedback evasion in Chapter 2: hindsight narratives are the most honorable form of that old gene). Validation window: Changes with the organizational stage, early 3 months, mid-term 6 months, late 9 to 12 months; all stages look at leading indicators on a monthly basis. The stage can change the verification cycle, but value acceptance cannot be canceled.

This standard applies to everyone, including founders, but the value indicators are different for different positions. After it runs, it is a flywheel: People investment → Create attributable value within the window → Personal ROI > 1 → Release more cash and management bandwidth → Continue to invest in people and long-term capabilities. If it is greater than 1 in each round, personnel investment snowballs; if it is continuously less than 1, the scale itself becomes a liability. An organization does not become stronger with more people, but every person invested must make the next growth easier. The flywheel spinning in the opposite direction has a familiar name, called "big company disease": the number of people is increasing, the quality of judgment per person is declining, and all the extra manpower is used to absorb the coordination costs generated by the extra manpower.

Give this standard a complete example and see how it looks on paper. A newly appointed head of marketing at a mid-stage company (validation window 6 months) has five lines on his ROI card. Value indicators: The direct value is the gross profit contribution of qualified leads that can be attributed within 6 months; the agency indicator of long-term value (written down now and not allowed to be discussed after expiration) is the verification standard for precipitated delivery channels ≥ 3, and the monthly trend of brand natural traffic turning positive. Full cost: Salary and benefits × 1.7 (including experience coefficients for recruitment, management, tools and opportunity costs, each company can calibrate itself). Attribution method: The gross profit of leads is traced to the source of CRM, and the channel standard is based on the verification standard of "newcomers can achieve the same level of results". Leading indicators (monthly check): launch ROI trend, standard document output rhythm. Stop condition: In the third month, the leading indicators are all below the lower limit and there is no credible explanation. Review the market in advance without waiting for the window to end. It took about twenty minutes to write this card, and what it bought was: the review six months later is no longer a narrative competition, but a table comparison with the caliber of card construction.

The AI ​​era has added a new denominator term to this standard, as well as a new kind of waste. The new waste is: Spending scarce judgment on free execution, letting the most expensive judge do the sorting, retrieval, and formatting that AI can complete in three seconds is equivalent to using limited reactants to do the work of the solvent. Auditing this waste is simple: look at the calendars of the people with the strongest judgment and count how many hours are being executed. That number multiplied by their real hourly rate is the monthly rent you pay for "Reluctant to Change Process."

7. Three typical mismatches: using the resources of layer A as layer B #

Each of the three layers of limiting resources has its own accounting currency. The typical form of mismatch is to use the currency of one layer to pay the bills of another layer. The three most common ones:Mismatch 1: Use money to buy judgment and budget to replace thinking. When decisions are uncertain, the most comfortable action is to spend money: buy a consulting report, purchase a system, hire another executive, so that "money has been spent" creates the illusion that "a decision has been made". MD Anderson in Chapter 2 is the ultimate example of this mismatch: Sixty-two million dollars buys not judgment, but sixty months of deferred judgment. Money can buy the input of judgment (information, tools, talents), but not the judgment itself; for organizations that treat procurement as decision-making, the more sufficient the budget, the faster the judgment will shrink.

Mismatch 2: Use time to make up for judgment and work overtime to cover up gaps. When you are unsure about the direction, try all directions over a longer period of time. This was considered a stupid method in an era when execution was expensive, but it was pure self-deception in an era when execution was free: Chapter 3 said that the feeling of execution without standards is exactly enrichment. The team has been working overtime for three consecutive months and the output is mediocre. Most of the time, it’s not that the people are incompetent, but that a judgment that should have been made in the first week has been decently delayed by three months of busyness. Overtime is the most expensive painkiller for judgment gaps. It relieves pain but does not cure the disease.

Mismatch 3: Use judgment to execute, overkill monthly rent. This account has been calculated at the end of the previous section: let the person with the strongest judgment in the organization fill the calendar with execution tasks. This is the most insidious of the three mismatches, because it appears to be hardworking and hands-on, and may even be praised as a virtue. The test just needs to ask: How much worse would the results be if an AI or a human who was half as cheap were to do this? The answer is "almost" every hour is an idling of judgment resources.

The three mismatches point to the same root cause: failure to realize that the three levels of resources are not commutable. More money cannot make your judgment better, long time cannot make the direction correct, and strong judgment should not change the amount of execution. The first discipline of limiting resources is to let each layer's currency pay its own layer's bills.

8. Boundaries of Judgment #

Three boundaries, stop this conclusion.

First, ROI > 1 is not short-termism. Long-term value is clearly allowed in the numerator: standards, capabilities, and trust can all be accounted for. Amazon’s AWS was all negative in terms of direct returns in the first few years. The line of defense is no longer "only recognizing short-term returns", but "long-term value must define proxy indicators in advance": the only way to distinguish long-term investment from indefinite waste is to make it clear before investing "what signals to watch midway, when to review, and under what conditions to stop." It is difficult to explain the "long-term layout" of these three items, so they will be treated as waste by default.

Second, the three-layer limiting resource is "the one that is currently the most stuck", not the identity tag. In early stage companies, both money and judgment are often scarce at the same time, and the limiting resource changes position the day the financing is received. The key point of management is not to memorize "the company is short of money and the individual is short of time", but to know which one you are stuck in at any time. The limiting resource will migrate, and identifying it is a judgment in itself.

Third, tell the truth about the quality of the sample. The Amazon sample's standard text and capital flow are all public first-hand materials, which are of sufficient quality; Zhang Yiming's statement is a first-hand interview, but it describes the strategy rather than audited cause and effect; Evergrande's figures are from public documents, which are complete; the origin of the ROI formula is my own self-report, and I will discount it based on the self-report. This chapter also lacks a sample: a case of deviating from words and deeds by "saying AI is a priority but not including AI in the budget". There are a lot of companies like this out there, but none of them are willing to have their budgets publicly dissected. Until a nameable sample is found, this slot is filled by the following Monday test: your own budget sheet, which is the sample closest to you.

What to Do Monday Morning (No. 1 perspective) #

Two tables, forty minutes:

First shot: Budget development. Calling out all expenditures in the last quarter, there are only two categories: execution (output of work that can be replaced by AI or outsourcing) and judgment construction (standard refinement, context construction, talent's thinking ability). Work out the ratio. This ratio is your company’s true strategy. If it doesn’t tell the same story as your strategic statement, trust the proportions, not the manifesto. Then look at the line on AI-related spending: If you’ve said “AI first” in an all-hands meeting and this line is close to zero, you’ve just completed forensics on the missing sample in this chapter at your company.

Second: ROI Card. Build a card for each member (starting with yourself): Freeze a 3, 6, or 9-12 month validation window by current organizational stage and write down value metrics, full costs, attribution methods, and stopping criteria. There are two rules: proxy indicators of long-term value must be written now and are not allowed to be replenished when due; look at the leading indicators every month and review them based on the caliber when the card is established when they are due, and no change of ruler is allowed.

Note (individual and team perspective): Do the same thing for your own time: record your time flow for a week and divide it into three columns: "Execution", "Judgment" and "Neither Execution nor Judgment". The third column is usually intimidatingly large; in the first column, mark the parts that the AI ​​can pick up today. Where should your limiting resource flow? It will naturally appear after the two annotations are completed.

Quotable Lines1. Don’t listen to what the company says, watch where the money goes. #

2. Excess resources talk about scale, and scarce resources talk about ROI.
3. The most expensive waste in the AI ​​era is spending scarce judgment on free execution.
4. An organization does not become stronger with more people, but every person invested must make the next growth easier.
5. An ROI greater than 1 is a snowball, and an expansion less than 1 only replicates losses faster.
6. The verification cycle can be changed by stage, but value acceptance cannot be canceled.
7. Making one more product does not mean adding another revenue line, but multiplying the complexity.
8. To correct deviations and only cut down on business without changing standards is tantamount to cosmetic treatment of diseased areas.
9. Long-termism, which only emphasizes benefits but ignores costs, is another Interference Method.
10. If you spend the wrong money, you’ll get a bill; if you spend the wrong time, you won’t even get a receipt.
11. A single investment is allowed to fail and the ranking criteria are not allowed to drift.
12. A lot of money cannot make your judgment better, and a long time cannot make your direction correct—limiting resources cannot be exchanged.
13. Overtime is the most expensive painkiller for judgment gaps: it relieves pain but does not cure the disease.

Chapter Acceptance Self-Check (compare with the five acceptance standards of the chapter) #

1. The assertion can be restated in one sentence ✓, and is an inference of the core assertion (adaptive insight/judgment is scarce → scarce resources require ROI discipline).
2. Whiteboard framework diagram ✓ (three-layer funnel: money → time → judgment, the same problem at each layer; aside from the human flywheel).
3. External comparison and data ✓: Winner side Amazon (first-hand text of shareholder letter + numbers) + S5 Zhang Yiming (first-hand interview with "Finance", including double-sided accounting of self-acknowledged costs); loser side Evergrande (public document-level data, including details of "correction and non-change of standards"); missing samples of "deviation between words and deeds" have been truthfully marked and filled in with Monday testing; new ROI > 1 on the internal side Standard parentage explanation (, 2026-06 forced by labor budget constraints).
4. 13 golden sentence candidates ✓.
5. "What to Do Monday Morning" Two tables from the No. 1 perspective + personal notes ✓.