S5: Build your money data

English edition. S1–S6 are available in English; S7–S9 are published in Japanese.

Building wealth is data design. Seeing where money comes from and where it goes is wealth.

What you get from this skill

Every month, your business and your personal finances close.

They close on the same day each month. Last month’s income and spending, fixed costs, cash, assets, and debts are numbers by early the following month
You read the same indicators every month: month-over-month change, how many months your cash lasts, distance to your safety line, upcoming taxes
When a threshold is crossed, an alert fires. Nobody has to remember to check
You can see where this month sits against a three-year plan

At that point a one-person company is looking at the kind of indicators a board looks at. Many small businesses can’t produce last month’s numbers on request. A solo operator whose books close every month is already ahead of most of them.

Part 1: why data

Owning your capital has always been hard

Before banks and currency, how did people save? Livestock. Turn the harvest into cattle and sheep — they don’t rot, and they multiply. That was capital. And it vanished easily: a flood, a predator, and the savings were gone. When one event can wipe out your savings, owning your capital is very hard.

We don’t keep cattle anymore. Instead we live inside something far cleverer at draining us: inviting shopping, subscriptions that quietly multiply, advice with no visible price. Where your money goes and where you actually make it gets harder to see every year.

So building wealth isn’t willpower or investing tricks. It’s a design problem: how to make income and spending visible as data.

Wealth is visibility

People picture wealth as a big balance, a lavish life, or big revenue. Those are results. The thing itself is visibility.

A wealthy person: someone who knows exactly where their money comes from and where it goes.

Look at enough businesses and the split is instant: those with visibility have money; those without don’t.

Stop the leaks before chasing more

If less flows out than flows in, savings grow even on a small income. Keep outflow minimal and you reach a state where you’re fine even when nothing comes in. Owners in that state stop grasping. Without panic, they don’t take bad deals.

Hand over your data and you become the one being sold to

If capital is data, that data flows to software platforms — “we’ll manage all your assets for you.” Put AI on top and you become the person receiving “here’s what you should do with your assets.” A subscription that tracks all your properties is, to the companies behind it, a superb list of renovation and rebuilding leads. Funds are similar: you get reports on what your money is in, but not as your own data, in a form you can give to your own AI. You grow, but you don’t hold the data.

Don’t stop the spending — see it

Telling a partner to justify every purchase wrecks the relationship, and nobody keeps a paper ledger anymore. The goal isn’t to stop spending; it’s to see how much there is. Once you can see it, you can discuss it without a fight. This is where AI agents change everything: no ledger habit required, and no more “money spread out in the open with nobody watching.”

Without visibility, decisions have no ground

High income without visibility means you can’t explain where $1,000 a month disappears — so every change in circumstances knocks you around. Companies do this too: a popular business with money coming in, where nobody, owner included, can see the numbers. Meetings have no data, so they go in circles. It works until the owner gets sick or the best employee leaves.

“Should I invest?” “Should I pay down debt first?” “Is my emergency fund enough?” For a company: retain earnings, reinvest, or repay to cut interest? None of these can be answered without your own financial data. And combined with what you learned about your market in S2 and S3, you can see things like “cash will be tight next year.”

The wealth ladder

Step 1: Secure your safety
  Make monthly cash flow positive; build an emergency fund
  → Goal: income minus spending is positive and stable
  ↓
Step 2: Clear high-interest debt and cut risk
  Payday loans, credit-card balances, other high-interest debt
  → Goal: zero high-interest balances
  ↓
Step 3: Start investing small with what's left
  Long-term and diversified, with money you can afford to lose
  → Goal: surplus cash is invested automatically
  ↓
Step 4: Scale with experience
  Expand into business investment, property, and more
  → Goal: both linear and compounding income are running

Step 2 is about high-interest debt — not a normal-rate loan from a bank. Borrowing more on top of expensive debt to start a business, betting on a big comeback, is a fantasy. Pay down and reduce risk first. Taking steps in order isn’t retreat; it’s the foundation for the next step. Step 4’s two engines are designed in S8 (Japanese).

Copy-paste path: the shortest route through S5

Paste these into your AI agent one at a time, in order. Replace everything in 【brackets】 first. Run personal and business finances separately (Step 1). Details and reasons are in Parts 2 and 3.

Never give the AI credentials. No account numbers, card numbers, passwords, or logins. Give it downloaded statements (CSV) and balances — nothing else.

① Create the structure and ledgers (Step 1)

Create ontology/finance/ (for the business, create finance/ in the business's own folder instead).
Inside: monthly/ (with this month's folder 【YYYYMM】), assets/, investment/
Ledgers: accounts.md (accounts and cards), recurring.md (recurring contracts), fixed_assets.md (fixed assets), tax_calendar.md (tax calendar)
Create headings and empty table frames only.
Never mix personal and business finance folders.

② List every account (the master list)

I'll name every bank account, card, and payment service I use, one by one. List them in finance/accounts.md.
Each row: name (no numbers) / type / business or personal / how statements arrive (CSV, accounting-software feed, none)
Flag anything used for both business and personal, and summarize those at the end.

【e.g. checking account at …, credit card …, PayPal, Amazon】

③ Read one statement and run the first analysis (Step 4)

Read the CSV in finance/monthly/【YYYYMM】/.
- List every subscription that looks unnecessary
- Which spending is routine?
- What do fees add up to per year?
- Any day-of-week pattern?
- Who have I been paying for, and how much?
I'm not deciding to stop anything yet. Just list what you find.

④ Write the first version of your bookkeeping rules (Step 5 · business)

Read the business statements in finance/monthly/【YYYYMM】/ and tentatively assign categories.
Ask me only about the ones you're unsure of, one at a time.
Write my answers into finance/bookkeeping-rules.md in the form "A is B" (e.g. "AI tool subscriptions are software expenses").
The test for a business expense: can I tell the business story of this cost *today*?

⑤ Build the tax calendar (Step 6)

From the tax returns, assessments, and payment notices I'll give you, build finance/tax_calendar.md:
- Confirmed: due date, tax, tax year, amount, how it's paid, status (always cite the document)
- Forecast: estimates from last year's figures, clearly marked "estimate"
- When the heavy payments end
- Questions for my accountant
Don't judge whether the amounts are right or suggest tax strategy — my accountant does that.

⑥ Close last month (Step 8 · monthly)

Read the ledgers in finance/ and the data in finance/monthly/【YYYYMM】/, and close 【Month YYYY】.
Order: collect → reconcile → adjust → read
1. Does every account in accounts.md appear in this month's data? (anything missing = "not imported")
2. Differences between book and actual balances (keep as "unexplained" until the cause is known)
3. Book the monthly share of annual contracts and of depreciation
4. The seven indicators (with month-over-month) and the alert checks
5. Has the month closed? (check each condition)
Save to finance/monthly/【YYYYMM】/monthly-close.md in the format from Step 9.
Draft the "What happened" section in 3–5 lines; I'll rewrite it.

⑦ Make one decision (Step 8)

Read this month's close and propose 3 candidate decisions for this month, based on the numbers.
For each: the numbers behind it, and what happens if I don't decide.
I make the decision. Log the one I choose in the action log.

Part 2: build the monthly close

Everything in this part serves one purpose: what you need to close every month.

Step 1: design `finance/`

Part of the close: where the data lives

Under the Ontology you built in S1, create finance/:

ontology/finance/
  ├── monthly/          monthly data and closes
  │     ├── 202609/     one folder per month
  │     └── 202610/
  ├── assets/           snapshots of assets and debts (updated monthly)
  ├── investment/       investment tracking
  ├── property/         optional, if you hold a lot of property
  │
  ├── accounts.md       accounts and cards   ← the master list for reconciliation
  ├── recurring.md      recurring contracts  ← source for prepaid/accrued items
  ├── fixed_assets.md   fixed assets         ← source for depreciation
  └── tax_calendar.md   tax calendar         ← the next 12 months of payments

The top half is data; the bottom half is ledgers — fixed files you reuse every month. Build them once; update when something changes.

Ledger	Why it exists
`accounts.md`	the master list that tells you what’s missing — every account, card, and payment service, each assigned to business or personal
`recurring.md`	spreads annual contracts across months
`fixed_assets.md`	books depreciation monthly
`tax_calendar.md`	when and how much you’ll pay over the next year

Two rules:

① Never mix personal and business finance/. This is where AI gets confused most. Personal money data goes in ontology/finance/ by default — inside the Ontology, the AI can read it alongside everything else that informs your decisions (people, businesses, your action log). The business gets its own finance/ in the business’s folder. If you’d rather not keep money data where the AI reads widely, manage it in a separate folder outside ontology/ — with the same structure (month folders and four ledgers).

② One folder per month under monthly/. Each month’s records end up in one place, and year-end filing gets much easier.

Step 2: how data gets in — five doors, and CSV

Part of the close: the inputs

Door	Good for	Weakness
Manual entry	cash, small items	doesn’t last
Word of mouth	”this payment is coming next month”	rarely recorded
Photos of receipts	evidence	poor for analysis
Email alerts	bank transaction notices	set up per service
CSV	the main road	you download it by hand

CSV is the main road. Most banks, card issuers, brokerages, and payment services offer CSV export, because people use it for bookkeeping and tax filing. Check what each of yours provides.

Drop CSVs straight into finance/ — AI reads them like Markdown. For data you’ll read repeatedly, convert to structured JSON for speed (Step 10).

Don’t leave files in Downloads. “Read the CSV in my Downloads folder” works, but Downloads isn’t storage. Move every file into finance/.

Keep receipts as evidence — that matters. Analysis and categorization happen on the AI side.

Step 3: gather your personal accounts in one place

Part of the close: the personal data source

Connect your bank accounts, cards, brokerage, and payment apps to one personal budgeting app that pulls transactions automatically. Then once a month, export a CSV into finance/. That’s the whole personal routine. Card statements settle at month-end, so the end of the month or the first days of the next is the practical time.

As of October 2026: account-aggregating budgeting apps exist in most countries (bank connections through open banking in the UK and EU, through aggregation services in the US). A modest paid tier is usually enough; free tiers tend to limit history or connections. Features and prices change — check what’s available where you live.

Step 4: run your first analysis

Part of the close: feeling that data drives decisions

Before building the whole system, run it once by hand. One statement, one question (prompt ③). That alone produces findings.

What actually came up. In one class, participants loaded their own statements and asked these questions on the spot. Findings included: a duplicate phone plan still running after switching carriers; a daily coffee adding up to a noticeable monthly sum; ATM fees for the year and expired reward points; taxi rides clustered on weekends (19 on Saturdays, 13 on Sundays); subscriptions piling up, and nearly all household goods bought from one retailer.

The amounts aren’t the point. People can’t make level-headed decisions when the information is missing and unorganized. If that’s true for one person, it’s truer for a company.

You don’t have to decide to stop anything. The outcome isn’t “no more taxis.” It’s that you can now reconsider, with data, whether those rides are needed.

Step 5: let AI run the business bookkeeping

Part of the close: business data and automated categorization

Personal data you download by hand. Business data can go further: connect your accounting software to your agent, and have the AI not just fetch transactions but categorize and record them.

MCP is not a sync tool or a downloader. It’s how the AI goes inside the software and does the work for you.

The target pipeline:

Bank, cards, payment apps → accounting software pulls transactions automatically
  → AI reads uncategorized items and records them per your rules
  → a human approves only the exceptions
  → monthly, AI reads the reports and writes a financial summary into finance/

Four design points make it work.

① Split business and personal at the source. Fix each card and account as business or personal at the transaction level — the name on the account doesn’t matter. Business ones connect to the accounting software; personal ones to the budgeting app. Avoid cards used for both. If “everything on this card is business” is true from the start, the heaviest judgment in bookkeeping — business or personal? — disappears. Move subscriptions to the business card; use a toll tag for business driving. Shared costs like rent can stay personal and be allocated.

② Write the rules in words. AI can categorize because your rules are written down — in the S1 form, “A is B”:

Bookkeeping rules (example):
- AI tool subscriptions are software expenses
- Novels are personal. Other books are research, so they're business expenses
- The dedicated home office is 10.8% of the home, so rent and electricity are allocated at that rate
- The overseas co-living stay is not an expense: I can't tell its business story today

The test is one question: can you tell the business story of this cost today? Not a future dream — a present connection you can state in one sentence. The first version comes from one session (prompt ④); exceptions are added as they appear. From the second month, the AI applies the rules and you judge only new patterns.

Allocations work the same way: design the evidence, not the percentage. “The floor-area share of a room used only for work.” “A mileage log generated by matching sales meetings in the calendar to toll records.” Then ask your accountant not “what percentage should I use?” but “with this evidence, how far can we go?”

③ Never fully automate. The AI records; you handle exceptions and policy. Never remove the approval gate. Have it ask “may I do this?” before acting, keep a log of what it did, and turn on the audit trail in your accounting software, so you can see what the AI booked and what a person changed. Trustworthy AI bookkeeping is measured by how checkable it is, not by how automated.

④ Where sovereignty lives: the software is plumbing, the intelligence is the Ontology. Once built, you notice: the accounting software connects to banks and keeps up with tax law and filing formats. All the intelligent work — rules, allocation evidence, expense judgment — happens in your Ontology. Does that make the software unnecessary? No. Bank connections can’t be rebuilt by an individual, compliance for a modest monthly fee is good value, and above all you need a neutral ledger. An AI that records entries and then declares its own books correct is a closed loop that no accountant, tax authority, or bank can trust. Dropping the software for homemade books isn’t sovereignty; it’s isolation.

Real sovereignty is this: as long as your rules, judgment, and master financial data live on your side, the accounting software is a replaceable part. If prices rise, you’re not a hostage; if something better appears, you move. Use vendors fully; depend on none. The flip side of “hand over your data and you become the one being sold to”: keep the judgment on your side, and you’re not.

As of October 2026: whether an AI agent can connect to an accounting platform (through an official MCP server or API) varies by platform and by agent — one agent may connect where another is refused. Choose by two things: does the software offer an official connection, and does it actually connect from your agent? Try it on the day.

Stuck in a settings screen? Send a screenshot. Faster than any manual; it’s how class participants got their connections working.

A real case: the person behind this framework processed the bookkeeping for a business launch this way — about 380 transactions across seven months categorized by AI; the human only answered rule questions (“novels are personal,” “this room is used only for work”); 31 confirmed entries were posted in one batch through the connection, and the trial balance was checked automatically. The first version of the rules came from that single day’s conversation.

Step 6: look at taxes first

Part of the close: the biggest, most predictable future outflow

Most people who hit a cash crunch don’t trip on living costs or inventory — they trip on taxes and mandatory contributions. Yet these are predictable, because three dates are always out of step:

① Which year's income it's on (the tax year)
     ↓ months later
② When the amount is set (filing, assessment)
     ↓ months later
③ When you pay (the due dates)

Examples: in the US, self-employed people generally pay estimated tax in quarterly installments during the year; in the UK, Self Assessment can require payments on account toward the next year, based on the previous year’s bill. In many systems, a high-income year makes the following year’s payments heavy — and if income has dropped by then, the bills arrive with no cash to meet them.

Being taxed isn’t the frightening part. It’s having it arrive without warning.

Policy belongs to your accountant; monitoring belongs to your Ontology. Whether the amounts are right, tax strategy, and which rules apply are your accountant’s domain; this framework doesn’t replace that. The Ontology’s job is to watch: every month, have the AI compute “how much, in which month” for the next twelve months and the cash you’ll need, with a warning before each due date.

List the taxes — income tax, self-employment or payroll taxes, sales tax or VAT, property tax, vehicle tax; for a company, corporate tax and payroll obligations
Write how each is determined — based on last year’s income, or this year’s? That decides how predictable it is
Enter confirmed amounts from returns, assessments, and notices — always cite the document
Fill unconfirmed ones with estimates from last year — marked “estimate,” replaced when the notice arrives

## Tax calendar

### Confirmed
| Due date | Tax | Tax year | Amount | Paid via | Status |

### Forecast
| Month | Tax | Basis | Amount | Confidence |

### When the heavy payments end
  this year's income → next year's payments, as one line

### Questions for my accountant

Always write when the heavy load ends. After a one-off high-income year, the following payments are heavy — but temporary, and when they end is structurally fixed. Without writing that down, you’ll plan as if that level lasts forever and become needlessly pessimistic. And the income you need after the heavy period ends is often a smaller, concrete monthly target.

Bundling payments can hide them. If you pay taxes or utilities through a payment service billed to your card, the card statement may show only the total charge, not the items. Compare the amount charged to your bank with the sum of the card’s line items each month: the gap is what you paid in taxes and bills. The breakdown only exists in the notices themselves — so build the tax calendar from notices, and reconcile once a year.

Step 7: how your relationship with your accountant changes

When this is in place, the books are categorized before your accountant sees them. The split becomes: AI records and analyzes; the accountant sets policy and signs off, and meetings turn into deciding direction while looking at the numbers. Some accounting firms are already shifting from bookkeeping toward advisory work. How that plays out, nobody knows yet.

Part 3: read the close and run it

This is the body of S5.

Step 8: turn the close into a procedure

Pick a closing date. Start right after month-end, once card statements settle; aim to finish within ten business days, then work toward five. Same order every month:

collect → reconcile → adjust → read → decide

Collect — personal: export the CSV into the month folder. Business: the AI records uncategorized items; a person approves only the exceptions.
Reconcile — cash on hand vs. books; every account balance vs. the bank. Keep differences as “unexplained” until you know why. Never delete them.
Adjust — clear advances and reimbursements, book prepaid and accrued expenses, monthly depreciation, month-end inventory if you hold stock.
Read — compute the indicators and run the alerts (Step 9).
Decide — make one decision from the numbers; save the report in the month folder; update assets and investments.

Skip “adjust” and month-over-month stops meaning anything. That’s the difference between a transaction feed and a monthly close. Expense an annual contract in the month you pay it, and that month’s profit sinks. Skip monthly depreciation, and the month you buy equipment looks like a loss. So annual contracts go into the ledger and are booked monthly:

Recurring contracts (source for prepaid items)

Annual (spread monthly)
  card fees, insurance, annual software plans → annual ÷ 12, booked monthly

Monthly (no adjustment)
  rent, utilities, phone, monthly subscriptions → expense in the month paid

Decide “closed” by conditions, not memory:

every account in the master list appears in this month’s data
book balances match actual balances
zero unexplained differences
no double-counting
prepaid/accrued items and depreciation are booked
an asset snapshot exists

If any one is missing, the month isn’t closed. Bonus: a full year of closed month folders flows straight into year-end filing.

Step 9: choose the indicators you read every month

Same ones every month — otherwise there’s nothing to compare.

#	Indicator	Tells you
1	Is the month positive or negative?	the top-line verdict
2	Total fixed costs and month-over-month	structural or temporary increase
3	Fixed costs as a share of revenue	can the business carry its fixed costs
4	Cash on hand and months of runway	how much time you have
5	Distance to your safety line	how far from danger (Step 11)
6	Taxes due in the next 12 months, and cash needed	predictable outflows (Step 6)
7	Assets, debts, investment performance	is it accumulating

Runway, two ways. Indicator 4 is the heart of this skill. Not the balance — how many months it lasts.

months of runway = balance ÷ monthly cash need
monthly cash need = living costs (or fixed costs) + that month's tax payments

Compute it for cash and for liquid assets:

Cash runway — can you pay right now? Under one month, every month-end is a tightrope
Liquid-asset runway — including what you could sell quickly (stocks, funds). Exclude property — it isn’t liquid

Separately they tell different stories. On cash alone you may look one month from trouble while liquid assets cover two years. Or plenty of investments but thin cash can freeze every monthly decision. Only both together tell you whether to worry.

Write alert conditions in advance — seeing isn’t enough; you’ll miss things:

- the month turned negative
- fixed costs rose month over month
- cash is approaching the safety line
- a tax payment is near
- an account wasn't imported
- an unexplained difference exists

Thresholds are yours to set from your own numbers; what generalizes is the shape of the conditions.

The monthly close, as a file in finance/monthly/YYYYMM/:

## Monthly close: Month YYYY

### 1. Profit and loss
  recurring income
  recurring costs (living or fixed / variable)
  taxes and contributions
  one-off costs            ← non-recurring, kept separate
  ─────────────────
  recurring result
  capital spending (outside P&L)  ← equipment isn't an expense
  total cash out

  accrual adjustments
    annual contracts spread    +
    monthly depreciation       +
    capital spending removed   ▲

### 2. Balance sheet
  assets (cash, securities, property) / debts / net worth

### 3. Indicators (seven, with month-over-month)

### 4. Alerts (results of the conditions)

### 5. What happened
  3–5 lines on what this month was about

### 6. Close status
  closed / not closed — and what's left

Always write section 5. Next month you won’t remember. A few lines explaining a loss makes the next comparison instant. Separating one-off costs and capital spending is the point of this format — the month you bought equipment, paid a cancellation fee, or had several tax payments land shouldn’t be mixed with your normal costs.

A real case: the first close didn’t close. When the person behind this framework first ran this procedure on their own previous month, the income-and-spending analysis was already detailed — yet the month didn’t close. Three months of taxes and contributions didn’t exist as line items at all (paid through a bill-payment service, they never appeared on the card’s itemized statement; only “amount charged vs. sum of items” revealed them). Writing the account master list immediately exposed an account that wasn’t being imported. A double-count fixed the month before had come back when another connection was added. Asset snapshots were missing for two months, so four of the seven indicators couldn’t be computed. A currency conversion never appeared on the asset table. Every finding came from the balance side — none would have shown up from analyzing spending, however carefully.

Step 10: keep it cheap enough to run every month

Store structured data. Convert CSVs to JSON instead of re-reading raw files each time.
Limit full re-reads — weekly, say, not every run.
Start a fresh conversation for each calculation. Long threads with history and data attached get more expensive. Keep work folders separate; run number-crunching in its own conversation.
Match the model to the job.

Job	Model	Reasoning effort
calculation, totals, tidying	lightweight	lowest
ideas, planning, policy	top-tier	higher

Using a top model for arithmetic is almost always waste. (Model names change every few months; the principle — light for calculation, top-tier for thinking — doesn’t.)

Step 11: read it over three years

Plan sustainability over three years, matching the S2 business plan. Then draw two lines on your fixed costs:

Line	Meaning
Defense line	below this, the business stops
Self-running line	you pay yourself and still retain earnings

Amounts depend entirely on your business; only the method generalizes:

Use months as the unit — not “how much short” but “how many months of lost revenue until it stops”
Keep the two lines separate — not stopping and paying yourself are different levels
Draw them from your own data — total fixed costs → which to cut, which to protect → months of cash needed

On the personal side, Step 1 of the wealth ladder (the emergency fund) is your defense line. Same shape of lines for company and person keeps decisions consistent.

What to cut, and what to protect. Read not only “what can I cut” but “which spending protects my ability to keep going”:

If your work is physically demanding:
→ invest in systems that make the work efficient
→ protect time and money for rest
→ don't treat health spending as something to cut

While the business is small:
→ the center of the plan is sustainability, not growth
→ move to the next step once the foundation is stable

Building wealth is turning ordinary, unremarkable days slightly positive, over and over. The first requirement is a setup in which you don’t collapse.

Step 12: assess risk objectively

Running a business on rough mental math is full of bias. “About right.” “It was fine last month.” Those small judgments quietly eat profit:

By feel:
"Materials are about $300 a month."
→ Actually $348. A $576 error per year.
→ Five items like that: nearly $3,000 a year.

From the books:
→ real numbers show what can be cut
→ give them to AI and patterns and anomalies surface automatically

Give the AI your daily notes and activity log alongside the closed numbers — the money data alone can’t say why a cost jumped.

Read the last three months of expense data and my daily notes.
Which cost items rose month over month?
In which months were fixed costs high relative to revenue?
What are the likely causes?

Loop it. The first analysis is usually thin because assumptions are missing — target client count, expected price, launch date. Add them and ask again:

Pass 1: current data only → "margins are thin"
   ↓ add the missing assumptions
Pass 2: with targets and deadlines → "N clients a month reaches the defense line"

Quality comes from the number of round trips.

Step 13: what to bet on, and how much

Don’t gamble. Win with systems and algorithms. Make waste visible, tighten costs, keep monthly cash flow positive, grow your cash. When it’s time to reinvest, risk roughly a tenth of total assets at most — an amount whose loss won’t be fatal.

No:  go all in on the deal you "just know" will work
     → gut calls are bias; this can be fatal

Yes: test within a tenth of total assets
     → you survive failure; success leaves knowledge and a track record
     → the data improves the next decision

Markets keep speeding up, and you’ll face change far more often than before. Bet everything and one miss takes you out. Small, repeated bets keep you in the game even while losing individual rounds — and the one who stays longest compounds and wins. This is a ceiling for investments, not a gate on everyday spending; indicators 2 and 3 handle that.

Where to invest next tends to emerge from your S4 role map and ecosystem design. Track investments in finance/investment/ and link them to the people database in S6, so you can see which people, companies, and areas actually pay off.

In the early years, paying yourself less to build retained earnings — for future borrowing capacity and the next investment — often yields more over a decade. Allocation is designed in S8 (Japanese).

Practice: close last month

Don’t stop at reading. Close one month for real. Only your own numbers show what’s missing.

Time: the first time, 2–3 hours of desk work, plus time inside each service’s settings (adding connections, checking balances — only you can do that). Expect the first close to take two sittings. After that, about 30 minutes.

Setup (first time only)

Pick the closing date and target — within ten business days to start
Create personal and business finance/ folders and month folders
List every account, card, and payment service, assign each to business or personal, and eliminate shared ones
Write a one-page closing checklist — item, owner, deadline — and update it every month

The master list decides everything. You can’t see what’s missing by looking only at what arrived.

Collect — does every listed account appear in this month’s data? Record missing ones as “not imported.”
Reconcile — count cash, match it to the books, match every balance. Keep differences as “unexplained.”

No transactions isn’t the same as not imported. Before chasing a broken connection, check whether that payment ever appears as line items at all — bill-payment services often don’t. The “charged amount vs. sum of items” gap tells you.

Changed a connection? Check for double-counting. Adding a feed can import the same purchase twice — once from the order, once from the card. And a line item that vanishes without the contract ending is a missing import.

Adjust — advances, prepaid and accrued items, monthly depreciation, month-end inventory.
Read — compute the indicators, compare with last month, run the alerts. Take an asset snapshot — without balances there’s no runway; this one page drives most of the indicators.
Decide — one decision from the numbers; save the report.
Hand over — send the monthly numbers to your accountant. Policy is theirs; monitoring is yours.

From next month: do what you can mid-month (weekly entry, payments matched the day after); record how many business days the close took, and shrink it.

Stage	Business days to close	State
1	20+	closes, but slips into the month after next
2	10–15	closes the same way every month
3	5–10	in time for decisions

For most small businesses, a steady close within ten business days is the realistic goal.

Are you done?

The question isn’t “do you have a sense of it?” It’s “is last month closed?”

Before this → S4: Multiply your revenue streams (and Milestone 1, in Japanese)
Next → S6: Manage your team with AI
Allocation and capital planning → S8 (Japanese)

SOVREN Framework is open-source. The English edition covers S1–S6; S7–S9 and the milestones are published in Japanese. Nothing here is tax, legal, or investment advice — confirm with qualified professionals where you operate.

Where it is practiced

SOVREN Framework is a free, open canonical text. It is tested in live classes in Yokohama, Japan, where people implement it in their own businesses with an instructor.

About the classes (Ontology incubation) Read the session reports