Wealth

I don't think about this as a diversification problem. At any given stage, capital has a highest-leverage destination. Right now, mine is the business. I'm reinvesting into the thing I have the most operational control over and the most information about. Passive market exposure is where surplus goes once the business no longer needs it. Running that order backwards produces mediocre outcomes at both ends.

The working hierarchy is roughly: business equity, then broad market exposure, then property, then anything speculative. Property carries high entry costs, poor liquidity and a maintenance burden comparable to running a business. It lacks the compounding leverage of something you control directly. At my current stage it makes more sense to own less of what I understand deeply than more of what I understand loosely.

There's a practical corollary: two businesses maximum, one primary and complex, one secondary and simple. Beyond that, neither compounds, because the binding constraint isn't capital. It's attention. Owning three businesses in practice means owning three underperforming ones. I've watched that happen to enough people to treat the limit as structural rather than aspirational.

The honest problem is that this whole hierarchy depends on the business actually working. If it doesn't produce meaningful equity, the sequencing argument quietly degrades into a rationalisation for skipping passive investing during the best compounding years available. I've held the view long enough to recognise that "invest in yourself first" is one of the most comfortable excuses in personal finance. At some point surplus has to flow into something that doesn't need my attention on a Tuesday.

I've revised this more than once. Earlier versions were far steeper: near-zero passive exposure during any build phase. That turned out to be too rigid. It created a false binary between building the business and building a position. Now I run them in parallel. The business gets priority on new capital. The passive side runs continuously even when the contributions are small enough to feel pointless. They aren't. That's the entire mechanism.

Know which asset class deserves your best capital right now — not which one deserves it in the abstract.

The distinction is trivial to state and genuinely hard to act on. Almost all income early in a working life is treadmill income (service work, consulting, employment), where output scales directly with hours. The work is progressively shifting the balance. Route treadmill income into things that generate returns without proportional effort. Reduce the share of total income that switches off the moment you do.

The practical implication is that monthly income is the wrong progress metric. The more useful one is the ratio of asset income to living costs. When that ratio approaches one, the financial pressure driving most decisions disappears. That pressure is where the genuinely bad decisions live. Every poor commercial decision I can point to was made from a position of needing the outcome.

Which makes runway the number I actually track. Not how much I earn, but how many months I can operate without active income. Longer runway is more optionality: time to take a strategic risk, room to wait for a better opportunity, permission to say no. Runway is the thing being bought. Framing it that way makes a lot of otherwise-difficult spending decisions resolve themselves.

Cost discipline is the instrument that extends runway without needing more income. It's the half people skip because it's less exciting than earning. It isn't austerity. It's separating two categories. Costs that compound against you: lifestyle inflation, impulse upgrades, subscriptions nobody has audited in a year. Costs that compound for you: education, health, relationships, tools that genuinely multiply output. Those two categories look identical on a bank statement and behave nothing alike.

The goal isn't a bigger income number. It's a smaller ratio of required effort to a covered life.

I run a three-tier framework based on cost. Below the first threshold, the decision gets under ten minutes and then it's closed. Between the first and second, a short delay and a quick cost-benefit pass. At the top tier: a full opportunity-cost audit. What else could this capital be doing over the next twelve months? Is that alternative genuinely better, or just more comfortable?

The point is to spend the right amount of attention on each category rather than to be restrictive across the board. The failure mode most people fall into isn't overspending. It's agonising over small purchases while waving through large ones on momentum. The small ones feel discretionary. The large ones feel inevitable. They aren't.

There's a philosophical corollary I apply at every tier: experiences get a deliberate bias. Physical goods depreciate reliably and predictably. The return on a trip taken or a skill developed compounds differently and less legibly. That's precisely why it's systematically underweighted. A purchase that produces a memory very rarely regrets itself the way an impulsive upgrade does.

The one rule that applies across all three tiers: never pay for the same lesson twice. The first time something goes wrong it's tuition. The second time it's carelessness. The correct response to an expensive mistake isn't resolve. Resolve decays. The fix is a micro-system, a checklist or a default, that makes the same error structurally difficult to repeat.

Match cognitive effort to financial weight. Overthinking cheap decisions is itself a cost.

Before anything else, I ask whether I actually understand the business. Not whether I've heard of it. Not whether the sector broadly makes sense. Whether I can explain, without looking anything up, how it earns money and who its customers are. And what would have to stay true for it to keep earning more in ten years. If I can't answer that cleanly, I stop there. Everything downstream is irrelevant without it. The analysis that follows a failed first gate is usually just elaborate confirmation.

Past that threshold there's a quantitative screen: consistency of returns on capital, sustained earnings growth over a long window, credible forward projections. These are filters rather than guarantees. Failing them doesn't rule something out permanently. It means considerably more scrutiny before proceeding. Things that clear all three go on a watchlist, not into the portfolio.

The watchlist is where most of the time goes, and price discipline is the part I find hardest. The rule is to buy only when the price sits meaningfully below what I estimate the thing is worth. Not when the business looks compelling. Not when the research feels finished. Specifically when there's enough margin to absorb my own estimate being wrong. That's structurally difficult to hold. You can spend months watching something you understand well, want to own, and never get offered at a price that justifies it.

The framework fails in one consistent direction: I apply it least rigorously to things that feel culturally obvious or that I'm already attached to. I've made purchases that would have been screened out had I run the gate honestly first. A process applied selectively isn't a process. It's a rationalisation engine with extra steps. There is always an available argument for why this particular case is the exception.

I also hold a small position in a category that would fail my own first gate outright. I know that. I keep it explicitly carved out: named as speculative, sized accordingly, held in a separate mental account rather than blended into the main framework. Whether that separation is principled or merely convenient is a question I haven't fully settled. What I have settled is that naming the carve-out is better than pretending the framework is applied uniformly when it plainly isn't.

A research process applied inconsistently isn't a framework. It's a rationalisation system.

I used to think accurate tracking meant tracking everything. I spent real time with tools built to capture every transaction and produce detailed category breakdowns. They were accurate in theory and abandoned in practice, every time. What I run now is deliberately coarser. Balances at the macro level (cash, investments, liabilities) on a monthly cadence. Recurring costs smoothed to averages. Large one-offs logged when they happen. That's close to all of it.

The reasoning is that I'm not optimising for precision, I'm maintaining a signal. The signal answers two questions: is cash growing month over month, and is the overall position moving the right way? If both are yes, nothing below that level changes my behaviour. If either is no, I go and look. The system exists to tell me which of those two situations I'm in. Not what I spent on food last October.

That buys a tolerance for good-enough numbers. The expense model runs on averages with a buffer for variance, and I know the buffer is imprecise. What matters more: I've maintained the thing consistently for years with almost no friction. A model with a meaningful margin of error that I actually use produces better decisions than a forensically accurate ledger I stop updating.

The honest failure: for a long stretch my expense buckets were set too low. The modelled surplus looked healthy while the cash balance wasn't actually growing, which meant the model was simply wrong, and comfortable about it. Rebuilding the expense side with real numbers was unpleasant because it revealed a much smaller surplus than I'd assumed. That's the specific vulnerability of coarse tracking. Bad inputs generate a false sense of security. The model gets more confident as it drifts further from reality.

The safeguard is the monthly reconciliation: does the actual balance match what the model said it would be. Skipping that check is where the whole thing breaks down, because the model doesn't announce that it's stopped being true. It just keeps producing numbers, and you keep making decisions on a picture that expired months ago.

Track what changes your decisions. Not everything that changes.

Some percentage of what I earn each year quietly disappears into things that didn't work. Tools bought and abandoned. A strategy tested and killed. The occasional genuinely expensive lesson. Most people treat these as failures of judgement. I budget for them as a normal operating cost, in the same way a business budgets for wastage.

The reason to name it in advance is that it changes the response. An unbudgeted mistake gets classified as a catastrophe, and catastrophes trigger over-correction: one bad outcome reshaping an entire strategy that was otherwise working. A budgeted mistake gets classified as tuition, which is usually the accurate reading and leads to a proportionate response.

The corollary matters more than the budget. Tuition is only tuition if you actually extract the lesson, and the extraction has to be structural rather than emotional. Feeling bad about a decision is not a control. A checklist, a changed default, a rule that makes the failure mode harder to reach. Those are controls. The measure of whether a lesson landed isn't how much it stung. It's whether the system changed shape afterwards.

Budget for being wrong. It's the cheapest way to keep being wrong from getting expensive.