4 min read
What transfer pricing taught me about where profit actually lives
Three months of benchmarking models in EY's Ho Chi Minh City office, and the strange, clarifying discovery that a company's profit has an address — and someone has to defend it.
- Career notes
- Markets & investing
- Saigon life
Before this summer I would have told you that a company's profit is a number near the bottom of an income statement. I'd now tell you it's a number with an address, a passport and a paper trail — and that somewhere, someone spends an entire career arguing about which of those it should have.
Transfer pricing is the least glamorous phrase in finance and one of the more interesting things I have been paid to think about.
The premise, in one paragraph
A multinational is legally not one company. It's dozens, in dozens of countries, selling to each other constantly. When the Vietnamese manufacturing entity sells to the Singaporean distribution entity, somebody has to set that price — and that price decides where the profit lands, and therefore which government gets to tax it.
The governing idea, the OECD's arm's length principle, is disarmingly simple: price it the way two unrelated companies would have. Fifteen models later, I stopped calling it simple.
What the work actually is
I spent the summer building financial benchmarking and valuation models across logistics, packaging and insurance clients. The mechanics look like comparable-company analysis, but the question underneath is different from anything I'd done in a valuation class.
You are not asking what a company is worth. You're asking what a normal margin looks like for a business that performs this exact function, carries this exact risk, in this exact market — and then whether your client's entity is sitting inside that range or outside it.
The pattern, every time:
Characterise the entity. What does it actually do — manufacture on order, distribute at limited risk, hold the intangibles? Everything downstream depends on this and it is judgement, not arithmetic.
Build the comparable set. Independent companies doing something close enough, screened and then read one by one, because screens are confidently wrong about a surprising number of them.
Test the result. Return on assets, operating margin, whatever the transaction calls for, against an interquartile range rather than a single number.
Write down why. Every rejection from the comparable set is a decision someone may ask you to justify in three years.
The part that surprised me
Every number in a transfer pricing file is defensible or it is worthless. There is no third state.
In a valuation, being roughly right is often enough — you build a range, you argue for a point inside it, and reasonable people disagree. Here the range is the answer, and it has to survive a reader whose professional purpose is to disagree with it, sometimes years later, with the enormous advantage of knowing how the year actually turned out.
That changes how you write, not just how you calculate. I started putting my assumptions where a sceptical reader would find them first instead of tucking them into a footnote, which is a habit I'd recommend to anyone who builds anything in Excel.
Cyclical and defensive, learned the hard way
Sector matters more than I expected. Logistics margins swing with freight cycles, so a comparable set that looks tight in one year sprawls in the next and you need multi-year averaging to say anything honest. Packaging is steadier and rewards a narrower set. Insurance is essentially its own planet, with its own vocabulary and its own idea of what an asset is.
The lesson generalised: the shape of a business's volatility is part of the comparison, not noise around it. I'd been treating volatility as something to smooth. It's information.
What I took back to Indiana
Ask what an entity does before you ask what it earns. The function determines the return, not the other way round.
A range you can defend beats a point estimate you can't.
“Reasonable” is a claim, not an adjective. If you can't say what it's reasonable compared to, you haven't said anything.
Write the assumption down where the reviewer will hit it, not where it's tidy.
I don't know yet whether tax is where I want to spend a career — I suspect not, and I suspect the modelling instincts come with me wherever it is. But I've stopped reading consolidated financials the same way. Somewhere under that one clean profit number is a map, and the map has borders drawn on it, and someone in an office in District 1 spent a summer defending exactly where they fall.