A family office is usually discussed as a structure: jurisdiction, licensing perimeter, AUM thresholds. But to the person running it, it is an enterprise with its own P&L — fixed costs, headcount, vendors, an IT budget and obligations to a single client who is also the shareholder. The difference between a working office and a decorative one rarely runs along the tax regime. It runs along two questions: what does it cost to service a unit of capital, and how many days does it take the family to get a reliable answer to "how much do we have and where".
The economics here are harder than they look. An office's costs are almost entirely fixed: salaries, rent, systems, audit, lawyers on retainer. Portfolio returns are volatile; costs are not, so the cost ratio rises in precisely the year the family is least willing to tolerate it. Hence the rule to adopt before the first hire: budget in advance and in basis points.
What follows is the operating contour of the office itself. Investment strategy is a separate conversation about the investment policy statement; jurisdiction choice and the SFO/MFO boundary are covered in the family office overview.
What the office costs, and what the bill is made of
The cleanest reference in absolute dollars comes from J.P. Morgan Private Bank's Global Family Office Report 2026 (333 single family offices across 30+ countries, average family net worth US$1.65bn). Median annual operating budget by asset tier: up to $250m — $0.9m; $250–500m — $1.7m; $501–999m — $3.3m; above $1bn — $6.6m, against $6.1m a year earlier. The bank names competition for talent, not vendor inflation, as the main driver of the increase.
The shape of the curve matters more than the sums: assets multiply between tiers while the budget grows far more slowly. Forge Community's benchmarking study puts average SFO operating costs at 0.41% of AUM, with the smallest offices (around $200m) running at 55 bps and the largest (around $12.5bn) at 17 bps. KPMG Private Enterprise with Agreus, in the Global Family Office Compensation Benchmark Report 2025, records the most common band at 0.6–1% of AUM — up from 0.1–0.5% in the 2023 reading.
The composition of the bill is stable: personnel accounts for 60–70% of operating costs per the UBS Global Family Office Report 2026, and Forge gives a similar picture. External providers absorb roughly 26% of the budget (J.P. Morgan 2026), with more than 40% of offices sending over 30% of spend outside.
The threshold follows directly. A minimally viable set — a head, an investment person, a controller — costs roughly the same at $100m as at $500m. On $100m of capital, a $1m budget means 100 bps a year and consumes a meaningful share of real return. Below roughly $100–250m an SFO loses to an MFO or a service model, not because "there isn't enough money" but because there is nothing to spread the fixed cost across.
The team: roles, size and money
The core is consistent: an executive director accountable for the office as an enterprise; a CIO; a controller running accounting and consolidation; the tax and legal contour; a family secretary and lifestyle function. Per UBS 2026, non-investment staff make up 40% of headcount.
Teams are smaller than most people imagine. KPMG 2025: 38% of offices have five employees or fewer, and 19% have twenty or more. Morgan Stanley with Botoff Consulting, in the Single Family Office Compensation Report, gives average headcount by tier: 8.4 people below $500m in assets and 28.3 above $2.5bn. Public substance thresholds sit lower still: Singapore's Section 13O regime requires only two investment professionals, at least one from outside the family.
Money. KPMG 2025 gives US base salaries: CEO $396–500k, CIO $330–500k, CFO $264–330k. Morgan Stanley and Botoff measure total compensation including bonus and long-term incentives, so medians run higher: CEO around $800k, CIO $926k, CFO $500k — and for the same CIO role the spread by office size runs from $558k to $1.51m. Bonuses reach 85% of employees, the most common band being 21–30% of salary (KPMG).
Retention is not built on bonus. A bonus closes the year; what holds a person for ten years is a share in the outcome. Morgan Stanley and Botoff: long-term incentive plans operate at 54% of all SFOs and 62% of investment-focused ones, and in 2025 co-investment (57%) overtook deferred compensation (56%) for the first time since 2015; carried interest and phantom carry were used by 30% of LTI-running offices in an earlier reading of the same series. Prevalence scales with size — 47% below $1bn against 72% above.
Hiring stays painful, and KPMG names the source: the office competes with industries where pay is transparent and structured while offering an opaque construction with no external benchmark. More than half of the offices in the Morgan Stanley sample report recruiting difficulties. A separate storyline is succession inside the office: per UBS 2026, 57% of families have a wealth succession plan but only 35% have one for the office itself. A key person walking out with the passwords, the bank relationships and the reporting logic is a scenario the industry is less prepared for than a generational handover.
What to keep in-house and what to send out
The line runs not through the complexity of a function but through whether the family loses control and context by delegating it. UBS 2026 (307 offices, average AUM $1.3bn): in-house — strategic asset allocation (86%), financial reporting (80%), portfolio risk management and next-generation preparation (75% each), succession planning (67%). Outsourced — tax planning (64%), legal (62%), cybersecurity (56%). J.P. Morgan 2026 agrees in substance: legal work (52%) and trade execution (45%) go out most often, while aggregation stays in.
Function
Where it belongs
Why
Allocation, major decisions, manager selection
In-house
This is the office's mandate
Relationships with banks, managers, registrars
In-house
Accumulated context is not transferable
Consolidation and reporting
In-house (the data) + vendor (bookkeeping)
Ownership of the number cannot be delegated; the keying can
Tax returns across jurisdictions
Outsourced
Local expertise is cheaper to buy than to hire
Custody and trade execution
Outsourced
Bank infrastructure is not reproducible at any budget
IT and cybersecurity
Outsourced, with an internal risk owner
The vendor supplies hands; accountability stays with the office
The investment function as a whole is a special case. An outsourced CIO buys discretion without building a team: on figures cited by Crain Currency, strong CIOs and CFOs command upwards of $500k plus bonus and carry, and the external model cuts that line by more than 30%. The adjacent construction is the EAM sitting on top of a custody bank. For smaller families PwC describes the intermediate forms: the embedded family office inside the operating business, and the virtual family office — a technology core plus external specialists assembled per task.
Reporting and the tech stack as the nervous system
Consolidation is not a reporting ritual but a precondition for decisions. Without one picture across all banks, entities and trusts, four things stay invisible: the true currency position, liquidity (how many years of spending are covered by assets sellable within a week), real concentration, and performance against a benchmark. Adoption is partial: per UBS 2026, 52% of offices use external financial reporting software. RBC and Campden Wealth, in the North America Family Office Report 2025 (141 respondents, average wealth $2.0bn), record a jump: 69% now have automated investment reporting, against 46% a year earlier.
The stack breaks into layers: portfolio and reporting systems (Addepar, Masttro, Landytech and dozens of competitors); an accounting layer with a multi-entity general ledger; an aggregation layer pulling from custodians; GP portals for private markets; document management. The typical mistake is universal — Excel as the system of record. A spreadsheet is an excellent calculator and a poor database: it does not version, does not reconcile, does not survive the author's departure and does not hand data to other systems.
The difficulty is that data arrives in different ways. The liquid side comes in automatically — custodian files, bank statements, APIs and SFTP. Illiquids come in by hand: portals, emails and PDFs in each manager's own format, with a quarterly lag, preliminary marks and later NAV revisions. Simple's 2025 review records the shift in priorities: integration and data quality have displaced visual reporting at the top of the list. The failure signal is simple — if a number cannot be traced back to a source transaction, and one or two people understand the logic of the final workbook, the office has no system of record.
The office's risk contour
Three items that look trivial in the budget and cost more than everything else when they go wrong.
Counterparty checks. Sanctions and reputational EDD are not just for banks: direct deals take the office into sectors and geographies outside its competence, and what needs checking is not only the numbers — litigation history, adverse media, ultimate beneficial owners, sanctions exposure. Thomson Reuters stresses that this is not a one-off exercise: files are refreshed regularly because a counterparty's circumstances change after closing too.
Payments. The same logic applies here as in private banking and in the EAM construction: whoever makes the decision does not execute the payment. The minimum is dual authorisation, value limits and confirmation of new bank details by callback to an independently sourced number. This is a direct extension of family cybersecurity: per Deloitte's Family Office Cybersecurity Report 2024 (354 SFOs, average AUM $2.0bn), 43% of offices suffered a cyberattack in the preceding 12–24 months, rising to 62% among offices above $1bn; phishing featured in 93% of those cases. There is no incident response plan at 31%, no disaster recovery plan at 50%, and no cyber policy at 63%. UBS 2026 confirms the gap from the other side: only 41% of offices have formalised cybersecurity controls.
Insurance and the archive. The standard perimeter in WTW and Marsh McLennan Agency practice: D&O and trustee liability for directors of holding and trust structures, fiduciary liability, professional liability, employment practices, crime/fidelity against internal fraud, and cyber. The claims sound surprising in a family context — improper delegation of authority, unequal treatment of family branches, deviation from the investment strategy; which is exactly why the authority perimeter belongs in the family charter and the IPS rather than in email. Corporate books, minutes, powers of attorney, UBO registers and access keys must be recoverable without any particular employee.
How to tell the office is working
Portfolio return is a poor measure of office quality: it measures the market and the managers, not the operating machine. Four things are worth measuring, and all of them can be assembled without an outside consultant.
Cost to service a unit of capital — all-in, in basis points, splitting internal costs from external providers, tracked over three years. Time to close the reporting period — working days from quarter-end to a signed pack; the most honest indicator of the state of the data. Share of assets visible in one system — the percentage of portfolio value that reaches the report without manual assembly; illiquids are usually where it sags. Response time to a family request — the median and, more tellingly, the 90th percentile.
A fifth measure is the discipline of reviewing providers, and here the industry has a blind spot: per UBS 2026 only 31% of offices run a regular review process for external providers, even though those providers absorb about a quarter of the budget.