Why I Track Family Office Headcount, Not AUM (And What the Numbers Show)

Originally published: 2024-09 | Last verified: 2026-09-13 Structure counts are from MAS parliamentary replies and, for Hong Kong, the InvestHK announcement and a Legislative Council reply; the Singapore AUM figure is from MAS. All are linked inline. IP headcount, per-structure and compensation figures are my own estimates, assembled from recruiter commentary and Hubbis practitioner coverage, as the piece explains; an earlier note claiming verification against named recruiter reports overstated that. The IP timing and minimum-asset rules reflect MAS Circular FDD Cir 05/2026. Two of the three predictions below were withdrawn in September 2026, with the reasons given in place. The third was tied to Hong Kong FIHV approvals, which the government does not publish; it now rests on the counts that are published, and the Hong Kong headcount range is reconciled with the official office count. Family office labour market data is reported by industry recruiters with varying methodology; please confirm against multiple sources for current cycle figures.

When the wealth media reports on the Singapore family office boom, the metric that anchors the coverage is almost always AUM (Assets Under Management). Singapore’s asset management industry reached a record S$6.7 trillion of AUM in 2025, and regional coverage is framed the same way. AUM is the natural metric — it’s denominated in dollars, it’s reportable across firms, and it scales cleanly with the things wealth managers actually charge fees on.

It’s also the wrong leading indicator for understanding what the family office ecosystem is actually doing. AUM is a heavily lagging signal: it reflects past wealth creation, past inflow decisions, and past asset price movements rather than current ecosystem activity. The number that I find materially more useful for reading the ecosystem in real time is Investment Professional headcount — the count of qualified investment professionals (IPs) actively employed across the family office population in a given jurisdiction.

This piece explains why headcount is the better signal, what the headcount numbers across SG / HK / Dubai are showing as of mid-2026, and what they imply that the AUM numbers don’t.

Why headcount leads AUM

The mechanical reason is straightforward, though narrower than it first looks. Setting up a credible family office structure requires hiring IPs, and the first hire happens before the wealth structure is fully populated. A Section 13O setup in Singapore requires two IPs, and a Section 13U three. Under the rules in force from July 2023 until the July 2026 circular, the roster had to be in place from the start of the award. For SFO awards approved from 1 August 2026, a fund may apply one IP short and complete the roster by the end of its first basis period, but it must already hold at least S$20 million (13O) or S$50 million (13U) of qualifying assets when it applies. So under the new rules the minimum capital is in before the last IP, not after. For a mid-tier family close to that minimum, which is where much of the growth has been (see below), most of the capital can arrive ahead of the full roster. What still leads is the first IP, who has to be employed at application and whose search starts months before it, set against the capital above the minimum, which a fund doesn’t need at application. On my read, that capital typically follows the first hire by 4-9 months.

The same logic applies in reverse. When wealth holders decide to wind down or scale back family office activity, the IP layoffs precede the AUM withdrawal by a similar lag. Practitioners I talk to who track the SG family office ecosystem closely consistently report that they see the hiring pipeline slow down 6-12 months before the AUM growth slows. The IP labour market is the leading edge.

These lead times measure different gaps, which is why they differ across this piece. The 4-9 months is for a single structure: the first hire against the capital above the minimum. The 6-12 months is the aggregate turn: the hiring pipeline slowing against AUM growth slowing. The longer gaps later in the piece add publication delay on top. AUM totals are annual (the S$6.7 trillion above is for 2025 and comes from an August 2026 reply), and MAS has not put out a family office AUM figure even when asked in Parliament (see below), so a headcount shift can take 12-18 months or more to show up in any AUM number a reader can actually see.

The signal-to-noise reason is also worth naming. AUM moves with global asset prices, with currency fluctuations, with mark-to-market revaluations of illiquid holdings, and with reporting timing differences across firms. A 10% move in the SG family office aggregate AUM can reflect underlying inflow growth, equity market movements, USD/SGD shifts, or reporting reclassifications — and disentangling them is genuinely hard. IP headcount has none of these noise sources. A 10% move in IP headcount is a 10% move in actual hiring, and that has a clean interpretation: the ecosystem is either expanding or contracting at the operational level. The catch is measurement. No one publishes an IP headcount series, so the figures below are my estimates, with ranges 30% wide or more, and at that width only a large move can be read; a 10% change sits inside the range.

The current SG headcount picture

Pulling together recruiter commentary, Hubbis practitioner coverage, and the implied IP count from MAS-reported 13O/13U structures — none of which publishes an IP headcount directly — the Singapore family office IP population in mid-2026 sits in a range I’d estimate at 4,500-6,000 total active IPs across all SG-registered family office structures.

The range is wide because the categorisation is genuinely fuzzy. The lower bound counts only the IPs employed at MAS-approved 13O/13U structures (roughly 2,000-2,500 structures × 2-3 IPs per structure × adjustment for vacancies and IPs serving multiple structures; MAS counted more than 2,000 on tax incentives at end-2025). The upper bound includes IPs at non-13O/13U single family offices, IPs at multi-family offices serving SG-resident families, and adjacent professionals (investment analysts who don’t qualify as IPs but who serve the same function in shop). Practitioners I trust converge around 5,000-5,500 as the realistic working number.

The growth trajectory for this population over the last five years is the more revealing data. The SG family office IP population has roughly tripled from 2020 to 2026 — from an estimated 1,500-2,000 to the current 5,000-5,500 range. That is slower than the 13O/13U structure count, which roughly quintupled over the same period, from about 400 at end-2020 to over 2,000 at end-2024. An earlier version of this paragraph described the IP growth as the larger of the two, which had the comparison backwards. IPs growing more slowly than structures is the compression discussed next. Both IP endpoints are my estimates, and the 2020 pair is one estimate stated two ways rather than two figures that confirm each other: the 1,750 midpoint spread across about 400 structures works out to about 4.4 IPs each, which is the 4-5 per structure used below and in the chart.

The headcount number tells a sharper story than the AUM number. MAS said in 2022 that it had no estimates of the aggregate AUM held by SFOs, and its August 2026 reply, asked for family office AUM, gave only the industry-wide total, so any per-structure AUM figure is an inference; mine is that it has been roughly stable. Headcount per structure, in contrast, has been compressing. The 2020 average was 4-5 IPs per structure (because the smaller pre-substance-update population skewed toward larger UHNW family offices); the 2026 average is closer to 2-3 IPs per structure (because the post-substance-update growth has been weighted toward mid-tier holders meeting the minimum IP requirement).

This compression of IPs per structure is the leading-edge signal that the ecosystem is shifting toward smaller, leaner shops — exactly the mid-tier dominance pattern I described in my pillar piece on the SG family office boom.

The HK and Dubai comparison

Applying the same headcount lens to Hong Kong and Dubai produces a useful three-way comparison.

Hong Kong. Estimated family office IP population in mid-2026: 2,500-3,500. That is my estimate, and it counts investment professionals at the offices that run an in-house investment team. At 3-4 IPs each it implies roughly 625-1,170 such offices, well short of the over 3,380 single family offices that InvestHK and Deloitte count at end-2025. The two figures describe different populations: the market count takes in every single family office operating in the city, including, on my read, many where family members or outside managers make the investment calls. The same study puts the staff those offices directly employ at over 10,000 full-time professionals, not limited to investment staff, or about three per office, which fits a minority of offices holding most of the investment professionals. The HK IP number has grown notably over 2024-2025 from a 2023 trough but has not approached the SG growth rate; the published market count rose from about 2,700 at end-2023 to the 3,380-plus above.

Why the staffed HK offices run more IPs each than SG’s (closer to 3-4 on average) is my read; the government does not publish who applies for its family office tax concession. One checkable pointer: that concession, for family-owned investment holding vehicles (FIHVs), requires at least HK$240 million of assets under management, about S$39 million at September 2026 rates and nearly twice the S$20 million minimum for a 13O fund, though below 13U’s S$50 million. The staffing floor, two full-time qualified employees, is the same as 13O’s two IPs. Singapore’s 13O scheme therefore admits smaller shops than Hong Kong’s concession does. It is a pointer rather than proof, because take-up of the HK concession has been, in the government’s words, “a relatively small number” of applications. The HK headcount story is “growth, but at the upper-tier end” — consistent with the Greater China integration corridor I described in my SG / HK / Dubai triangle piece.

Dubai. Estimated family office IP population in mid-2026: 1,500-2,500, again my estimate, as is the per-structure ratio below. On my estimates the Dubai number has grown fastest in percentage terms over the last three years, off a low base. The IPs-per-structure ratio in Dubai sits higher than either SG or HK (closer to 4-5 IPs per structure) because, on my read, the Dubai population is weighted heavily toward residency-relocation UHNW with full family office build-outs rather than mid-tier compliance-floor structures. The Dubai headcount story is “small total but high-quality average” — consistent with the residency relocation play.

The three-way comparison produces a useful insight: on my IP estimates, Singapore leads the count, Dubai the average per structure, and Hong Kong the recovery slope. The count lead is the weakest of the three. My Singapore range takes in multi-family office staff and adjacent analysts, my Hong Kong range counts only offices with an in-house investment team, and the one published staff figure I have for either city, Hong Kong’s 10,000-plus, counts every role, with no Singapore figure on the same basis to set against it. As with the structure counts in my piece on the Hong Kong comeback, the like-for-like direction of the gap cannot be established from published data. The AUM-based comparison would compress all of this into a single “who’s biggest” question. The headcount-based comparison, even on estimates, preserves the distinct character of each ecosystem.

What headcount predicts that AUM doesn’t

Three predictions I have drawn from the headcount data. Two have since been withdrawn; I have left them in, with the reasons, rather than delete them.

Prediction 1 (withdrawn September 2026): SG IP wage compression in 2026-2027

The SG family office IP labour market has been a sellers’ market through 2022-2025 — IP candidates with relevant Asia experience have commanded materially escalating compensation packages, with the senior IP tier (10+ years experience) running at all-in compensation of USD 250K-400K+ for the right candidate. This has been driven by the supply-demand imbalance: the structure formation rate exceeded the IP supply growth.

In the mid-2026 version of this piece, I read the 2024-2026 hiring data as showing this was flipping: the IP supply looked on track to catch up with and exceed the structure formation rate by mid-2026, partly because the mid-tier dominance pattern had reduced IPs-per-structure and partly because dedicated IP training programs at SG private banks and accelerated returnee inflows had expanded supply. The implication I drew was IP wage compression through late 2026 and into 2027, and a different cost environment for recruitment budgets than the 2022-2025 norms suggested.

The argument was that the AUM data would not show this, because AUM lags, and that hiring-side signals were where any wage compression would show up first. That part of the method still stands; it is the hiring-side signals that now point the other way.

Update, September 2026: I no longer hold this prediction. My later read of the IP market, in the family office hiring squeeze piece, points the other way, with compensation still rising and the supply gap widening, and I would weight that later read over this one.

Prediction 2 (withdrawn September 2026): A meaningful structure count contraction in 2027

Structure formation ran hot through 2024, partly because faster MAS processing (most new applications approved within three months, as MAS describes it; covered in my piece on the three-month approvals) accelerated throughput. The question in the mid-2026 version of this piece was whether the inbound applicant pipeline could sustain that rate through 2027.

I argued that it could not. Wealth holders who had committed to family office formation but not yet completed IP recruitment were, on that reading, the leading indicator of the pipeline, and on my read of recruiter commentary that count had been compressing through Q1-Q2 2026. I took that to mean formation would moderate by late 2026 and that 2027 might see flat or modestly contracting structure counts. The advisory implication I drew was a 2027 in which new formation slows and deepening existing structures (additional fund vehicles, AUM scaling within existing wrappers) becomes the main work.

Update, September 2026: I no longer hold this prediction either, for two reasons. First, the hiring-squeeze read that I now weight over Prediction 1 expects IP demand to keep outgrowing supply through 2027-2028, driven partly by continued SFO formation. A flat or shrinking structure count in 2027 doesn’t sit with that, and I can’t lean on that piece for one conclusion while ignoring the other. Second, the indicator has changed meaning. For SFO awards approved from 1 August 2026, a fund can apply one IP short and finish the roster after approval, so a family with an incomplete roster may already hold an award rather than be waiting to form one. MAS’s latest count, more than 2,000 SFOs on tax incentives at end-2025, doesn’t settle 2027 either way. I would treat 2027 structure counts as an open question, and work on existing structures as coming on top of new setups rather than replacing them.

Prediction 3: HK structure count growth materialises through 2026-2027

The HK headcount growth pattern through 2024-2025 — a meaningful rebound from the 2023 trough, focused on my read at the upper-tier end — leads the structure formation data by the usual lag. An earlier version of this prediction named HK FIHV approvals as the check, measured against a 2023-2024 baseline. That check does not exist. FIHV figures count tax concession applications, not approvals of family offices, and the government does not publish them: in a July 2025 Legislative Council reply it said only that “a relatively small number” had been received for the 2022/23 and 2023/24 years of assessment, and that “it may not be appropriate to disclose relevant data.” The series that are published are InvestHK and Deloitte’s market count of single family offices, about 2,700 at end-2023 and over 3,380 at end-2025, and FamilyOfficeHK’s count of family offices it assisted to set up or expand in the city: 26 in 2023, 95 in 2024 and 50 in the first five months of 2025, up 19% on the same period a year earlier. Expect both to keep rising through 2026-2027. Neither breaks offices down by size, so the published counts will not show whether the growth sits at the larger structures, as my headcount read implies.

This is the structural HK comeback that the press has been heralding for two years and that has been visible in my headcount estimates for 18 months. If the market count and the FamilyOfficeHK figures keep climbing through 2026-2027, the structure-count data will have confirmed it; if they stall, my headcount read was wrong. AUM figures come later again, and only indirectly, through annual industry-wide totals and private wealth reports, which is why I put the AUM-side confirmation around 2028. Practitioners watching only AUM are reading the ecosystem two years late.

What this changes for the practitioner

Three actionable implications.

First, build IP labour market data into the regular practice review. If you advise wealth holders on family office setup, the IP availability and pricing picture in your jurisdiction is the leading indicator of how your client experience will evolve over the next 12-18 months. There is no published headcount series to pull it from: the working sources are recruiters’ salary guides for pricing, and conversations with search firms (Selby Jennings, Robert Walters, Page Personnel and the wealth-recruiting boutiques) for availability. Most advisors I talk to don’t track it systematically. They should.

Second, when reading AUM-based figures, know which ones you are reading. MAS’s annual figure is for the whole asset management industry, not family offices, and private reports such as BCG’s or McKinsey’s rest on their own estimates. Reverse-engineer the headcount changes those figures imply and compare them with the recruiting-side data. If the AUM growth implies headcount growth that the labour market data doesn’t support, the AUM growth is being driven by mark-to-market or by inflow into existing structures rather than by new structure formation. That distinction matters for the read on ecosystem health.

Third, when evaluating a specific family office setup mandate, ask the client about IP recruitment status alongside the AUM and structure questions. The IP recruitment status tells you whether the family is actually moving into operational setup or whether they’re still in the contemplation phase. AUM commitments and structure intentions can be ambiguous; an IP search firm engagement is unambiguous.

The headline AUM numbers will continue to drive the press cycle on the family office ecosystem. They will continue to be the wrong leading indicator. The headcount numbers tell a different story — about ecosystem maturity, about labour market dynamics, about the leading edge of formation and contraction — and that story is more accurate for understanding what is actually happening in the room. That is the metric I keep coming back to.