Where to work: what each firm does and what the registers say.

Four ways to make money

Job postings say “hedge fund”, “bank” or “fintech”, but those words group very different businesses together. What best explains the daily work of a quantitative profile is not the name of the sector but where the firm's revenue comes from: from there follow the horizon of decisions, the risk tolerance, the value of a millisecond and the way of measuring whether someone does their job well. Lesson 7.5.1 describes the roles (researcher, developer, trader, risk, data and ML engineer); this one places them in the places where they are practiced.

flowchart TB
  I[Source of revenue] --> P[Other people's assets]
  I --> F[Client flow]
  I --> C[Own capital]
  I --> S[Selling services]
  P --> G[Asset managers and funds]
  F --> B[Brokers and banks]
  C --> T[Prop and market makers]
  S --> D[Fintech and data]

• Other people's assets. The firm manages other people's money and charges a percentage on it (management fee) and, sometimes, on the return (performance fee). Its revenue grows with the assets and with the ability to retain them.
• Client flow. The firm charges for executing, holding in custody, financing or distributing: per-order commissions, interest margin on cash and securities lending, retrocessions for distributing funds.
• Own capital. The firm puts its balance sheet at risk. It earns if its statistical edge exceeds the costs and the inventory risk; nobody pays it for trying.
• Selling services. Data subscriptions, software, platforms or analysis. The client pays even if they lose money in the market.

A large firm combines several of these; an investment bank has all four. But each desk or each subsidiary usually lives off one, and it is worth knowing which before the interview (lesson 7.5.3).

Seven types of firm

Two nuances the table does not capture. The first: “hedge fund” is not a European legal form. In the EU, whoever manages strategies of that kind is an alternative investment fund manager (AIFM; GFIA in Spanish) and, in Spain, it can also be an SGIIC that manages hedge-fund-type IICs (IIC de inversión libre). The second: “prop firm” designates two things that are nothing alike. One is a proprietary trading firm, with its own capital and employees. The other sells retail customers evaluations in simulated accounts and promises them capital if they pass; its model and its risks are studied in lesson 7.5.4.

Legal forms in Spain

Each type of firm has a license, and the license says what it can do. Article 128 of Spain's Law 6/2023 on Securities Markets classifies Spanish investment firms (empresas de servicios de inversión, ESI):

• Sociedades de valores (SV, securities firms). They may operate “both on behalf of others and on their own account” and provide all investment services. It is the legal form of a broker that also trades with its balance sheet or makes markets.
• Agencias de valores (AV, securities agencies). They “may only operate on behalf of others”. They receive, transmit and execute client orders, manage portfolios or advise, but they do not trade on their own account.
• Sociedades gestoras de carteras (SGC, portfolio management companies) and empresas de asesoramiento financiero (EAF, financial advisory firms). Discretionary portfolio management and advice, without being able to hold client funds or securities.
• Credit institutions. They are not ESIs, but the same article allows them to provide all investment services if their authorization covers them. That is why a bank can have a trading desk.

The quotations from Law 6/2023 are an unofficial translation. Outside that article are the SGIICs, which manage investment funds and companies (IIC, instituciones de inversión colectiva, including hedge-fund-type ones), the SGEICs, which manage closed-ended investment entities (private equity and venture capital), and the crypto-asset service providers (CASPs) authorized under MiCA, which are studied in lesson 3.7.5.

The Spanish ecosystem in the CNMV lists

The CNMV publishes one list per type of entity, an HTML page with one block per entity: name, registration number and date, and address. Lesson 4.4.1 already read the list of European investment firms operating under the freedom to provide services to check a broker; here the others are used to measure the ecosystem. The lists are in Spanish, and the code reads them as they are. The real code that downloads and counts them:

import html
import re
from urllib.request import Request, urlopen

BASE = "https://www.cnmv.es/portal/consultas/listadoentidad.aspx"
LISTAS = {"SV and AV": "?id=1&tipoent=0", "EAF": "?id=1&tipoent=2",
          "SGIIC": "?id=2&tipoent=0", "SGEIC": "?id=4&tipoent=0"}
for tipo, q in LISTAS.items():
    req = Request(BASE + q, headers={"User-Agent": "Mozilla/5.0"})
    with urlopen(req, timeout=30) as r:
        pagina = r.read().decode("utf-8", errors="replace")
    texto = html.unescape(re.sub(r"<[^>]+>", "\n", pagina))
    # "Official registration number and date", one per entity
    n = texto.count("Número y fecha de registro oficial")
    print(tipo, n)

On October 23, 2026 it printed 114 SV and AV, 92 EAF, 122 SGIIC and 165 SGEIC. The list of Spanish credit institutions uses another format (a code of the Banco de España, Spain's central bank, instead of a registration number) and had 102. The next cell processes a fictitious extract with the format of the lists (with less indentation), in Spanish like the original, to work without a network connection:

import re
import pandas as pd

LISTA = """
ALCOR MERCADOS, SOCIEDAD DE VALORES, S.A.
  Número y fecha de registro oficial: 901 - 12/03/1990
    Dirección: C/ MAYOR 1 - 28013 MADRID
BRIO PATRIMONIOS, A.V., S.L.
  Número y fecha de registro oficial: 902 - 02/06/2017
    Dirección: C/ ARAGO 2 - 08007 BARCELONA
CENIT GESTION, SGIIC, S.A.
  Número y fecha de registro oficial: 903 - 10/04/2001
    Dirección: GRAN VIA 3 - 48001 BILBAO
DUNA CAPITAL PRIVADO, SGEIC, S.A.
  Número y fecha de registro oficial: 904 - 01/09/2026
    Dirección: C/ COLON 4 - 46004 VALENCIA
EBRO ASESORES EAF, S.L.
  Número y fecha de registro oficial: 905 - 15/01/2025
    Dirección: C/ ALFONSO I 5 - 50003 ZARAGOZA
FARO QUANT A.V, S.L.
  Número y fecha de registro oficial: 906 - 24/07/2026
    Dirección: C/ AZOR 6 - 28691 VILLANUEVA
"""
patron = re.compile(
    r"^(?P<nombre>\S[^\n]*)\n"
    r"\s*Número y fecha de registro oficial: (?P<num>\d+)"
    r" - (?P<fecha>\d\d/\d\d/\d{4})\n"
    r"\s*Dirección: (?P<direccion>[^\n]*)", re.M)
ent = pd.DataFrame([m.groupdict() for m in patron.finditer(LISTA)])
ent["fecha"] = pd.to_datetime(ent["fecha"], format="%d/%m/%Y")
ent["cp"] = ent["direccion"].str.extract(r"- (\d{5})\b")[0]
ent[["nombre", "num", "fecha", "cp"]]

                                      nombre  num      fecha     cp
0  ALCOR MERCADOS, SOCIEDAD DE VALORES, S.A.  901 1990-03-12  28013
1               BRIO PATRIMONIOS, A.V., S.L.  902 2017-06-02  08007
2                 CENIT GESTION, SGIIC, S.A.  903 2001-04-10  48001
3          DUNA CAPITAL PRIVADO, SGEIC, S.A.  904 2026-09-01  46004
4                    EBRO ASESORES EAF, S.L.  905 2025-01-15  50003
5                       FARO QUANT A.V, S.L.  906 2026-07-24  28691

The columns are the name, the registration number, the date and the postal code (cp). The list mixes sociedades and agencias de valores, and the legal form of each one is only in the name, written without a fixed rule: “SOCIEDAD DE VALORES”, “S.V.”, “A.V.”, “AV” or, as in a real entity on the list, “A.V,” with the comma attached. Management companies write “SGIIC” or “S.G.I.I.C.”, and EAFs, “EAF” or “EAFI”. A classifier based on regular expressions has to accept those variants. The province comes from the first two digits of the postal code:

REGLAS = [
    ("SV", r"SOCIEDAD DE VALORES|\bS\.?V\b"),
    ("AV", r"AGENCIA DE VALORES|\bA\.?V\b"),
    ("SGIIC", r"\bS\.?G\.?I\.?I\.?C\b|GESTORA DE INSTITUCIONES"),
    ("SGEIC", r"\bS\.?G\.?E\.?I\.?C\b|ENTIDADES DE INVERSI[OÓ]N"),
    ("EAF", r"\bE\.?A\.?F\.?I?\b|ASESORAMIENTO FINANCIERO"),
]
PROVINCIAS = {"28": "Madrid", "08": "Barcelona", "48": "Bizkaia",
              "46": "Valencia", "50": "Zaragoza"}

def tipo(nombre: str) -> str:
    for etiqueta, regla in REGLAS:
        if re.search(regla, nombre.upper()):
            return etiqueta
    return "unclassified"

ent["tipo"] = ent["nombre"].map(tipo)
ent["provincia"] = ent["cp"].str[:2].map(PROVINCIAS).fillna("other")
ent[["nombre", "tipo", "provincia"]]

                                      nombre   tipo  provincia
0  ALCOR MERCADOS, SOCIEDAD DE VALORES, S.A.     SV     Madrid
1               BRIO PATRIMONIOS, A.V., S.L.     AV  Barcelona
2                 CENIT GESTION, SGIIC, S.A.  SGIIC    Bizkaia
3          DUNA CAPITAL PRIVADO, SGEIC, S.A.  SGEIC   Valencia
4                    EBRO ASESORES EAF, S.L.    EAF   Zaragoza
5                       FARO QUANT A.V, S.L.     AV     Madrid

With the real lists, that classifier assigns a type to each of the 493 entities on the four lists and splits the 114 on the first one into 40 sociedades and 74 agencias de valores: they are the two bottom bars of the header figure. The sociedades de valores, highlighted, are the only Spanish investment firms that may trade on their own account. The same aggregation gives the geography:

# (entities with postal code 28xxx, total of the list)
madrid = pd.DataFrame(
    {"en_madrid": [88, 101, 96, 53], "total": [114, 122, 165, 92]},
    index=["SV and AV", "SGIIC", "SGEIC", "EAF"])
madrid["pct"] = (100 * madrid["en_madrid"] / madrid["total"]).round(1)
madrid

           en_madrid  total   pct
SV and AV         88    114  77.2
SGIIC            101    122  82.8
SGEIC             96    165  58.2
EAF               53     92  57.6

More than three quarters of the SVs and AVs and of the SGIICs have their registered address in the province of Madrid; the private equity management companies (SGEICs) and the EAFs are somewhat more spread out, with Barcelona in second place (35 SGEICs and 14 EAFs). The registered address is the registered office: it does not say where each employee works, but it does say where the management and, almost always, the desks are.

How they make money, with CNMV data

The CNMV investment firm statistics publish every quarter the aggregate balance sheet and income statement of sociedades and agencias de valores. Table 1.4 breaks down the fees received by service, in thousands of euros accumulated since the start of the year (note 2 of the table), so the fourth quarter column is the full year. This is the 2025 extract, with the figures from the table published with data up to June 2026. The items keep the CNMV's Spanish wording; the top ones are the distribution of IIC (investment funds), order handling and execution, “other”, investment advice and portfolio management:

import io

CUADRO = """partida;miles
2.1.1. Tramitación y ejecución de órdenes;170.106
2.1.2. Colocación y aseguramiento de emisiones;10.296
2.1.3. Depósito y anotación de valores;32.196
2.1.4. Gestión de carteras;68.167
2.1.5. Asesoramiento en materia de inversión;68.920
2.1.6. Búsqueda y colocación de paquetes;1.147
2.1.7. Operaciones de crédito al mercado;0
2.1.8. Comercialización de IIC;232.984
2.1.9. Internalización sistemática de órdenes;0
2.1.10. Intermediación en instrumentos derivados;11.279
2.1.11. Elaboración de informes y análisis financieros;11.077
2.1.12. Servicios y actividades con criptoactivos;0
2.1.13. Otras;106.013
"""
com = pd.read_csv(io.StringIO(CUADRO), sep=";", thousands=".")
com[["codigo", "servicio"]] = com["partida"].str.extract(
    r"^([\d.]+?)\.? (.+)$")
total = com["miles"].sum()
print(f"Sum: {total} thousand euros; table total: 712186")
com["pct"] = (100 * com["miles"] / total).round(1)
print(com.sort_values("miles", ascending=False)[
    ["servicio", "miles", "pct"]].head(5).to_string(index=False))

Sum: 712185 thousand euros; table total: 712186
                             servicio  miles  pct
              Comercialización de IIC 232984 32.7
   Tramitación y ejecución de órdenes 170106 23.9
                                Otras 106013 14.9
Asesoramiento en materia de inversión  68920  9.7
                  Gestión de carteras  68167  9.6

The sum of the items falls one thousand euros short of the published total: each item is rounded to thousands, and a difference of one unit is rounding, not a transcription error. Checking the sum is the first thing to do when copying a table. And the result contradicts the usual image of a broker: the top item is not order execution but fund distribution, a third of the fees. Execution, brokerage in derivatives and block trade sourcing, the closest thing to trading, add up to 25.6%. For a job seeker this has a direct reading: in many Spanish investment firms the business is distributing and advising, and the quantitative work, when there is any, is in portfolios, risk and data rather than in strategies.

The income statement also says how much staff costs weigh:

pyg = pd.Series({"margen_bruto": 649_333, "personal": 309_801,
                 "generales": 166_567, "antes_impuestos": 162_612})
print((pyg / 1000).round(1).to_string())
print(f"Staff costs / gross margin: "
      f"{pyg['personal'] / pyg['margen_bruto']:.1%}")

margen_bruto       649.3
personal           309.8
generales          166.6
antes_impuestos    162.6
Staff costs / gross margin: 47.7%

The keys are the gross margin, staff costs, general expenses and profit before tax, in millions of euros. Staff costs were the sector's largest expense in 2025, almost half of the gross margin. It is an aggregate figure for 114 very different entities and says nothing about individual salaries: the statistics do not publish them and this lesson gives no salary figures without a primary source that publishes them.

Europe in the ESMA register

ESMA (the European Securities and Markets Authority) maintains a register of entities authorized under MiFID II and other rules (UCITS and alternative fund managers, trading venues, systematic internalisers). The web interface queries a Solr index that supports counts by field (facets). The query below worked on October 23, 2026; it is not a documented API:

import json
from urllib.parse import urlencode
from urllib.request import Request, urlopen

URL = ("https://registers.esma.europa.eu/solr/"
       "esma_registers_upreg/select?")
params = [("q", "type_s:parent"), ("rows", 0), ("wt", "json"),
          ("fq", "ae_status:Active"),
          ("fq", 'ae_officeType:"Head office"'),
          ("facet", "true"), ("facet.limit", 60),
          ("facet.pivot", "ae_entityTypeLabel,ae_homeMemberState")]
req = Request(URL + urlencode(params),
              headers={"User-Agent": "Mozilla/5.0"})
with urlopen(req, timeout=30) as r:
    datos = json.load(r)
pivote = datos["facet_counts"]["facet_pivot"]
print(pivote["ae_entityTypeLabel,ae_homeMemberState"][0])

The response is a tree: one node per entity type with its total and, inside, one node per home Member State. The next cell reads an extract of that response with the real values of the query (the top five States of each type and Spain):

import json

RESPUESTA = """[
 {"value": "investment firm", "count": 5319, "pivot": [
  {"value": "germany", "count": 1865}, {"value": "italy", "count": 463},
  {"value": "austria", "count": 458}, {"value": "spain", "count": 312},
  {"value": "france", "count": 268}]},
 {"value": "aifm", "count": 2882, "pivot": [
  {"value": "france", "count": 566}, {"value": "poland", "count": 384},
  {"value": "spain", "count": 337}, {"value": "luxembourg",
  "count": 273}, {"value": "hungary", "count": 137}]},
 {"value": "ucits management company", "count": 1143, "pivot": [
  {"value": "luxembourg", "count": 286}, {"value": "france",
  "count": 224}, {"value": "spain", "count": 110},
  {"value": "ireland", "count": 92}, {"value": "italy", "count": 45}]}
]"""
filas = [{"tipo": t["value"], "total": t["count"],
          "estado": e["value"], "n": e["count"]}
         for t in json.loads(RESPUESTA) for e in t["pivot"]]
esma = pd.DataFrame(filas)
esma["pct"] = (100 * esma["n"] / esma["total"]).round(1)
print(esma[esma["estado"] == "spain"].to_string(index=False))

                    tipo  total estado   n  pct
         investment firm   5319  spain 312  5.9
                    aifm   2882  spain 337 11.7
ucits management company   1143  spain 110  9.6

Spain is the fourth State by number of active investment firms and the third in alternative and UCITS fund managers. But the ESMA and CNMV counts do not count the same thing: the register's 312 Spanish “investment firms” are almost three times the 114 SVs and AVs on the national list. Classifying the 312 names with the lesson's rules yields 39 SVs, 75 AVs, 92 EAFs and 106 names without any of those abbreviations; 103 of them have “BANCO”, “BANK”, “CAJA”, “CAIXA” or “CRÉDITO” in their name. Credit institutions that provide investment services (Article 128(3) of Law 6/2023) appear in the European register under that label, and the CNMV lists them separately. Comparing countries with this count requires knowing which national legal forms each authority loads under the same label: a count without a definition is useless for comparison.

From the register to the job posting

With the map and the sources, checking a potential employer takes a few minutes and avoids surprises:

flowchart LR
  A[Job posting] --> R{In a register?}
  R -- yes --> L{Matching license?}
  R -- no --> W[Warnings]
  L -- yes --> E[Interview]
  L -- no --> P[Ask]

1. Identify the entity that is hiring. The trade name is not enough: the contract is signed with a company, and the legal notice on its website gives the company name and the registration number.
2. Look it up in the register that matches what it says it does. Fund management: SGIIC or SGEIC at the CNMV, or UCITS management company or AIFM in the ESMA register. Proprietary trading or market making: SV or credit institution. Crypto-assets: authorized CASP (lesson 3.7.5).
3. Check that the license matches the activity. An agencia de valores that advertises a proprietary trading desk, or an EAF that talks about “managing the firm's capital”, deserves a question in the interview.
4. If it does not appear in any register, look it up in the CNMV warnings about unauthorized entities. Not appearing in a register is not illegal if the activity does not require authorization (a software company, a consultancy, a trader with their own money), but then it is not a regulated financial entity, with what that implies for job stability and for what it can promise.

The step from each type of firm to a specific position, and what is asked in interviews, is in lesson 7.5.3; the training that job postings ask for and the risks of evaluation prop firms, in 7.5.4.

Exercises

1. Classify these names, in the format of the ESMA register, with the lesson's tipo function extended with a rule for credit institutions. How many fall into each group?

NOMBRES = ["ALCOR MERCADOS, SOCIEDAD DE VALORES, S.A.",
           "BRIO PATRIMONIOS, A.V., S.L.", "CENIT BANCO, S.A.",
           "CAJA RURAL DEL EBRO, S. COOP. DE CREDITO",
           "EBRO ASESORES EAF, S.L.", "GALENA INVESTMENT, S.A."]
REGLAS_ESMA = REGLAS + [
    ("credit", r"BANC|BANK|\bCAJA\b|CAIXA|CR[EÉ]DITO")]

def tipo_esma(nombre: str) -> str:
    for etiqueta, regla in REGLAS_ESMA:
        if re.search(regla, nombre.upper()):
            return etiqueta
    return "unclassified"

pd.Series([tipo_esma(n) for n in NOMBRES]).value_counts()

credit          2
SV              1
AV              1
EAF             1
unclassified    1
Name: count, dtype: int64

2. The figures in table 1.4 are accumulated since January. With the fees for order handling and execution accumulated to June, September and December 2025 (84,337, 124,996 and 170,106 thousand euros) and to March and June 2026 (51,838 and 97,771), compute those of each quarter. Which quarter had the most order fees?

acum = pd.Series([84_337, 124_996, 170_106, 51_838, 97_771],
                 index=["2025Q2", "2025Q3", "2025Q4",
                        "2026Q1", "2026Q2"])
anio = acum.index.str[:4]
trim = acum.groupby(anio).diff().fillna(acum)
trim = trim.drop("2025Q2")  # without 2025Q1 it cannot be separated
trim.astype(int)

2025Q3    40659
2025Q4    45110
2026Q1    51838
2026Q2    45933
dtype: int64

3. A job posting asks for a “quant researcher” for “a proprietary trading firm based in Madrid” and does not give the company name. List what you would check before and during the interview with the sources of this lesson.