HNW UK — Find British Millionaires by Name, Every Morning

Built for Haris King (SSM) by Saad Belcaid.

This scraper finds UK millionaires. By name. With the company they own. With a one-line reason to call them today.

---

The dumb-simple version

Every morning, the UK Government publishes the financial accounts of thousands of British companies — for free. Inside those filings: profit, turnover, who owns the shares.

This scraper reads all of it. Then it answers one question: "Whose company made enough profit last year that this person is now rich?"

It outputs a list. One row per rich person. Their name. Their company. Roughly how rich. And one line that says why now.

Three things you should know:

1. Same profit, different wealth. A £2M-profit software business makes its 75% owner ~£14M rich. A £2M-profit restaurant makes them ~£5M. Identical profit. The difference is what the market pays per £1 of profit in that industry — 8× for software, 3.5× for restaurants. Industry is the dial.
2. The signal is the filing. When a company files fresh accounts, that owner's wealth just got re-papered today. That's why the row exists — not last year, not last month, today.
3. Ceased = freshly liquid. When someone was a major owner but no longer is, they sold. They got cash. They're in money-decision mode. The scraper surfaces them too — you'll find these in the column psc_status = ceased. Most lead-gen actors throw these away. We don't.

UK only. Free data only. No browsers, no proxies, no AI in the hot path.

---

Max-value playbook (do this in order)

Day 1 — first run, get volume

companiesHouseApiKey: <your free CH key>
backfillDays: 14            # process the last 2 weeks of filings in one run
minPersonalWealth: 5000000  # default £5M floor
# everything else: defaults

Hit run. Wait ~15 min. Expect 200–500 rows in your dataset. That's your first batch.

Day 1, hour 2 — triage by signal type

Open the dataset. Sort or filter by psc_status. Three lists, in this order:

List	Filter	Why first
Hot — fresh exits	psc_status = ceased AND ceased_on in last 12 months	They sold, they got paid, they're picking advisors right now. Call this week.
Warm — top wealth band, decision-makers	estimated_personal_wealth_band ∈ {£25M+, £50M+, £100M+} AND ownership_band = 75-100%	Top of the pyramid. Sole/majority owners are the decision-maker, not a junior partner. Call within the month.
Steady — your industry, sole/majority	targetSicPrefixes = your sector AND ownership_band ∈ {50-75%, 75-100%}	Bread-and-butter. Sequence them across your normal cadence.

Day 2 onwards — daily drip

companiesHouseApiKey: <your key>
backfillDays: 0    # today only
# everything else: same defaults

Set up a Schedule with cron 0 8 * * 2-6 (8am UTC, Tue–Sat). The actor's persistent dedup means only NEW people surface every morning — typically 10–40 fresh leads per day. Steady drip into your CRM with zero overlap with yesterday's batch.

---

Filter recipes (copy-paste)

"I want freshly-liquid Brits this month"

pscStatus: ceased
ukResidentOnly: true
backfillDays: 30
minPersonalWealth: 5000000

"I want central London £10M+ founders"

wealthBands: ["£10M+", "£25M+", "£50M+", "£100M+"]
ownershipBands: ["75-100%", "50-75%"]
regionPostcodePrefixes: ["SW", "W1", "EC", "WC", "N1", "NW1"]

"I want UK SaaS / IT founders"

targetSicPrefixes: ["62", "63"]
ownershipBands: ["75-100%"]
minPersonalWealth: 5000000

"I want pharma / biotech / healthcare sellers (recent exits)"

targetSicPrefixes: ["21", "72", "86", "87"]
pscStatus: ceased
backfillDays: 60

"I only care about ultra-HNW (£25M+)"

minPersonalWealth: 25000000
ownershipBands: ["75-100%", "50-75%"]

"I want Manchester / NW England HNW"

regionPostcodePrefixes: ["M", "BL", "OL", "SK", "WN"]

"I want Scottish HNW"

regionPostcodePrefixes: ["EH", "G", "AB", "DD", "PA", "FK", "KY", "ML"]

---

Why "ceased = gold" deserves its own paragraph

Most actors filter ceased PSCs as "old data." They drop them. We don't.

A ceased PSC is someone who held significant control of a profitable UK business and no longer does. There's basically one reason that happens: they sold. And in the 6–18 months after a sale, that person is doing every one of these things:

• Picking a new wealth manager
• Buying property (sometimes multiple)
• Setting up a family office or trust structure
• Looking at their next venture or angel deals
• Giving to charity strategically (tax-driven)
• Restructuring inheritance plans

If you sell any of those services — wealth management, real estate, tax/legal, M&A advisory, charity advisory, business consulting, executive coaching, family office services — ceased PSCs from the last 12–18 months are the warmest lead category in your dataset. Not just warm — time-sensitive. Their decisions get made, they sign with someone, and the window closes.

Use pscStatus: ceased + filter the dataset by ceased_on year ≥ (this year − 1). That's your urgent-call list.

---

Output — one row per HNW individual

Every row has these fields. The ones in bold are the ones you'll use most.

Field	Example
person_name	John Smith
psc_status	active (still owns) or ceased (sold — call urgently)
psc_relationship	direct (PSC of this co) or via_chain (PSC of a parent holdco)
ceased_on	2024-07-06 (date they exited, if ceased)
nationality	British
country_of_residence	United Kingdom
ownership_band	75-100%
control_summary	"ownership of shares 75 to 100 percent, voting rights 75 to 100 percent"
holding_chain	["Acme Holdings", "Acme Topco"] (set when discovered through a parent chain)
company_name	Acme Software Ltd
company_number	12345678
registered_address	123 High Street, Manchester, M1 1AA
locality	Manchester
postal_code	M1 1AA
country	England
sic_codes	["62012"]
sic_descriptions	["Business and domestic software development"]
industry_multiple_applied	8.0
company_ebitda	4,200,000
company_turnover	18,500,000
company_employees	45
estimated_personal_wealth	29,400,000
estimated_personal_wealth_band	£25M+
signal	"Majority/sole owner of Acme Software Ltd — business and domestic software development, £4.2M EBITDA, fresh filing 2026-04-29"
filing_date	2026-04-29

direct vs via_chain

When you see the same person in two rows for two related companies (e.g. an operating co AND its holding co), one row will be direct and the other via_chain. The direct row is the canonical one — it's where the person is actually listed as PSC. The via_chain row was found by tracing through ownership.

Keep both for full coverage; or filter psc_relationship = direct to de-double-count.

---

How we estimate wealth (in plain English)

One rule. Three inputs. No AI, no magic.

wealth = profit × industry multiple × % they own

Translated:

"How much profit did the business make × how much someone would pay to buy a business like that × what slice of it this person owns."

That's the back-of-envelope every UK M&A advisor scribbles before they pick up the phone. It's not a valuation. It's a triage signal — good enough to know who's worth calling.

What's a "multiple"? (the SIC story)

When someone buys a private business, they pay a multiple of its yearly profit. Software gets 8×. Restaurants get 3.5×. Why? Because software profit is recurring and scalable. Restaurant profit shows up if you're standing in the kitchen at 11pm. The market knows the difference and pays accordingly.

To pick the right multiple per company, we use the company's SIC code — the UK Government's industry tag, declared on every annual filing (5 digits, e.g. 62012 = software, 56101 = restaurants). The scraper matches the SIC code against an internal table of ~85 industry prefixes and uses the most specific match.

Same profit, very different millionaire

Owner	Industry	Profit	Multiple	Stake	Estimated wealth
Software founder	62012 (software)	£2M	8×	75%	£14M → £10M+ band
Restaurateur	56101 (restaurant)	£2M	3.5×	75%	£5.25M → £5M+ band
Restaurateur (50% partner)	56101 (restaurant)	£2M	3.5×	37.5%	£2.6M → £1M+ band, dropped

Identical revenue. Identical profit. The difference is which industry they're in and how much they own.

Industry multiples — selected examples

(Full coverage of ~85 SIC prefixes lives in src/wealth.py. Lookup matches the LONGEST prefix, so 62012 software gets 8× even though 62 IT services is 6.5×.)

Sector	Multiple	Why
Biotech / pharma R&D (7211)	9.0×	IP-heavy, clinical assets, pharma buyer demand
Fund management (6630)	8.5×	Recurring AUM fees, sticky
Software development (6201)	8.0×	SaaS scaling economics
Insurance brokers (6622)	8.0×	Subscription-like renewals
Pharma manufacturing (21)	8.0×	Regulated, branded
Data / cloud / hosting (6311)	7.5×	Cloud-services premium
Specialist medical / dental (8622, 8623)	7.5×	Private healthcare consolidation
Hospitals (861)	7.0×	Recurring care revenue
Insurance (65)	7.5×	High recurrence
Veterinary (75)	7.0×	Active consolidation play
Medical instruments (3250)	7.0×	High-margin specialty mfg
Pre-primary education (8510)	7.0×	Nursery roll-up trend
IT consultancy (6202)	6.5×	Tech services premium
Banking (641)	5.5×	Regulated, capital-heavy
Mgmt consultancy (7022)	6.0×	Repeat-engagement
Real-estate letting (6820)	5.5×	Rental recurrence
Wholesale (46)	4.5×	B2B recurrence
Manufacturing (general) (28, 29)	5.5×	Capital-heavy operations
Construction (41–43)	4.0–4.5×	Project-based, cyclical
Retail (47)	4.0×	Margin pressure
Restaurants (56)	4.0×	Fixed-cost squeeze
Pubs (5630)	3.5×	Toughest
Default fallback	4.0×	Used when SIC isn't in the table

These are conservative. Real M&A deals trade above and below these depending on growth, recurrence, scale, customer concentration. We tune low on purpose — the doctrine is under-estimate wealth, never over-estimate. A £14M owner-operator on this scoring is, in reality, almost always richer than that.

Ownership-band midpoints (from Companies House PSC API)

Band	What we use
75–100%	0.875
50–75%	0.625
25–50%	0.375

Companies House only reports owners who hold at least 25% of the shares or voting rights. Anyone who shows up holds at least a quarter of the company.

Wealth bands the scraper emits

<£1M, £1M+, £5M+, £10M+, £25M+, £50M+, £100M+

The default floor is £5M. Raise it to £10M+ if you only want serious money. Drop it to £1M+ if you want a wider net of owner-operators. Or use wealthBands to keep specific bands only.

---

Input options

Field	Default	Description
companiesHouseApiKey	(required)	Free REST key from CH developer portal
Wealth filters
minPersonalWealth	5000000	Wealth floor in £
wealthBands	[]	Restrict to bands. e.g. ["£25M+","£50M+","£100M+"]
ebitdaPreThreshold	0	Auto-derived from minPersonalWealth when 0. Override only for tuning.
Person filters
ownershipBands	[]	e.g. ["75-100%"] for sole/majority owners only
pscStatus	all	active / ceased / all. Use ceased for fresh-exit lists.
ukResidentOnly	false	Drop non-UK residents
Geographic filter
regionPostcodePrefixes	[]	e.g. ["SW","W1","EC"] for central London
Industry filters
targetSicCodes	[]	Exact 5-digit SIC matches
targetSicPrefixes	[]	SIC prefixes, e.g. ["62"] for IT
Date range
backfillDays	0	Process last N days in addition to today (max 60)
bulkFileDate	(today)	Override the end date YYYY-MM-DD
Advanced
maxCorporatePscHops	3	How deep to trace holding-co ownership chains
skipWeekendCheck	false	Run anyway on Sun/Mon (no-op since CH won't have published)

---

Use cases by industry

Wealth-management & private-banking outreach

minPersonalWealth: 10000000 + pscStatus: ceased for the past 12 months. These are the people picking a new advisor right now.

Estate planning / tax / legal

minPersonalWealth: 25000000 + ownershipBands: ["75-100%"]. Patient outreach, but high-LTV.

Family offices

minPersonalWealth: 50000000 + pscStatus: all. UK ultra-HNW universe is ~12,000 individuals; this surfaces the ones whose wealth is fresh-papered today or who just exited.

Real estate / property advisory

pscStatus: ceased + regionPostcodePrefixes matching where you sell. Recent exits buy second/third homes; they self-identify by location.

Connector OS Station

Pipe the dataset into Station as the demand side. The signal field plugs straight into the I Layer for evaluation against your supply network (advisors, brokers, vendors).

Flow: scrape → dataset → paste dataset ID into Station → match against your supply → scored introductions.

---

How it works under the hood

• Parses XBRL filings using the UK Government's own stream-read-xbrl
• EBITDA = operating profit + depreciation (conservative floor estimate)
• Rate-limited to stay within Companies House's 600 req/5min limit
• Normalizes UK CH company numbers (zero-pad, strip prefixes) so malformed bulk-feed values don't lose leads
• Resolves corporate PSC chains by registration_number, falling back to exact-name search when CH didn't store the number — so PE/holdco structures still surface real humans
• Tags each row direct vs via_chain so operators can de-double-count when an operating co AND its holdco both file separately
• Surfaces both active and ceased PSCs — the ceased ones are the highest-value HNW signal in the dataset
• Strips honorifics (Mr/Mrs/Dr/Sir) before deduping so the same person doesn't appear under multiple titles
• Stores (person + company) keys in a persistent KV store so duplicates never reappear day-over-day

---

Costs

• Companies House API: free (600 req / 5 min)
• Apify compute: ~10–15 min per run × 1024 MB ≈ trivial CU on the standard plan
• Multi-day backfill (backfillDays > 0): N× the per-day cost

---

Built by Saad Belcaid for Haris King's SSM workflow. Data sourced from Companies House (public record). Wealth estimates are heuristic, not valuations.