How accurate are scraped property addresses?

Portal addresses are usually marketing strings — partial, abbreviated, or deliberately vague (e.g. "a quiet street in SW18"). They cannot be reliably normalised to a real-world address without a separate paid address-resolution service.

How to Scrape UK Property Listings (And Why You Probably Shouldn t)

Q: What is the alternative to scraping property listings?

Use a licensed UK property data API like Homedata. A search returns matched listings with tenure, price and movement; an address reveal resolves any listing to a real UPRN and normalised full address; a property call adds EPC and last-sold; and a listings call returns thirty years of listing events. No IP bans, CAPTCHAs, or legal exposure.

The naive scraper — and why it usually doesn't work

Every developer trying to build a UK property tool eventually writes some version of this:

import requests
from bs4 import BeautifulSoup

url = "https://www.example-portal.co.uk/properties-for-sale/london"
headers = {"User-Agent": "Mozilla/5.0"}

resp = requests.get(url, headers=headers, timeout=10)
soup = BeautifulSoup(resp.text, "html.parser")

for card in soup.select(".property-card"):
    title = card.select_one("h2").get_text(strip=True)
    price = card.select_one(".price").get_text(strip=True)
    addr  = card.select_one("address").get_text(strip=True)
    print(price, "—", addr, "—", title)

On most UK property portals, this code will not return what you expect. Common outcomes from a single requests.get():

HTTP 403 / 429 — bot detection blocks the request before HTML is served.
An almost-empty HTML shell — listings are JavaScript-rendered, so requests never sees them. resp.text contains a loading spinner and no data.
A Cloudflare or DataDome challenge page — your scraper sees a CAPTCHA, not listings.
Geo-blocked or rate-limited — works once, fails on call ten.

Even when the basic request succeeds, the HTML structure changes regularly, the selectors break, and you're back to maintenance. This is the easy lesson. The hard one comes next.

What you actually get back

Look closely at a typical scraped listing. Here's what it really contains:

A marketing address — "Quiet residential street, SW18" or "Off Lavender Hill". Not a real, postable, normalised address.
A rounded asking price — sometimes "Offers in Excess of", sometimes "Guide Price", sometimes "POA".
Bed/bath counts — fine, but already available in any open listing feed.
Estate agent name and listing photos — both copyrighted, both off-limits for redistribution.

What's not there:

The UPRN — the only stable identifier that links a property to Land Registry, EPC, planning, council tax, or flood-risk data. Portals strip UPRNs deliberately.
The full real address with house number, flat, and full postcode. You see the marketing string, not the property.
The last sold price, EPC rating, tenure, floor area, listing history, or whether this property has been listed before under a different agent.

In other words: even if your scraper survives the WAF, the IP rotation, and the legal exposure, the data you extract is marketing copy. It's not a property record. It can't be matched, joined, deduped, or trusted.

Why even "successful" scrapers fail at scale

Problem	What it costs you
IP bans / CAPTCHAs	Need residential proxies (~£200–£2k/mo), still get blocked.
HTML structure changes	Selectors break weekly. Permanent maintenance burden.
JavaScript-rendered listings	Forces headless browsers — 50× slower, 100× more expensive.
No UPRN, no real address	Can't join to any official dataset. Marketing strings only.
No history	Re-listings, agent changes, price reductions are invisible.
Portal ToS	Every major UK portal forbids automated extraction; Homedata doesn't endorse working around that.

The data you literally cannot get without a licensed source

This is the part most scraping tutorials skip. The data that actually moves the needle in a property product — UPRN-matched listings, full normalised addresses, EPC, Land Registry sold price chains, listing-event history (Added → Reduced → Sold STC → Withdrawn), tenure, and floor area — is not exposed on portal pages. It's licensed. The portals don't sell it directly. The Land Registry sells partial pieces of it. Open-data projects don't have it.

There are essentially two ways to get it: pay a portal a substantial fee for a private feed, or use a licensed UK property data API.

The Homedata alternative — a search, a reveal, real data

Homedata is a UK property data API covering 29 million properties. One search returns matched listings with tenure, asking price and how that price has moved; an address reveal returns the real UPRN and the normalised full address for any listing you need to resolve. No scraping, no proxies, no CAPTCHAs, no legal exposure.

import requests
from uuid import uuid4

API_KEY = "your_homedata_api_key"

# Real listings from one search; the UPRN and full address are a reveal per listing
resp = requests.get(
    "https://api.homedata.co.uk/live-listings/search/",
    headers={"Authorization": f"Api-Key {API_KEY}"},
    params={"postcode": "SW18 4AA", "transaction_type": "Sale", "page_size": 50}
)

for listing in resp.json()["results"]:
    # A search result is the listing, not the property behind it: it carries the
    # street and the asking price, not the UPRN or the full postal address.
    # /listing-address/ unlocks those per listing — priced per engagement, and
    # free on a listing you have already revealed. A 403 with no_organization
    # means the key predates organization linking; ask support to relink it.
    revealed = requests.post(
        "https://api.homedata.co.uk/listing-address/",
        headers={
            "Authorization": f"Api-Key {API_KEY}",
            # Required: without it the reveal returns 400 missing_idempotency_key.
            # Reusing a key makes a retry free rather than charging twice.
            "Idempotency-Key": str(uuid4()),
        },
        json={"listing_id": listing["id"]},
    ).json()

    print(revealed["uprn"], revealed["full_address"], listing["latest_price"])

Sample response (a single listing, abridged):

{
  "count": 128,
  "page": 1,
  "page_size": 50,
  "total_pages": 3,
  "results": [
    {
      "id": "c61d5383-1214-4e6f-b63c-1153863bc50f",
      "street": "Lavender Hill",
      "postcode": "SW18 4AA",
      "transaction_type": "Sale",
      "latest_status": "For sale",
      "latest_price": 875000,
      "bedrooms": 3,
      "bathrooms": 1,
      "property_type": "Terraced",
      "ownership": "Freehold",
      "is_reduced": true,
      "times_reduced": 2,
      "days_on_market": 41,
      "added_date": "2026-04-22"
    }
  ]
}

What Homedata gives you that no scraper can

Real UPRNs on reveal — the foundational identifier for property datasets in Great Britain. Not stripped, not derived, not guessed.
Normalised addresses on reveal — house number, flat, full postcode, BS7666-conformant. Joinable to any OS, ONS, or government dataset.
Confirmed matches — every listing is reconciled against Land Registry, EPC, and our property characteristics graph. You're not guessing.
Listing movement — is_reduced, times_reduced, days_on_market and added_date on every search result. Portals show you "today"; we show you the movement.
Tenure and price on every search result — with the UPRN and full postal address a single reveal away.
No IP bans, no proxy budget, no CAPTCHA solving, no legal exposure.

Our recommendation

For any project where the data is going to be used commercially, shown to customers, sold to clients, or put in front of an investor — do not scrape. Use Homedata, or another professional UK property data provider. Pick the one that fits your stack and budget. Any of them will cost you less than a single legal letter, and you'll get genuinely matchable data instead of marketing strings.

Scraping a few public pages to learn HTTP and HTML for a personal project is a different conversation — but at the point where the output leaves your laptop, you should be on a licensed feed.

Try it

Pay as you go — 100 tokens = £1, no subscription needed. Most hobby projects fit comfortably under it.

FAQ

Does Homedata recommend scraping property portals?

No. Homedata does not endorse scraping. The output is unstable, the data is incomplete (no UPRN, no real address, no joined records), and licensed providers exist for a fraction of what scraping infrastructure costs to build and maintain.

Can I get UPRNs by scraping property listings?

No. Portals deliberately strip UPRNs from public listings. Without a UPRN you cannot reliably match a listing to Land Registry, EPC, planning, or council tax records. UPRN-resolved listings are only available through licensed providers.

What's the difference between a marketing address and a real address?

A marketing address is a string an estate agent writes — "Lovely Victorian terrace, SW18", "Off Northcote Road", "POA". A real address is the BS7666-normalised representation with house number, building name, postcode, and a stable UPRN. You can join real addresses to government data; you cannot join marketing strings.

What is the alternative to scraping property listings?

Use a licensed UK property data provider. Homedata gives you UPRN-matched listings, real addresses, EPC, sold price history, and full listing event history through a handful of REST calls: a search returns matched listings with tenure, price and movement; an address reveal resolves any of them to a real UPRN and normalised address; /property/{uprn}/base/ adds EPC and last-sold; and /property_listings/ returns thirty years of listing events. Several other professional providers cover similar ground — pick whichever fits your stack and budget, but don't scrape.

How to Scrape UK Property Listings (And Why You Probably Shouldn't)