This article is the practical "how" companion to the pillar guide. Below is the sequence we use ourselves when running a commercial sourcing campaign, expressed in the order a buyer can follow without GalimAI access. Where the GalimAI portal automates a step, that is noted explicitly.
Before you start: define the buy box
Off-market sourcing fails when the buy box is fuzzy. Before any data step, write down, in one sentence each:
- Asset type: retail parade, single-let industrial, multi-let industrial, suburban office, regional logistics, mixed-use, development site, ground-up with planning, ground rent, etc. Be specific.
- Geography: postcode districts or local authority areas. London-wide is not a geography; "M3, M4, M14, M15 in central Manchester" is.
- Lot size: price range, GIA range, or annual rent range. Anchored to what you can actually transact.
- Tenure preference: freehold, long leasehold, ground rent, or open.
- Capital constraint: cash buyer, bridge-then-refinance, term-loan, JV-style. This shapes which seller archetypes are realistic counterparties.
Everything that follows is filter-driven. A vague buy box produces a vague long list and a wasted campaign.
The six-step methodology
1Identify the candidate title set
Start with HM Land Registry's freely available data products. The Commercial and Corporate Ownership Data (CCOD) dataset lists every UK property title held by a UK company. The Overseas Companies Ownership Data (OCOD) covers titles held by non-UK entities. Both refresh monthly. Filter to your geography and run a sanity-check pass to remove obvious mismatches (a residential block held in a property company, a freehold ground rent that is not the asset you want).
In the portal: the GalimAI commercial filter applies CCOD + OCOD natively plus the additional signals below in one pass.
2Resolve the corporate owner
Take each title's registered proprietor and look it up on Companies House (free, online). Capture: company name, registered office, all current directors, all persons with significant control, the SIC codes (sector classification), the date of incorporation, and any active or recent charges. This is the metadata you will signal-stack on next.
A common pitfall: the registered office is often an accountant's address, not the operating address. Do not write the letter to the registered office without first reading the latest accounts to confirm the actual decision-maker.
3Layer the catalyst signals
For each company on the long list, score the catalyst stack:
- Is there a bridging or development-finance charge with a maturity date inside the next 12 months? (Companies House charges register, charge instrument PDF.)
- Is any director over 65 with no younger director succession? (Directors list, dates of birth.)
- Are there overdue filings (annual accounts, confirmation statement)? (Filing history.)
- Has a London Gazette notice been published against the company? (gazette.gov.uk search; check the Companies House overview for the small "Insolvency" flag.)
- Is the company a long-tenure family-owned holding (incorporated pre-2003, single asset, family-name directors)?
One signal alone is noise. Two stacked signals start to be predictive. Three or more is where motivation-to-sell becomes meaningfully above the population baseline. Read more on stacked distress signals.
4Rank, do not just filter
A flat filter ("show me companies with bridging maturity in 2026") returns a list with no ordering. A ranked list ("show me companies sorted by combined probability of disposal in the next 6 to 18 months") returns the same set with the highest-probability targets on top. For a 100-letter campaign, you want the top 100 of the ranked list, not the first 100 of the unranked one.
In the portal: the sell-timing model produces this ranking automatically across more than 1 million property owners. Read the methodology.
5Write a letter that is not interchangeable
Generic letters fail. The letters that get read and replied to do three things consistently:
- Address the actual decision-maker by name. Not "The Directors". Not the registered-office accountant.
- Reference the specific holding, not "your portfolio". One sentence that proves the writer has done the homework: the asset type, the rough location, sometimes the tenure.
- Make the implicit catalyst easy to acknowledge without naming it. A line about "if circumstances change" works; "we noticed your bridging is overdue" does not.
The 160,000-letter outcome dataset we trained the response model on shows reply rates rising sharply when these three conditions are met together. Read more on the direct-to-vendor letter.
6Time the send to the catalyst, not the calendar
The single most common mistake we see is mailing on a fixed monthly cadence regardless of the signal state of the recipient. The catalyst-aware approach mails each recipient on the recency of their freshest catalyst: a Gazette notice within the last 60 days, a charge maturity within the next 90, a recent late filing on top of a long-tenure holding.
This is also where most operators run out of bandwidth without software support, because the signal state of every name on the list changes weekly. The best time to send letters goes deeper.
What "good" looks like on the metrics
A properly run UK commercial off-market campaign should produce response rates an order of magnitude above generic property mailshots. The mailshot baseline industry-wide sits around 0.3 to 0.6 percent. Signal-driven, named-recipient, catalyst-timed campaigns we observe consistently land in mid single digits for verified positive replies, with the top deciles of the ranked list outperforming the median by a wide margin.
For a buyer, that means a campaign of a few hundred well-chosen letters delivers a handful of genuine, off-market conversations. Most buyers do not need more than that.
What this approach cannot do
Two boundaries worth being honest about.
- It does not extend to Scotland or Northern Ireland. Registers of Scotland and Land and Property Services NI operate separately from HM Land Registry; the title-to-company link does not exist in the same accessible form. GalimAI covers England and Wales only.
- It does not surface owner-occupied residential. The methodology assumes corporate ownership. Family homes held in a personal name are not on Companies House and not in the addressable population.