Six stages, and five of them can say no.
Most sourcing is a filter over listing text. Ours is a filter over the registers, and the difference shows up as the properties that never reach you.
01
Channel
We work unsold auction lots available afterwards by private treaty, receivership and probate stock, insolvency notices, and independent agents who list days before the portals reach them. Being early is most of the advantage.
02
Register checks
Flood zone, conservation area, Article 4 direction, listed status, licensing, planning history, tenure and title. Checked against the national registers before a property is put to you — never after, and never as a warning attached to something we sent anyway.
03
Condition read
The floor plan and photographs are read as a building: room dimensions off the plan, structural state off the pictures. It is the difference between a works budget and a works guess.
04
Appraisal
Purchase at what comparable lots actually fetch rather than what the guide implies. Works from a costed schedule. Finance, holding costs, both ownership structures, stamp duty and the VAT position. A pessimistic case alongside the base case.
05
Ranked against its own market
A return is judged against comparable properties on the same route in the same local rental market. Income deals and conversion deals are not comparable, and one blended hurdle rate flatters whichever is already priced.
06
A person decides
Nothing reaches you automatically. The filter is the plainest one there is: would we buy this ourselves, with our own money, at this price?
Measured against our own stock, with the sample stated.
We test our assumptions against real listings, real sales and real register results rather than industry rules of thumb. Three findings shape almost everything we do.
18 of 25
known deal breakers appeared in none of 950 real auction listings
Not knotweed, not subsidence, not a flying freehold, not a regulated tenancy, not flood risk. Meanwhile the registers put 12% of the lots we checked in a flood zone and 22.3% in a conservation area, against 1.8% of listings that mentioned one. A listing discloses what attracts a bidder and omits what deters one.
1.23×
the median hammer price against guide, in the £40k–£150k bands
Guide plus ten per cent is a reserve, not a price. Measured across 237 sold buildings publishing both figures. Cheap lots run hot and expensive lots hug the guide, so an appraisal built on the guide is wrong before it starts.
32.4%
of Class MA commercial-to-residential applications are refused
Class Q is refused 37.8% of the time and Permission in Principle 45%. Permitted development is not permission, which is why every conversion we price carries a case where consent does not arrive.
The refusals are the product.
A desk that answers everything is a desk that guesses. These are the places we deliberately stop.
We will not price a room we cannot measure
A room under 10.22 square metres must be let as a single, and the room rate we hold is a double rate. Where a property carries undersized rooms, we price those rooms at nothing rather than guess.
We will not call a thin register a clear result
Article 4 coverage is patchy nationally. Outside proven coverage the honest answer is unknown, and unknown is never treated as a pass.
We will not put a language model near a number
No AI anywhere in a scoring, gating or financial calculation. Same inputs, same output, every time — otherwise a figure cannot be defended to you.
We will not source three popular strategies
Supported living, ground rents, and rent-to-rent or lease options. Whatever the headline return, they are not defensible to package for someone else.