Every price argument in a property file rests on a table of comparable evidence, and that table is the one document in the pack a reader can check without trusting its author. That is the whole point of it. A figure produced by an adjustment nobody can see is an opinion with a citation attached; a table in which every line carries its date, its area, its basis of measurement, its adjustments and its source can be re-worked by the other side’s adviser in an afternoon and will either survive on its own terms or fail on them.
Eight fields are the minimum for that to happen. Each is present because its absence breaks a specific thing a reader needs to do, and each can invalidate a line on its own: a comparable with no as-of date, a size measured on a basis nobody can identify, a price that turns out to be a guide rather than a consideration. The table below names the field, the reason it is there and the failure it prevents. It is deliberately short. A table with forty columns is usually a download, not evidence.
Table 1. The eight fields of a usable comparable, the reason each is in the table, and what it can invalidate when it is missing or wrong. Source: constructed for this piece; the third column is the diagnostic.
| Field | Why it is in the table | What it can invalidate |
|---|---|---|
| Date, with the character of the date — contract, completion or asking | Fixes the moment the price belongs to, and distinguishes an agreed price from a completed one | A price from eighteen months ago is history presented as today; a contract price abandoned before completion is a price nobody paid |
| Address or reference, and the unit it identifies | Establishes that the comparable is a different asset from the subject, and which asset it is | Two files citing “the same building” can hold two different units at two different rents, and the table then double-counts one transaction |
| Property type and segment | Fixes what market the line is evidence for: office, retail, industrial, residential; prime or secondary | A retail unit in an office set is a different asset class with a different rate, and it moves the mean as if it belonged |
| Size, with the date it was measured | Makes the rate computable at all, because price divides by area | An area with no date is an area from a plan nobody can identify, and a floor since extended or reinstated is a different area |
| Basis of measurement — net internal, gross internal, gross external, stated in the same words on every line | Makes the rate reproducible against the subject and against every other line | A rate per gross metre set beside a rate per net metre understates the subject’s area by five to fifteen per cent, and the error runs one way through the table |
| Condition, with the basis on which it was assessed | Distinguishes the asset that sold from the asset being valued | A refurbished comparable in a set of unrefurbished ones lifts the range by the cost of a refurbishment nobody has counted |
| The price, and what it is — consideration, asking price, guide price, price on application | States what claim the number supports | An asking price treated as a transaction is the single commonest failure in the business, and one question about a sold price ends it |
| Source, with the retrieval date | Makes provenance checkable and the line arguable | A figure with no source cannot be tested, and a figure from a portal is a claim by a seller rather than a record of a deal |
Date
The date does two jobs and the table needs both. The first is chronological: a price agreed in one market and a price agreed in another are not the same evidence, and the only way a reader can tell which is which is to see when each was struck. The second is character. An asking price has no transaction date, only the date it was listed; a price agreed at contract is not a price paid until completion, and a file that records the first as if it were the second has recorded an intention as a fact.
A comparable older than the question needs is labelled with its date and treated as history. That is a small discipline with a large effect, because thin markets are exactly where the temptation lives: the last fitting transaction is eighteen months old, the market has moved, and the file uses the figure in the present tense. The correction is not to inflate it for time by a guess. It is to show the date, state the adjustment the office’s own index supports, and let the reader see how much of the conclusion rests on a number that old.
Where the file supplies no index, the date adjustment is zero and the table says so. A zero is a statement — nothing here is adjusted for time — and it is a defensible one. A two per cent uplift recalled from memory is not.
Address or reference
The field exists to prove that the comparable is a different asset from the subject, and to prove which asset it is. Both halves matter. A table whose addresses are vague cannot be checked at all; a table whose addresses are precise but whose unit is unstated can contain the same transaction twice under two descriptions, once as “12th floor, tower two” and once as the building’s headline sale, and no arithmetic will reveal it.
The discipline is one line, one asset, one identifier: the address or the file reference the office holds it under, the floor or unit where the asset is part of a building, and the tenure — freehold, leasehold with a term stated, or a share of a company that holds it. Tenure is not decoration. A leasehold flat with eleven years left and the same floor area as a freehold one is a different asset at a different rate, and a set that mixes them without saying so has mixed two markets.
Type and segment
The segment is the market the line is evidence for, and it belongs in the table rather than in the analyst’s head. Office, retail, industrial, logistics, residential, mixed use: each has its own rate structure, and each divides again into prime and secondary, and again by lot size, because a buyer of a small unit and a buyer of a building are different buyers with different capital.
The test is whether the comparable would appear in the subject’s own buyer’s search. If it would not — a retail parade in an office set, a logistics shed in a workshop set — it is not evidence, whatever its rate per square metre. Where a line is kept as a cross-check rather than as evidence, it is labelled as a cross-check and weighted below the set, and the table shows the weight rather than leaving it to the reader.
Size
The size field carries two things: the number and the date it was measured. The number is there because every rate in the table is a price divided by an area, and a rate without an area is a figure with no denominator. The date is there because areas change: a floor extended, a plant room re-designated, a measured survey superseding a plan. An area with no date is an area from a document nobody can find, and if the comparable’s area has since been altered, the rate belongs to an asset that no longer exists in that form.
Basis of measurement
This is the field that most often decides whether the table means anything, and the one most often left to whatever the listing said.
Three bases circulate and they are not interchangeable: net internal area, which is the usable space inside the walls; gross internal area, which adds the shared structure within the building’s envelope; and gross external area, which measures to the outside face. The gaps between them are not small — five to fifteen per cent between net internal and gross external is ordinary, and in some buildings more — and a rate per square metre computed on one basis set beside a rate computed on another is a comparison between two different measurements wearing the same units.
The arithmetic is brutal and it runs in one direction. Take one line from the set in the next section but measure it the other way. Comparable E sold for 2,379,000 and the office’s record gives 610 square metres. On that basis the rate is 3,900.00 a square metre. If the 610 is gross external and the net internal figure is 585, the rate per net square metre is 4,066.67 — 4.3 per cent higher, on the same price — and the line would sit 5.4 per cent below the next line in the table rather than the 9.3 per cent below it that the uncorrected figure shows. Nothing about the transaction changed. The basis did.
So the basis is named on every line, in the same words, and the subject is measured on the same basis before anything is divided. Where a basis is unknown, the line says unknown and the rate is not computed, because a rate without a basis is a number that will be quietly compared with one that has a basis, and that comparison is where the error enters the pack.
Condition
Condition is the field analysts most want to leave qualitative, and the table cannot afford it. What is being recorded is not a grade but the difference between the asset that sold and the asset being valued: newly refurbished, refurbished within five years, original specification, shell and core, or requiring works, with the basis stated — who inspected, when, against what standard.
A three per cent adjustment for condition is a claim that a refurbishment is worth three per cent of the rate, and that claim is arguable only if the table says what the difference consists of. Where the comparable is superior and is adjusted down, the reader can test the number against the cost of the works if the file has one. Where the file has none, the adjustment is labelled as the office’s convention, and the honest alternative — drop the comparable and say why — is always available.
The price, and what it is
Four kinds of number arrive in this field and they are not ranked together. A consideration is the price in a completed transaction. A contract price is agreed and may yet fall away. An asking price is a claim by a seller, and a guide price is a claim by a seller’s agent designed to attract interest rather than to state a figure. A price on application is the absence of a number.
Only the first is a measurement of what a buyer paid. The second is evidence of what one buyer was willing to agree at a moment. The third and fourth are cross-checks, useful for testing whether the set’s range is plausible and useless as its foundation, and they carry the weight the table gives them — which, in the worked example below, is 0.4 against 1.0.
The field also has to say whether the price is the whole consideration. Where the sale included stock, plant, a rent-free period, a reverse premium or a leaseback, the headline is not the price of the property, and the table states what was included and what was done about it. A comparison of two headline figures where one carried a rent-free period and the other did not is a comparison of two different deals.
Source, and the date it was read
Provenance is the last field and the one that makes the table a document rather than a set of recollections. A source is specific enough to be checked: the registry record, the office’s own transaction file, the counterparty’s confirmation, the agent’s particulars with their date, the portal listing with its reference and the date retrieved. “The market” is not a source, and neither is a colleague’s recollection, though a colleague’s recollection that leads to a registry entry is fine once the entry is cited.
Two dates belong here and only one of them is the transaction date: the date the figure was read, and the as-of date of the underlying record. A file that reads a six-month-old portal listing today has a retrieval date of today and an as-of date six months ago, and only the second is evidence about the market.
The table, worked
The set below is constructed so that the arithmetic can be followed. Five transactions, one illustrative subject of 420 square metres, and one asking price logged and excluded. Each adjustment moves the comparable towards the subject: a positive figure means the comparable is inferior to the subject on that point and its rate is adjusted up, a negative figure means it is superior and the rate comes down.
Table 2. A constructed comparable set of five transactions and one excluded asking price, each reconciled from price and area to an adjusted rate. Amounts in constructed currency units; net adjustments are the sum of the lines above them. Source: constructed for this piece; no address, transaction or party is real.
| Comparable | Date | Type and segment | Area (m²) | Basis | Condition | Price | Rate /m² | Adjustments | Net | Adjusted rate |
|---|---|---|---|---|---|---|---|---|---|---|
| A | 90 days ago | Office, secondary | 380 | Net internal | Refurbished | 1,748,000 | 4,600.00 | date +2.0, size −1.0 | +1.0% | 4,646.00 |
| B | 90 days ago | Office, secondary | 520 | Net internal | Refurbished, superior | 2,236,000 | 4,300.00 | date +2.0, size +2.0, condition −3.0 | +1.0% | 4,343.00 |
| C | 60 days ago | Office, secondary | 400 | Net internal | Refurbished | 1,860,000 | 4,650.00 | date +2.0, size −0.5, condition +1.0, parking −0.5 | +2.0% | 4,743.00 |
| D | 30 days ago | Office, secondary | 450 | Net internal | Original, inferior | 1,935,000 | 4,300.00 | condition +2.0, floor −1.0 | +1.0% | 4,343.00 |
| E | 60 days ago | Office, secondary | 610 | Net internal | Original, inferior | 2,379,000 | 3,900.00 | date +2.0, size +6.0, condition −1.0 | +7.0% | 4,173.00 |
| F | asking, today | Office, secondary | 570 | Net internal | Refurbished | 2,394,000 asking | 4,200.00 | date +2.0 | +2.0% | 4,284.00 |
| Set | median 4,343.00 |
Read line E first, because it carries two lessons. Its headline rate, 3,900.00, is the lowest in the table by a margin that invites a story about a forced seller, and the margin is a measurement rather than a market: 610 square metres is a large lot, and the six per cent size adjustment accounts for part of the gap, while the two per cent time adjustment and the one per cent deduction for condition account for the rest. A reader who does not see those three lines sees a rate nine per cent below the set and cannot tell whether it is a signal or an error. Then its adjustments lift it to 4,173.00 — the direction of every one of them moving the comparable towards the subject, which is what the column is for.
Line F is the discipline of the exclusion log. It is the same segment, the same area band and the nearest date in the table, and it is not evidence of a price because nobody paid it. It is retained at a weight of 0.4 and excluded from the range.
Table 3. The adjusted set carried through to the subject’s 420 square metres at the low, median, mean and high rates. Constructed for this piece. Source: Table 2.
| Basis | Rate /m² | Subject area (m²) | Figure |
|---|---|---|---|
| Low — comparable E | 4,173.00 | 420 | 1,752,660 |
| Median of five | 4,343.00 | 420 | 1,824,060 |
| Mean of five | 4,449.60 | 420 | 1,868,832 |
| High — comparable C | 4,743.00 | 420 | 1,992,060 |
Three observations follow from the two tables, and each is a reading of the arithmetic rather than an opinion about the property.
The first is the compression. Unadjusted, the five rates run from 3,900.00 to 4,650.00, a spread of 19.2 per cent. Adjusted, they run from 4,173.00 to 4,743.00, a spread of 13.7 per cent. That is what adjustments are for: they are the work that turns a set of different properties into a set of statements about one property. A table whose adjustments widen the spread instead is doing the opposite of its job, and the usual cause is that the analyst is carrying differences the comparables do not actually exhibit.
The second is the shape of the range. Five lines support a range of 570.00 a square metre and a figure between 1,752,660 and 1,992,060 — 239,400 between the ends, which is 13.7 per cent of the low figure. That spread is the table’s output. A file that publishes 1,824,060 from the median and calls the set supportive is publishing one point from a range whose width is a fifth of the number, and the reader who needs the number for a decision needs the width as much as the point.
The third is that the median and the mean disagree, and the disagreement is information. The mean, 4,449.60, sits 106.60 above the median because the two highest lines pull it up and the two identical lines at 4,343.00 hold the middle. On five lines the median is the more robust statistic and the mean is the more informative one, and the honest course is to publish the range and say which statistic the central figure came from.
What the table cannot do
The table’s first limitation is structural and no amount of care removes it: an adjustment is a convention, not a measurement. There is no experiment that establishes that a refurbishment is worth three per cent of the rate, or that a smaller unit commands one per cent more, or that a floor two levels higher is worth one per cent. Those numbers come from the office’s experience and the market’s habits, they differ between offices and between segments, and two competent analysts working the same set will produce different ranges. What the table can do is show every adjustment, its direction, its size and its reason, so that the reader can substitute their own convention and see what moves. What it cannot do is claim the precision its two decimal places suggest.
The second limitation is the one that matters in this family of work. A comparable table establishes what similar assets fetched. It does not establish what the subject is worth. Value depends on the subject’s income, its tenure, its condition, its letting prospects and the buyer’s cost of capital, and on judgements about all of them that are made by a qualified person under a duty and with their own liability attached. The agent assembles the evidence; the table is the evidence; the valuation is not in the file.
The third is coverage. The table is built from comparables the office already holds or can source. Where the market is thin, the honest output is not a wider range invented to look cautious: it is a short table, a note of what is missing — a comparable in the same building, a transaction inside six months, a measured survey to confirm the area recorded for line E — and a statement of what each gap would change.
The decision at the end
Before any figure is published, the decision is whether the table is complete enough to publish at all: every line dated, every area on a named basis, every price characterised, every adjustment reasoned, and every exclusion logged with its reason. If a line fails any of those tests, it comes out of the set and goes into the log, and the range is narrower and honest rather than wider and unusable.
That is the closing decision this piece asks of its reader, and it is a decision about the evidence rather than about the property. Once it is taken, the range is the range, and the person who values the property takes it from there.
Where this goes
- The pillar: x402.review/2026/10/31/property/ — the six documents and the boundary this piece works inside
- The role: studio.x402.press/roles/real_estate_assistant — the agent that assembles the table
- The service it buys: x402.press/services/deepseek/ — the canonical price
- The review: x402.review/2026/10/31/deepseek-review/ — the measured handshake
The cost of a table
One table is eight paid model calls: one to read each source document, one to draft the adjustment lines, and one to check the arithmetic against the file. At the measured DeepSeek rate on Base — $0.00700 a short-prompt call under exact — that is $0.05600 for the table. The same eight calls at the metered ceiling of $0.00955 read $0.07640. A complete deal’s fifty calls, the six documents of the pillar, come to $0.35000, of which this table is sixteen per cent.
The figure that circulates as “DeepSeek’s price” is $0.00875, and it is not DeepSeek’s: it is Qwen’s measured short-prompt price, and the DeepSeek handshake asks for $0.00700. The difference is 25 per cent on every call in the file, and it is the reason this publication prices from the challenge in front of the call rather than from a catalogue.
Method note
Prices were read from the live 402 handshake at the DeepSeek endpoint on Base on 31 October 2026 at 09:19 UTC: $0.00700 for a short prompt under exact, $0.00955 for a long prompt under exact and the same figure as the upto ceiling at max_tokens: 1000. The 3,900.00, 4,600.00 and 4,650.00 rates in Table 2 divide the stated prices by the stated areas and reconcile to the cent, as do the adjusted rates, the four figures in Table 3 and the 239,400 span between the ends. The comparables, the subject and the adjustment percentages are constructed to make the arithmetic followable; no address, party, transaction or market is described, and none is real. Nothing in this piece is a valuation, a price prediction or advice: the agent assembles evidence and a qualified person values and advises.
x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.
Filed under Jobs · Tags: property, comparables, evidence, valuation
