Sales comparison
The adjusted grid, carried through to a rate and a value. Strongest where the comparables are close in kind, in time and in place, and the grid is what shows whether they were.
The valuation engine
Five places it refuses, and one place it concludes
The engine is deterministic. It adjusts each comparable with a stated reason, refuses a sample that cannot carry a conclusion, values by three approaches, reconciles them under weights somebody has justified, and checks every assertion the report will make against the documents on file before a word is drafted. What it will not do is produce a number that has nothing behind it.
The adjustment grid
A comparable is not comparable until you have said how it differs
Two properties are never the same property. The grid is where that difference is written down as a percentage and a reason, factor by factor, in an order that does not change between reports — which is what lets a reviewer read across the row instead of taking the rate on trust.
Transaction date
A sale from eighteen months ago is evidence about eighteen months ago. Adjusted for the movement since, with the reason stating what that movement is based on.
Location
Micro-market rather than city. Two roads apart can be a different rate, and the distance from the subject is recorded on every comparable.
Tenure
Freehold against leasehold, and the unexpired term where it is leasehold. A tenure difference nobody adjusted for is one of the commonest defects in a challenged report.
Size
Rate per unit area moves with the size of the lot or the unit, and not linearly. Adjusted with the reason naming the direction and why.
Age and condition
Effective age rather than year of construction, and condition assessed separately — a refurbished older building is not its build date.
Floor, frontage and view
The three that a spreadsheet usually collapses into one fudge. Each is its own adjustment with its own reason, so a reviewer can disagree with one without discarding the others.
Distress
A forced sale is not market evidence at market value. Adjusted explicitly, or excluded with the exclusion recorded.
The rationale is not optional and it is not a comment field. An adjustment cannot be stored without one, which means the grid cannot quietly degrade into a column of percentages that somebody once had a reason for.
What it changes
Eight comparables, before and after
The same eight sales, first averaged and then adjusted. The interesting figure is not the rate — it is how far the eight disagreed with each other before and after.
| Trimmed mean of the raw rates | ₹7,499.59 / sq ft | The eight rates disagreed by 25.3% |
|---|---|---|
| Mean of the eight adjusted rates | ₹7,779.07 / sq ft | After adjustment they agree within 3.1% |
The rate moved by under four per cent, which is the part that does not matter. The disagreement between the comparables fell from a quarter to three per cent, which is the part that does: it is the evidence that the adjustments are describing real differences rather than being applied to make a number arrive. A grid that widened the spread would be telling you something too, and the software would refuse the sample rather than average through it.
The trimmed mean survives in the system only as a pre-adjustment sanity statistic, labelled as one wherever it appears. Nothing in the software returns it under a name a reader could mistake for a conclusion.
When it refuses
Four ways a sample fails, and all four block
These are blocking errors rather than warnings, and that distinction is the whole of it. A warning on a busy day is a thing somebody clicks past; a refusal is a thing somebody has to answer.
Too few
Below the minimum count the conclusion is arithmetic performed on an anecdote. The refusal names the minimum rather than saying the input was invalid.
Too old
Past the maximum age a sale is history rather than evidence, and no date adjustment recovers that honestly.
Too far
Past the maximum radius the location adjustment is doing more work than the comparable is. The distance is on the record for every one of them.
Still disagreeing
If the adjusted rates span more than the configured threshold, the outliers are named and the valuation is refused. A wide spread after adjustment means these were not comparables — and averaging them would hide exactly that.
Three approaches
Reconciled, never averaged
Three methods produce three indicated values. Turning three numbers into one is a judgement, and the software treats it as one: the weights are somebody’s decision, each carries a written reason, and they have to sum to one.
Income
Net operating income, capitalisation and a discounted cash flow where the asset is tenanted. Reads the same lease schedule the rent roll does, so the two cannot disagree.
Cost
Replacement cost less depreciation. The method that matters most for the asset classes where comparable evidence barely exists, and the only sensible one for an insurable value.
- Weights must sum to one. A reconciliation that does not is refused by the API rather than normalised silently.
- Every weight carries a rationale in prose. A weight with no reason is not a judgement, it is a slider.
- The mandate’s purpose decides which methods are mandatory — a lending valuation of a tenanted property cannot be concluded on comparable sales alone.
- The basis of value constrains the same way: an insurable value does not get to lean on market evidence.
- Approaches that diverge beyond the threshold are refused rather than split down the middle. Two methods far apart is a finding, not an average.
A tenanted commercial property, valued three ways and reconciled. The weights and their reasons are the valuer’s; the arithmetic is the software’s.
- 0.25
- 0.60
- 0.15
The evidence gate
Nothing is drafted for a property that cannot support the report
The gate runs before the drafting rather than after it, so a property with no encumbrance certificate on file does not produce a carefully hedged paragraph about encumbrances — it produces a stop, naming the document that is missing.
- Ownership, tenure, encumbrance status, approvals held, area and age each have to resolve to a registered document, a title chain entry, an encumbrance row or an approval on file.
- An unsupported assertion blocks the render. It is never softened into hedged prose, because the hedge is the sentence a lender skips.
- “Clear and marketable title” with no chain of instruments behind it is blocked, and the block names what would satisfy it.
- A second pass reads the drafted text: the gate checks what the property can support, the scan checks what the draft actually claimed.
- There is no bypass. No force parameter, no permissive mode, no senior override — an unevidenced title assertion inside a signed valuation is the claim that ends a registration.
This is the one rule on this page with no exception anywhere in the software, and it is the reason the line exists in the shape it does. Everything else here improves a valuation. This one is what keeps it from being a liability.
Provenance
Every figure has a line, and every line has its comparables
The separation this product is built on: the arithmetic produces figures, the drafting produces prose, and the two are joined only where a figure can name where it came from.
- 01
A figure in the report
Rendered from a valuation line and nowhere else. A number written into a drafted paragraph with no line behind it is a blocking error, demonstrated against a case deliberately prompted to produce one.
- 02
The valuation line
Carries the amount, its basis, the approach it belongs to, the comparables it drew on and a reference to the source it was computed from.
- 03
The adjusted comparables
Each with its sale, its area, its rate and its adjustments — every adjustment with the percentage and the written reason.
- 04
The source sheet
The row that was imported, including the rows that were rejected and why. Nothing is dropped silently: parsed plus rejected plus duplicate equals what was submitted.
- 05
The document behind the assertion
For a statement of fact rather than a figure: the registered instrument, the encumbrance certificate or the approval, openable by the reviewer reading the sentence.
Every read and every export writes an audit row with the actor, the address it came from and the time. "Who exported this valuation, and when" is a query rather than an investigation.
The arithmetic underneath
Exact decimals, and one table of conversion factors
The dull half of the engine, and the half that determines whether a portfolio ties. Both of these are the kind of defect that surfaces months later, in a report somebody has already signed.
No floating-point number touches a rupee
Every figure is an exact decimal, parsed from text rather than from a float, and rounded once at the boundary rather than at every step. A two-hundred-property portfolio totals identically whether you sum it per property or in aggregate — to the rupee, not to the nearest thousand. Intermediate rounding is how a rent roll stops tying to its own lines, and it is invisible until it is embarrassing.
A figure that cannot be read is not a zero
A malformed amount raises rather than being coerced. A silent zero flows into a total and reconciles perfectly with nothing, which is worse than a row that stops and asks a person to look at it.
One conversion table, read by both halves
Square feet, square metres, acres, guntha, bigha and the rest come from a single shared source that the engine and the console both read, so the console cannot disagree with the report. The state-dependent local units are held as unverified against notified schedules and refuse to convert without an explicit instruction — an area wrong by a state’s factor is a valuation wrong by a multiple.
The console does no arithmetic
Money reaches the browser as a decimal string and is displayed, not calculated with. A figure computed in a browser is a figure the report cannot trace, which by this product’s own rules means it could never have been rendered anyway.
Bring a grid you have already had to defend
The fastest way to judge this is against a report that came back with questions. Which figure was challenged, what the reviewer could not trace, and how long the grid took to assemble by hand.
Get in touch
Talk to the people building it
No chatbot and no ticket queue. Tell us what your practice actually looks like — how many valuers, which asset classes, who reviews, which banks you are on the panel of — and someone who works on the software will reply.
info@legosphere.com