Why three surveyors give three valuations.
Instruct three chartered surveyors to value the same Hong Kong flat and you can get three different numbers, none of them wrong. Valuation is an opinion within a range, not a single true figure. Here is the margin of error the courts accept, the six reasons valuations diverge, and why consistency is the real case for an explainable model.
By the QPV Founder. Published 13/06/2026.
A valuation is an opinion within a range.
Property valuation is not measurement. There is no single correct figure waiting to be read off an instrument. A valuation is a professional opinion of the most likely transaction price, formed from comparable evidence under uncertainty. Two competent valuers, working to the same standards on the same property on the same day, can reach different numbers, and both can be right.
This is not a weakness of the profession or a sign that someone made a mistake. It is intrinsic to any estimate built from imperfect comparable evidence. The standards bodies that govern valuation, the International Valuation Standards Council (IVSC), the Royal Institution of Chartered Surveyors (RICS), and the Hong Kong Institute of Surveyors (HKIS), all formally recognise "valuation uncertainty" and require valuers to disclose it where it is material.
The practical takeaway: a single surveyor's number is one defensible estimate, not the truth. The useful question is not "what is the property worth" but "what is the range, and how wide is it for this property". That is exactly the question an explainable valuation should answer with a confidence band.
The law puts a number on how much valuations can differ.
English valuation negligence case law, which Hong Kong valuation practice and the RICS standards draw on, has developed an "acceptable bracket" or "permissible margin of error". A valuation is not negligent simply because it differs from another valuer's figure or from the eventual sale price, provided it falls within a reasonable bracket around the correct value. The leading modern statement of the brackets is K/S Lincoln v CB Richard Ellis Hotels Ltd [2010] EWHC 1156 (TCC), building on the older principle in Singer and Friedlander v John D Wood and Co [1977].
The brackets, as applied in that line of case law, are approximately:
- Around plus or minus 5 percent for a standard residential property with ample comparable evidence.
- Around plus or minus 10 percent for a one-off or less typical property.
- Around plus or minus 15 percent, or a little more in genuinely exceptional cases, for an unusual or complex property with no good comparables.
Two points matter for anyone reading a Hong Kong valuation. First, the courts explicitly accept that "two valuers may reach different valuations of the same property at the same time without either being negligent". The disagreement is lawful and expected. Second, a figure that falls outside the bracket is evidence that something may have gone wrong, but it does not by itself prove negligence; the claimant must also show the valuer's method fell below that of a reasonably competent valuer. The bracket is a screen, not a verdict.
Read this correctly. The plus or minus 5, 10 and 15 percent figures are the UK and RICS legal benchmark for the acceptable margin of error. They are not a measured Hong Kong statistic. There is no published study quantifying how far Hong Kong valuations of the same property actually diverge. Treat the brackets as the professional and legal standard for tolerable difference, not as data about the Hong Kong market.
Why two valuers land on different numbers.
The disagreement is not random and it is not bias. It comes from six legitimate differences in how two competent valuers approach the same property.
- Comparable selection. Which past transactions each valuer treats as genuinely comparable, and how many exist, is a judgment call. In a dense estate there may be dozens of like-for-like sales; for an unusual property there may be almost none. The fewer the comparables, the more the choice of which ones to use moves the answer.
- The valuation date. In a moving market a few weeks changes the figure. Hong Kong residential prices have been rising month on month into 2026, so two valuations dated a fortnight apart can legitimately differ on timing alone.
- Property-specific assumptions. Floor level, view, orientation, condition, fit-out, and whether a unit is treated as standalone or part of an en-bloc all carry premiums and discounts that each valuer weighs slightly differently.
- Method weighting. Sales comparison, income, and cost approaches can all apply, and how each is weighted is a matter of professional judgment, especially for properties that are part home and part investment.
- Data access. Valuers see different transaction datasets. One firm may have richer recent deal flow in a particular district than another, and the data each can see shapes the estimate.
- Scope and instruction. The purpose of the valuation and the instructing party's guidance change the basis. A mortgage valuation, an insurance reinstatement figure, and a probate valuation of the same flat are answering different questions. This is legitimate scope difference, not improper influence.
None of these is a defect. They are the normal degrees of freedom in expert judgment under uncertainty. Notice that the first reason, comparable density, is also why valuations diverge more in some Hong Kong locations than others, which is the subject of our companion piece on how valuation behaves by district.
The spread is wide where comparables are thin.
The acceptable bracket is not fixed at one number; it widens with the property. That is the whole logic of the 5, 10 and 15 percent gradient. The driver is comparable density: how many genuinely similar, recent transactions exist to anchor the estimate.
In Hong Kong this maps cleanly onto property type. Large, homogeneous estates with many near-identical units and constant turnover, the kind found across much of Kowloon and the New Territories, produce a deep pool of like-for-like sales. Valuations there cluster tightly, near the narrow end of the bracket. At the other extreme sit Peak houses, low-rise luxury on Hong Kong Island, sea-view premium units, pre-war walk-ups, and village houses. Each is close to unique, comparables are sparse or non-comparable, and view, floor, and condition premiums are large and subjective. Valuations there legitimately spread toward the wide end.
This is why "three surveyors, three valuations" is barely visible on a Mei Foo or Taikoo Shing unit and very visible on a Peak house. It also explains why the same property can be priced quite differently by two banks: each instructs its own panel surveyor, and the thinner the comparable evidence, the further apart two competent panel firms can reasonably land.
The profession admits the uncertainty in writing.
Valuation disagreement is not a loophole; it is built into the professional standards themselves.
HKIS. The Hong Kong Institute of Surveyors publishes the HKIS Valuation Standards, current 2024 edition, maintained by the General Practice Division and kept aligned with International Valuation Standards (IVS) and the RICS Red Book. General Practice surveyors are the professionals who produce property valuations in Hong Kong, and many are dual HKIS and RICS qualified.
RICS. The RICS Red Book Global Standards, December 2024 edition incorporating IVS, applies to RICS members operating in Hong Kong, with separate Hong Kong application guidance. The Red Book frames a valuation as an opinion within a range and requires valuers to disclose material valuation uncertainty rather than present false precision.
IVSC. The International Valuation Standards formally recognise valuation uncertainty as an inherent property of the exercise and require it to be communicated where significant.
There is also a forward-looking signal worth noting. The RICS professional standard on the responsible use of artificial intelligence in surveying practice came into effect on 09/03/2026, setting mandatory expectations for how AI tools are governed and disclosed in valuation work. The profession is not resisting modelled valuation; it is standardising it. An explainable model that publishes its method and its uncertainty sits comfortably inside that direction of travel.
How to read a number that could have been different.
If a single valuation is one point inside a range, the practical move is to stop treating any one number as final and start triangulating.
- Buyers and owners. If a bank valuation lands below your agreed price, do not assume it is the truth. Get a second, independent data point. If a second estimate lands near the bank's number, the bank looks fair and you can renegotiate or fund the gap with eyes open. If it lands meaningfully higher with a tight range, you have grounds to test another lender or ask for a fresh surveyor inspection.
- Brokers. Expect different panel firms to differ, especially on thinly traded stock. A defensible independent estimate is leverage when arguing a bank number is conservative.
- Risk teams and family offices. For anything you hold or lend against at scale, what matters is not one valuer's opinion but a consistent, auditable basis applied the same way every time, with the uncertainty made explicit.
The constant across all three is consistency. Human valuation gives you authority and on-site judgment, but a different result each time it is instructed. The complementary tool is one that gives you the same answer from the same inputs, every time, with the reasoning shown.
Consistency is the point of an explainable model.
An Automated Valuation Model does not abolish valuation uncertainty; nothing can, because the uncertainty is in the market, not the method. What a well-built, explainable model adds is repeatability and transparency around that uncertainty.
QPV is an independent, explainable AVM for Hong Kong residential property. Its V1 prototype dataset covers 7,096 Hong Kong transactions across 35 plus districts, with 914 properties valued end to end. Three things make its output different from a single human opinion or a black-box bank AVM:
- Repeatability. The same inputs always produce the same output. There is no instruct-and-hope variance between one run and the next.
- A confidence range, not a false point. Every valuation ships with a band, so the user sees how wide the uncertainty is for that specific property, narrow on a liquid estate, wider on an unusual unit, exactly as the margin-of-error gradient predicts.
- The comparables and weights, shown. Every valuation lists the comparable transactions used and the weight on each driver, so the estimate can be inspected and reproduced rather than taken on trust.
QPV does not replace a HKIS-registered chartered surveyor where a sworn, on-site valuation is required. It is the consistent, auditable second opinion for the far more common case where you simply want to know whether one number you have been given is fair, and how much it could reasonably have been different.
What people ask about valuation differences in Hong Kong.
How much can two valuations of the same property differ?
Under the UK and RICS margin-of-error case law that Hong Kong practice draws on, the acceptable bracket is roughly plus or minus 5 percent for a standard residential property with ample comparables, plus or minus 10 percent for a one-off property, and plus or minus 15 percent for an unusual or complex property with no good comparables. A figure inside that bracket is not negligent. These are the UK and RICS legal benchmark, not a measured Hong Kong figure; no published Hong Kong-specific dispersion statistic exists.
Is it normal for two surveyors to value the same flat differently?
Yes. The courts explicitly accept that two competent valuers can value the same property differently at the same time without either being negligent. Valuation is an opinion within a range. The difference comes from comparable selection, the valuation date in a moving market, property-specific assumptions, method weighting, the data each valuer sees, and the scope of the instruction.
Does a valuation outside the bracket mean the surveyor was negligent?
Not on its own. A valuation outside the reasonable bracket is evidence that something may have gone wrong, but negligence is not established by the bracket alone. The claimant must also show the valuer's method fell below that of a reasonably competent valuer. The bracket is a screen, not a verdict.
Why do two Hong Kong banks value the same property differently?
Banks lend against the lower of purchase price and bank valuation, and each bank instructs its own panel valuation firm. Two banks instructing two different panel firms on the same flat can land on different figures, each defensible within the acceptable bracket. The spread is widest on thinly traded property where comparables are sparse.
How does an AVM reduce valuation disagreement?
An explainable AVM is repeatable: the same inputs always produce the same output, with the comparables and weights shown. It does not remove uncertainty, which is intrinsic to valuation, but it makes the estimate consistent and auditable rather than dependent on which individual was instructed. QPV publishes a confidence range and the comparables behind every valuation, so users see the uncertainty rather than a single false-precise number.
See the range, not just a number.
Request access to the platform, or read the full methodology behind every QPV valuation.