Grading a Number: Our Research Method

How does Victaura Research grade a number?#

Victaura Research grades a number by labelling how it was obtained, before anyone discusses what it means. A grade is a short label attached to a figure. It states who produced the figure, by what method, and how far the reader can lean on it. There are four grades, measured, industry estimate, broker-reported and directional. The method asks for one on every statistic and every table in an article, together with a source written as institution, document, year.

The grade attaches to the number, not to the article or to the publisher. One report can hold a measured line and a directional line side by side. A statistics office publishes counts from a survey and also modelled projections. The projection does not inherit the status of the count. The principal is told which is which, line by line, because a modelled figure that looks like a counted one can be underwritten as if it had been counted.

The purpose of the grade is narrow. It does not make a figure true. It tells the allocator how much weight the figure can bear and what to do next. The next step may be to underwrite on it, to cross-check it, or to treat it as context only. This article applies the method to itself. Each figure below carries its grade and its source, and where access was second-hand, the grade says so.

The four grades: what each one means and what it permits#

Measured means that the source counted or recorded the thing itself, and says how. Examples are official statistics, registry data and audited results. The grade says nothing about relevance. A measured occupancy figure for a country is accurate for that country and silent about one building. Measured also does not mean current, so the date of the count travels with the figure every time.

Industry estimate means that a research firm or an industry body produced the figure with a stated method, and the reader cannot rebuild it. Research reports by consultancies, banks, industry bodies and the research teams of property firms fall here. The figure may be sound and the firm competent. It can be used for sizing a segment and for comparing places, not as a precise underwriting input.

Broker-reported means that an agency or a developer published the figure about the market it sells in. Asking prices, price ranges in an agency's market guide and payment plans in listings fall here. The figure is useful but not independent, because the publisher earns from the market it describes. The grade follows the figure and not the firm, so a property firm's wealth count can be an industry estimate while a price range in its guide to a market where it sells is broker-reported.

Directional means that the figure may indicate a direction, and its size cannot be relied on. Surveys with undisclosed samples, figures quoted in the press without a source document, and counts with no stated method are directional. A directional number can justify a question. It cannot justify a decision on its own, and it should never be added to a measured number in the same sum.

A calculation by the desk is not a fifth grade. Where Victaura Research combines figures from different sources itself, the source line names the desk and the inputs, in the form 'Victaura Research calculation from' followed by the documents used. It is not an index and not a proprietary dataset, and we do not have either. Any error in the arithmetic is ours. An assumption of the desk is labelled as an assumption, and each input keeps its own source, so the reader can redo the calculation or challenge an input.

What is the source hierarchy behind a grade?#

The hierarchy starts with the instrument that creates the right or the rule. For property, tax and residence, that is the statute, the regulation or the registry entry, read in its original text wherever possible. A summary on an agency or a competitor's website is not that instrument. Where a fact depends on a legal threshold, the rule is to cite the legal text or the authority's own notice, never an agency or a competitor's page, or to leave the threshold out of the article.

Primary sources come first, institutional research next, and the press last. Primary sources are statutes and regulations, regulators, statistical institutes, land registries, company filings and investor relations material. Institutes publish definitions together with their data, which lets a reader check what was counted. Institutional research means reports by property brokers, hotel consultancies and banks. Those from firms such as Knight Frank are useful and carry commercial context. The press is used mainly to confirm that an event happened and on what date.

The ranking is about checkability, not prestige. A small authority that publishes its definitions can outrank a large firm that does not. Equally, a high-ranking source can be the wrong one for the question. A national tourism figure is a poor guide to a single island resort. At each rung the test is the same. Can the reader open the document and find the same number, with the same definition, on the same date?

How is each figure checked against its source?#

Before an article enters the publication queue, an automated review reopens the sources it cites and checks each figure and each attribution against them. What a source does not support is corrected or removed. The review also looks at the date of each source, since a figure can be copied correctly and still be out of date. Every article published since 30 September 2026 has been through it. Most older articles have also been checked against their sources. That check is not yet finished for all of them, so a figure in an article published before 30 September 2026 may not have been through it.

The review tests the copy, not the source. It asks whether the number is in the document cited, with the same meaning and date, not whether the document is right. A modelled estimate that matches its report passes and is still a modelled estimate, so the grade matters after the review as much as before it.

When sources disagree, the article says so. The method asks each article to say where the evidence is thin, where sources disagree and which conclusions rest on assumptions, and most articles do this in a section of their own. This article does it in the Knight Frank section and in its section on weak points.

Agreement between sources is weaker evidence than it looks. Two press articles citing one report are one source, not two, and two industry estimates may share the same upstream data. A reader weighing two sources can ask whether each could have observed what it reports without the other. This is a judgment, not a test.

What does a measured number look like? Eurostat, and Ras Al Khaimah (RAK) as a contrast#

European tourism occupancy data is a clear picture of a measured series. Under Regulation (EU) No 692/2011 concerning European statistics on tourism, member states report on tourist accommodation establishments to Eurostat. National reference metadata for the occupancy dataset, for example the Danish and Estonian entries, point to that regulation and to Eurostat's methodological manual for definitions. The reader can open the definition. That is what makes the series measured and not merely published.

Measured still has limits, and the limits are written into the grade. The Latvian metadata entry notes that its accommodation survey is not a sample survey, so the estimates carry no sampling error. The same entry reports over-coverage, under-coverage and unit non-response, and says the precise size of the coverage errors cannot be calculated. The Croatian entry lists the activity groups covered under NACE, the EU classification of economic activities, as 55.1 to 55.3 (hotels, holiday and other short-stay accommodation, and camping grounds), and adds 55.9, other accommodation, for national purposes. Whatever sits outside the listed groups sits outside the count. A reader comparing two countries has to check that the same units were counted before comparing anything.

An authority's own count, with no method stated, falls short of measured. The Ras Al Khaimah Tourism Development Authority (RAKTDA) reported 1.35 million overnight visitors to Ras Al Khaimah (RAK) in 2025, up 6%, with tourism revenues up 12%, as relayed by Gulf News in a page last updated on 19 January 2026. For the first half of 2025 the authority reported more than 654,000 visitor arrivals, as relayed by the Emirates News Agency. The authority's own press releases, dated 19 January 2026 and 30 July 2025, state the same figures and were read at the source. Neither release says how visitors were counted, and a count with no stated method is directional under the definitions above, so both figures are graded directional. The authority is reporting on its own results, and we did not find a Statistics Centre publication for these counts in the sources checked.

A visitor count describes an emirate, not demand for any asset in it. The figure says how many people stayed overnight. It does not say where they stayed, at what rate, or whether supply grew faster than arrivals. Using it as proof of demand for a single building is a category error, and stating the unit counted next to the figure guards against it. Hotel occupancy and rate data would be a different series, with a different source and its own grade. Victaura holds an off-plan capital position on Al Marjan Island in Ras Al Khaimah, as the disclosure at the foot of this article states, and the principal should read this example with that interest in mind.

Can an industry estimate be used? The Knight Frank wealth count#

An industry estimate can be used to size a segment and compare places, not to count buyers, and the Knight Frank Wealth Report 2026 is a clean case. As relayed by trade press, the 20th edition counts people with more than USD 30 million in net assets as ultra-high-net-worth individuals and puts the global population at 713,626 in 2026, against 551,435 in 2021. We could not open the report itself for this article. The figures are therefore graded industry estimate, with second-hand access, and should be checked against the original before anyone reuses them. The automated review later found two of them, the global count and the 2021 baseline, in Knight Frank's own results article of 23 April 2026. The report itself was still not opened.

A count of private net worth cannot be a measured number in the statistical sense. Individual balance sheets are private, so any count of people above a wealth threshold is inferred from indirect evidence. That is a legitimate method and an estimate by construction. The figure 713,626 is precise to the unit and not accurate to the unit. That gap between precision in print and accuracy in fact is the reason the grade exists.

What the reader can do with the number is bounded. It can size a segment. It can show shares of the global count, which the coverage gives as 35% in the United States and nearly 31% in Asia-Pacific, and it can show direction over five years. It cannot be turned into a number of buyers for a specific market, because the step from a global count to local demand needs a second source and a stated assumption. Where we make that step ourselves, the source names Victaura Research as the author of the calculation and states the assumption.

Other country-level values from the same coverage are not used here. The two trade-press accounts we read round and word the same country projections differently, which is the kind of variation a second-hand reading produces. The global count, the 2021 baseline and the United States share appear identically in both accounts. The Asia-Pacific share appears only in the Family Wealth Report account. Knight Frank's own results article does not give either share as a share of the global count, so both are press-reported and graded directional. Those are the only values we repeat, and only with the grade attached.

What did the Henley 2026 report change, and what does it teach about method?#

The Henley Private Wealth Migration Report 2026 was released on 16 June 2026 without a new set of millionaire inflow and outflow estimates. Henley's own methodology page says the framework 'does not seek to measure actual millionaire inflows or outflows', and says that in some cases headline inflow and outflow figures have been drawn into political narratives or public debate. Its press release says the change should not be interpreted as a move away from wealth migration analysis, and its methodology page describes a framework of 12 weighted dimensions and 38 indicators. We found no formal retraction notice in the sources we checked. We therefore describe what the edition did, not what it formally withdrew.

The earlier headline counts had been challenged in public. Tax Policy Associates published an analysis dated 27 July 2025, updated on 18 August 2025 and again on 16 June 2026. It checked the last significant digit of the counts of millionaires, centimillionaires and billionaires in the Henley data for 2022 to 2025 and found that 42.5% of 299 were odd, against about half expected by chance, and put the probability of a result that low at 0.5%. For the 159 migration-flow numbers of the same years, the headline counts themselves, it found 41.5% odd and put the chance of that at 2%. Spear's reported on 19 August 2025 that Henley had initiated an independent audit of the data provider's methodology. Tax Justice Network, writing about the 2026 edition on 16 June 2026, says the firm had not provided any update on that audit. Tax Policy Associates and Tax Justice Network had both publicly criticised the earlier reports, and the digit test is an inference about a process, not proof of intent. We grade their account directional.

We treat the case as a lesson in method. A headline count was widely repeated, then its producer changed approach. Anyone who repeated the count was left with a series that its producer did not update in 2026. The lessons are three. A migration headcount is directional at most. Its definition belongs next to the number. A later edition of the same source is worth checking before an older figure is reused. Readers of our piece on where the world's wealth is moving should apply the same filter to any migration headcount, whoever published it.

One Henley figure is kept, with a narrow meaning and a low grade. The press release states that applications from UK addresses to the firm rose 15% from 2024 to 2025. That is a count of the firm's own applications. The release does not say how it was counted, so it is graded directional, and the firm is describing demand for its own services. It is silent about the market and cannot be generalised to all wealthy individuals. The Family Office Almanac's reading, that Henley changed method after years of criticism, is commentary and not a source for any figure.

What does Victaura Research not use as fact, and why?#

The rule is that residence thresholds and tax rates are not taken from agency or competitor websites. Such pages are marketing documents and can be out of date. A threshold changes the tax result for a principal, and an error there is expensive. The rule is to cite the statute or the tax authority's notice, or to leave the number out and say that it needs advice in the jurisdiction concerned. Not every published article meets the rule, and some still take a tax amount or a visa rule from a property portal or an adviser's summary. A threshold in any of our articles should therefore be checked against the statute or the tax authority's notice before it is relied on.

The press is used mainly to confirm events, not as the record for a statistic. A report that a law was passed, a ceasefire was agreed or a project opened is confirmable against a second report. A figure quoted in an article without its source document is not. Where the press is our only access to a report, as with part of the data in this article, the grade says so. The state of any ongoing event is written with its date, as of a stated day and with a source, and never as resolved or over.

Forecasts, ranges and momentum claims without a stated method are not presented as data. Where an article needs one as a working assumption, the source labels it as a Victaura Research assumption. Anonymous 'sources close to' statements are not used. A figure that an agency or a developer publishes about the market it sells in is graded broker-reported, and a seller's claim about its own scheme carries the same caveat. What turns a claim into something checkable is a document the principal can open, which is the subject of the documents a developer should show you.

A critic's analysis is used for what it documents, not for what it concludes. Tax Policy Associates published its digit counts and its method, and a reader can recompute the proportion of odd digits. Its reading of the pattern as a sign of human intervention is its own, and it leaves open whether that was deliberate. We cite the first and do not adopt the second. The same discipline applies to producers and to sellers, which is why the rule does not depend on whether we like the source.

How does Victaura Research correct a published number?#

A correction is a dated revision note visible to the reader, not a silent edit. When a figure, a source or an attribution is found to be wrong after publication, the article is corrected in place, and a note at the top states the date and what changed. Longer corrections are listed in a section at the end, with what the article said and what it says now. The article's dateModified field, which records the date of its most recent modification, is updated to match, so that the reader and search engines see the same date.

A source that changes its method can change the grade of an older figure. If a source later changes its method, as Henley did in 2026, a figure taken from an earlier edition may no longer deserve the grade it was given. Re-grading it does not require that the original was wrong. It records that the ground moved under the number, which is the normal fate of estimates and the reason every figure carries a year.

A reader who finds an error can report it through the contact page, and a confirmed error is corrected in the same way. A source document settles the question, not the reader's recollection or ours.

Weak points of this method, honestly disclosed#

Grading is a judgment, and judgments differ. Two analysts can place the same figure in different grades, in particular at the boundary between industry estimate and directional. We publish the grade so that disagreement is visible. We have no external audit of our grading and no measure of how often we are wrong, and we do not claim that the method is more reliable than the practice of any other research house.

Part of the access in this article is second-hand. The Knight Frank figures were read through trade press, and the automated review then found two of them, the global count and the 2021 baseline, in Knight Frank's own results article. The Ras Al Khaimah visitor figures were checked against the authority's own releases, which report on the authority's own results. Each grade says so. A figure read second-hand can carry a transcription error, and an original document would be better. The corrective is to open the original, which the references allow the reader to attempt.

The review can fail quietly. It tests whether a figure is in the source cited, so a source that is itself wrong passes, and when two sources share an upstream, their agreement can look stronger than it is. It does not catch every error. The hierarchy also favours what has been published, so a true fact that nobody has published stays outside our articles. The cost of that bias is that we omit where others might guess.

This article had no access to the Henley data and cannot rule on it. We describe the criticism, the producer's stated change of approach and the dates. We do not assert that the earlier series was wrong, fabricated or accurate. The critics' accounts are graded directional for that reason. On the audit, the latest statement we could read is Tax Justice Network's of 16 June 2026, that Henley had given no update, and we found no published audit result as of 8 October 2026.

How is the conflict of interest declared?#

Every article ends with a disclosure, and the disclosure is identical in each. The conflict of interest is structural. A research desk that sits beside a developer has an incentive to favour the places where the developer works. A fixed text at the foot, rather than a sentence composed for each article, means the author cannot soften it. The wider approach is described on our approach page.

Disclosure does not remove bias, and the grade is the second line of defence. A reader who assumes commentary may be influenced can look at the grade and the source and set the prose aside. That is why the method asks for a source the reader can open on each number, and for a document and not the front page of a website. Not every article meets that, and some references, mostly in older articles, point to the home page of a site. The test of an article is whether it survives with the commentary removed.

The article is classified as marketing material, and it is written to be read that way. It contains no recommendation, no price of any project and no claim about return. It describes a method. The reader who wants to test that description has the references, the grades and the correction procedure above.

Skin in the game disclosure. Victaura, through its parent Greystone B.V. (Netherlands), develops projects on Lake Como (Italy), in Nungwi, Zanzibar (Tanzania), on Gili Air and in Uluwatu, Bali (Indonesia), and holds an off-plan capital position on Al Marjan Island, Ras Al Khaimah (UAE). Readers should assume commentary may be influenced by, or benefit, these positions. This document is classified as marketing material under MiFID II Article 24(3). It is not investment advice.