NovResearch and data
We publish only dated, verifiable data with the source stated. Where comparable data is insufficient, we show that honestly rather than drawing a convincing trend line. Below are our data-led pages and methodology.
What should accompany every indicator
Every figure should state the indicator name, definition, unit, geography, property type, observation period, publication date, last-check date, source and a short method description. Where NovAsia calculates the value, disclose the formula, inputs and rounding as well.
The World Bank explains that metadata is needed to use a data source effectively and can include the source, structure, methodology, topical, geographic and temporal coverage, update date and maintenance process. Linking to a report without this context does not make the individual figure reproducible.
Different sources answer different questions
Official statistics and registers support officially defined indicators and legally relevant records within their stated coverage. A developer document supports the project's stated terms on a named date, but not a market outcome. An independent consultant report provides analysis under its own method and sample, while a listing portal records asking supply rather than necessarily completed transactions.
Do not raise the evidence status merely because several websites repeat the number; they may all reproduce one original source. Research should state the role of each source, trace the value to its origin and avoid presenting listing data as an achieved sale or rent without separate evidence.
A data point has at least three dates
The data period states when the observation applies. The publication date states when the source released it. The NovAsia check date states when the editorial team confirmed that the source and wording remained available and relevant. These dates should not be replaced with one generic “current” label.
A report published in February may describe the second half of the previous year and be checked by the editorial team later still. The user needs to see that gap and understand that a recent check date does not turn an older observation period into the current market.
When two price points are comparable
A price series needs the same price type—asking, contractual or completed transaction—measurement basis, currency, property type and condition, geography, project stage and included packages. A project's minimum price at one date and the average of available units at another do not form an honest trend.
Where comparability is incomplete, show two separate observations and describe the difference rather than calculating percentage growth or decline. A price-list change can also reflect a different unit mix, incentive, furniture package, schedule or methodology rather than movement in the same asset's price.
Research should disclose the coverage of its sample
A market review should state which districts, projects, property types and development stages are included, which are excluded and why, how many observations were found and the proportion for which the required field was confirmed. Where the sample contains only publicly visible or partner-accessible projects, make that limitation prominent.
In a separate data-centre report, Knight Frank Cambodia describes a bottom-up method using detailed information for each visited facility and a classification of development stages. The example shows the value of an explicit method. For housing, NovAsia should likewise define the unit of observation and avoid calling a partial list the whole market without coverage evidence.
Observation, estimate and forecast should look different
An observed value comes from a named source for a completed period. An editorial estimate is derived from available evidence under a disclosed method. A forecast concerns the future, while a developer target or stated completion date is a promise by a named party. These types should not share one unmarked series.
Where a figure is provisional or expected to be revised, say so beside it. UK Government Analysis Function guidance recommends presenting important caveats early and explaining how quality and method affect use rather than relying on a general statement that numbers may be imperfect.
Data and indicators
How to verify data (method guides)
Methodology and transparency
Do not hide a range or conflict inside an average
Where sources use different definitions, periods or samples, retain each value and explain the disagreement. A simple average of non-comparable numbers does not resolve the conflict. For a range, state the minimum and maximum observations, sample size and composition, and whether it represents listings, confirmed transactions or scenarios.
Uncertainty does not make information useless, but it changes the conclusion that can be supported. Government Analysis Function guidance says that transparent communication of quality, uncertainty and change helps users avoid conclusions that the evidence does not justify.
How to display a revision or correction
Beside a materially changed indicator, show the earlier value, new value, change date, reason and impact on the conclusion. A scheduled update for a new period, a recalculation under a new method and correction of an error are different events and should be labelled separately.
The Code of Practice for Statistics requires revisions and corrections to be released transparently with the nature and scale of the change explained. Good practice also preserves a revision history and clearly marks provisional and revised data. A formatting edit does not need a notice, but a changed price, period, method or investment conclusion does.
Need data for a specific project? Contact us — we'll send what is confirmed, with a check date and source state.
Message on TelegramTrust Center →