NovWhat I do when two credible sources disagree
How Ekaterina Andreeva handles conflicting official, company and market sources without simply choosing the more convenient number.
This article reflects the named expert’s practical perspective. See NovAsia’s editorial policy for how material is prepared and reviewed.
Two reputable sources can disagree without either one being useless.
A government series may count registered transactions, a consultancy may track a defined slice of new supply, and a developer may publish its own current list price. Put those numbers beside each other without definitions and you manufacture a contradiction.
My first check is therefore not “which source do I trust more?” It is “are they measuring the same thing?” Geography, property type, period, transaction status and methodology usually answer that question faster than reputation alone.
That sounds simple, but a surprising amount of disagreement disappears once the labels are unpacked. One source may report asking prices while another records completed transactions. One may cover an entire city, another only a prime district. A consultancy might exclude small projects or resale stock. A developer naturally reports its own inventory. The numbers can all be accurate within their own frames and still be unsuitable for direct comparison.
Units create another layer of false conflict. Price per square metre can refer to different definitions of area. A headline average may be arithmetic while another source uses a median. Currency conversion dates may differ. One dataset may include taxes or fees that another excludes. A reader sees two clean numbers; the editor has to see the definitions behind them.
If the definitions really do match, I move to timing and provenance. When a genuine discrepancy remains, it belongs in the article. The NovAsia market-data section is built around that discipline: comparable evidence is more useful than a tidy average made from unlike inputs.
Provenance matters because two websites do not always mean two independent sources. A media article may quote a consultancy release, an aggregator may repeat the media article, and a blog may copy the aggregator. Counting all three as separate confirmations would create artificial confidence. Whenever the claim matters, I want to know where it began.
The reverse can also happen. Two genuinely independent sources may use different methods and reach different results. In that situation, averaging the values can be the least informative option. The midpoint is mathematically neat but may correspond to nothing in the market. If one number is an asking-price series and the other is registered sales, the average merely hides the distinction.
Sometimes the disagreement is itself the point. A developer campaign and a resale asking price can both be current while describing different parts of the market. A later document may supersede an earlier timetable.
When two sources really do claim the same thing and still disagree, I narrow the question. Is there a newer edition? Did one source revise its methodology? Is there a footnote changing the geographic boundary? Does one figure include cancelled transactions or only completed ones? Can the original dataset be checked? These details are less glamorous than the final number, but they often explain why credible sources diverge.
There are also cases where the conflict cannot be resolved. The editorial response should not be to choose the figure that fits the narrative. It may be better to present both, explain the difference and identify the conclusion that remains supportable. Two datasets can disagree on the exact magnitude of a decline, for example, while still indicating the same direction of change. That common ground may be the strongest claim the evidence allows.
I treat dates the same way. If an older project document lists one completion date and a newer official document lists another, the discrepancy is not noise to be averaged. The sequence matters. A later document may have superseded the earlier one. The article should show that chronology instead of pretending there has always been a single stable answer.
This approach can make copy less dramatic. “Sources report different values because they define the market differently” is not as punchy as a single definitive figure. But it gives the reader something more useful: a way to understand why the numbers vary and which one fits the question they are trying to answer.
The editorial job is to explain what survives the disagreement and what no longer can be claimed with confidence. A good article does not need to eliminate uncertainty. It needs to prevent uncertainty from being disguised as precision.