Skip to content

Methodology

How every figure on this site is produced, what each series measures, and the floors below which nothing is published.

What each series measures

Every series states its basis in its name and on every figure. Asking means the price or rent sought in a current listing. Achieved means a price or rent that was actually paid. Valuation means a figure produced by a written valuation method. Sentiment means survey responses. The bases are never mixed in one figure, and an asking figure is never described as a price paid or as a value.

  • Dhaka Asking-Price Monitor. The median asking price per square foot for resale flats, citywide and by area, monthly. Asking basis.
  • Dhaka Land Asking-Price Panel. The median asking price per katha for residential plots, by area and plot-size band, quarterly. Asking basis.
  • Dhaka Months of Inventory. Active listings divided by the recent monthly pace at which listings leave the market, by area and price band, monthly. Asking-panel basis: it measures how fast stock turns over, not how many sales completed.
  • Dhaka Asking-Rent Monitor. The median asking rent per square foot per month for flats, by area, monthly. Asking-rent basis. Any yield shown is asking rent over asking price and is labelled an asking yield.
  • Dhaka Property Sentiment Survey. A monthly panel of market participants, published as net balances: the share reporting a rise minus the share reporting a fall. Sentiment basis.

Three further series are planned and are not published until their own conditions are met: a valuation index for a fixed basket of flats, which waits for its error rate; a household asset-class return table, whose components appear one by one as their sources pass their floors; and a measured metro premium, which waits for enough listing history on both sides of a milestone.

Disclosure floors

A figure is published only when the sample behind it reaches the floor for that statistic. The floors are enforced in code, not by judgement. A cell below its floor is shown as not yet published, with the sample so far, the floor, and the period in which it is expected to clear.

StatisticMinimum for publicationAdditional rule
Asking-price median, per celln of at least 30 active listingsAn area figure needs at least three cells passing and a combined n of at least 100
Land price per katha, per celln of at least 20Plot-size bands may be merged, with a flag
Months of inventoryat least 30 active listings and a three-month exit averageSuppressed if fewer than 8 exits in the window
Time to exitat least 50 listing spellsConfidence band shown
Achieved-price indexat least 30 pairs over four quarters and at least 8 in the quarterNo single pair counts for more than a quarter of the index by value; flats and land kept separate
Valuation error rateat least 20 pairsSigned bias reported with it
Achieved yield, per areaat least 10 managed units, and at least 20 citywideVacancy taken from the managed book
Verification ratesat least 25 verifications, and at least 15 per areaCheckpoint distribution shown
Sentimentat least 15 respondentsPanel composition class disclosed
Any internal aggregatenever fewer than 5 underlying transactionsNo cell in which one client could be identified

Sample sizes and intervals

Every published figure carries its n. Medians and indices also carry a 90 percent interval estimated by resampling the underlying listings.

Versions and the change log

The method has a version number, shown on every series page. Any change to a definition, a floor or a computation raises the version and is recorded in the change log below with its date and its effect. Figures are recomputed under the new version and the revision is shown.

Revisions

A release may revise the three periods before it, as late records arrive and duplicates are resolved. A revision table accompanies every release. A revision larger than 1.5 index points carries a note explaining the cause in the data.

Error and audit rates

  • Valuation error rate. Not yet published. The first estimate is published once at least 20 valuation pairs exist, and is then updated every quarter with its signed bias.
  • Duplicate-listing audit rate. Published with every release from the first release: the share of a random sample of listings that a manual check found to be duplicates the method had missed.

Conflict of interest

Aventra Portfolio is a real estate brokerage. It earns commission when property changes hands, so it has an interest in market activity. The statistics it publishes are produced by fixed rules that are published in advance, from records logged before the outcome of any transaction is known. No figure is adjusted by editorial judgement. Sample sizes, error rates and the change log are published on this page.

External review

An independent academic reviewer is invited each year to examine the method and publish a short note, which appears on this site whatever it says. A technical note describing the method in full detail is made available to the reviewer under agreement.

Change log

  • 2026-09-07, method version 2.0. Data Program v2.0 adopted. Series pages, area profiles and the methodology published with every cell pending. No figure has been released.