Good transaction analysis starts by asking what a record actually says, not what a chart appears to imply. For an Abu Dhabi pricing question, separate observed sale fields from inferred unit characteristics, state the sample count, and compare the same kind of property on the same measurement basis. A precise decimal cannot rescue an unsuitable comparison.
The ADREC market-data interface offers a Recent Sales view with filters including asset type, subtype, primary or secondary sale, district, community and project. That interface is a starting point, not evidence that every individual record contains every unit attribute an analyst wants. Check the actual record or export and note the retrieval date and any omissions.
Make a field dictionary before calculating
Create a short dictionary for the dataset in hand. Record the displayed field name, its documented definition, unit of measure and missing-value rate. Do not silently interpret a portal's date as the contract date or a displayed amount as the buyer's full cash cost. Fees, incentives, furnishing and payment arrangements may sit outside the visible fields.
| Field to check | Why it matters | Safe reading |
|---|---|---|
| Date | Defines the observation period | Use the date as labelled; do not invent a registration lag |
| Sale type | Primary and secondary deals may have different terms | Keep cohorts separate unless the question requires a mix |
| Property type and layout | A villa and an apartment are different products | Match the subject property before calculating a rate |
| Area | It is the rate denominator | Quote the documented unit and measurement basis |
| Amount | May not describe all financial terms | Identify the displayed amount, not an assumed all-in cost |
Floor, view, condition, occupancy and upgrades may be absent. If you enrich them from another source, mark the enrichment and its provenance. An inferred floor from a unit number is still an inference until checked.
Mean, median and the sample they describe
The mean answers "total divided by count." The median is the midpoint of ordered values. Consider five fictional sales at AED 1m, AED 1.1m, AED 1.2m, AED 1.3m and AED 5.4m. The mean is AED 2m because the total is AED 10m divided by five. The median is AED 1.2m. These fabricated values demonstrate why a single unusually expensive unit can move the mean; they are not current Abu Dhabi sales.
Neither number describes a specific home automatically. A median from five mixed towers can be less useful than a range from three closely matched layouts. Publish the period, location, sale type, property type, count, range and exclusions alongside any statistic. For a thin sample, show individual observations where permissions and privacy allow and resist a sweeping trend claim. There is no universal minimum count at which a median becomes reliable.
Verify the area denominator
A price-per-area calculation divides a displayed amount by an area. It says little until the area definition is known. Abu Dhabi has an area-measurement regulatory framework, which is a reason to check the applicable measured area, not a licence to assume that every source reports identical gross, sellable or internal space. Do not apply a generic gross-to-net conversion to a unit without its measured area and documentation.
When two sources use different or unclear denominators, do not combine their rates into one table. Compare total prices for genuinely equivalent layouts when feasible, or leave the rate comparison unresolved. Write "area basis not confirmed" instead of choosing a convenient conversion.
Separate price change from a change in what sold
A community median can rise because more expensive buildings supplied a greater share of sales, even if no matched unit changed value. Start with a grouped table by building, property type, bedroom count, size band and primary or secondary status. Compare each group's count and price distribution between periods. If the group mix moved materially, a community-wide percentage is a composition signal, not a clean like-for-like price index.
A cohort is a deliberately matched set. Strong candidates may include the same layout line in one building, but condition, floor, view, occupancy and deal terms can still differ. A repeat sale of the same unit removes many fixed differences, yet a renovation or changed transaction terms can alter the result. RICS's comparable-evidence guidance supports weighing evidence for relevance rather than treating a single sale as decisive.
Interrogate each candidate transaction
For each proposed comparable, ask whether the location, product, sale type, layout, size and measurement basis match the question. Then ask what you do not know about the home's condition and the deal. An unusually high or low record deserves investigation, not automatic deletion: it may be a different product, a special transaction or a data problem. Keep a reasoned inclusion and exclusion log.
Avoid a fixed "recent means three months" rule. A fast-moving launch may make an older record less relevant, while a quiet building may have no close recent sale. State the dates and why those observations remain useful. Likewise, do not attribute monthly changes to Ramadan or summer without a dated, comparable series that actually demonstrates the effect. A rolling window may reduce noise, but it also blends different market conditions.
A reproducible client-facing note
A defensible note can be short: "These six recorded secondary apartment sales were shown for this project between the stated dates. Four match the subject's layout and documented area basis; two were excluded because their sale type or area was unclear. The four retained prices span the stated range. Condition, incentives and occupancy were not confirmed, so this is context rather than a valuation." Fill the counts and dates from the actual records, never from this example.
Knownable's interactive map may help locate the relevant community. It does not verify a unit's condition or resolve an ambiguous area field. Good analysis is an audit trail: source, retrieval date, filter, included records, exclusions, arithmetic and limits. This is not investment, legal or tax advice. A professional valuation or legal interpretation needs the current underlying documents and qualified review.