A median always belongs to a selected sample
Write the sample size beside the median, with the product and date range. A result based on one disclosed sale describes that sale; it does not become a market benchmark by appearing in a chart. A small set needs closer reading of individual differences.
Show why each sale is included. If the evidence has mixed building forms, room configurations or pre-development assets, a precise calculation can answer the wrong question. Improve product matching before adding decimal places.
Distinguish no observed sale from no property
A property can exist and remain held without a recent transaction in the selected period. Stock, properties with observed sales and number of sale events are different counts. A property with more than one sale should not silently inflate the count of distinct properties.
Record what the source covers and what remains incomplete. If historical collection or enrichment is still in progress, a blank area may reflect the work queue rather than market inactivity. Avoid calling a selected database a complete census unless that coverage has been established.
Do not turn unknown fields into zero
A missing room count, bathroom arrangement, price or building classification is an unknown observation. Substituting zero changes averages, filter membership and the apparent availability of a product. Retain a distinct missing category.
Use the same population for a fraction’s numerator and denominator. If the numerator counts confirmed ensuite products, a denominator containing every keyword candidate produces a different measure. Label what the fraction describes before interpreting it as a market signal.
Align the dates behind price, rent and selling time
A sale and a rental advertisement several years apart do not describe one contemporaneous income-and-price observation. Keep each date visible. A later operating record may be relevant context but needs an explicit reason before joining it to a sale.
For selling time, distinguish campaign start, contract date, listing appearance and estimated dates. If the campaign record starts after the contract date, do not manufacture an ordinary marketing period from that interval. Missing or inconsistent chronology should stay visible.
Use a sensitivity range with stated evidence
Choose which observed comparisons support the base assumption and which differences justify alternative cases. Keep this reasoning visible in a feasibility sheet. An upside case is a scenario, not a prediction supported merely by the highest observed sale.
Bring the construction specification and operating brief back into the review. If the proposed product changes, rerun the comparison rather than preserving a convenient median. Keep source permissions and current filters with the saved result so someone else can reconstruct the decision later.
Documents to take to the next conversation
- Sample size, date range, product filter and source coverage.
- Distinct properties separated from sale events and total stock.
- Missing fields retained as unknown, with consistent denominators.
- Scenario assumptions linked to individual comparisons and differences.
Questions owners ask
Does a median from one or two sales establish the market?
It summarises that small selected set. Read the transactions, product fit and coverage before treating it as a project benchmark.
Why can a region have stock but few recent comparable sales?
Stock and recent transactions measure different things. Properties may remain held, and source coverage or classification may also be incomplete.