An office wants to know what proportion of the flats it monitors have a balcony. In the table there are multiple “yes” entries, a few “no” entries and several empty cells. If an empty cell is treated like a No, a easily calculable number results. However, it may answer a different question than intended. First it must be clarified what is actually known.
This article explains a method of evaluation using entirely fictional data. It can be applied in Austria and Germany without equating national property statistics. It is not about a current market share nor about a technical definition of a balcony. The professional meaning of an attribute and the actual quality of data collection must be clarified before a real analysis.
Define the examined attribute clearly
Describe what the evaluation is intended to capture. “Any outdoor area” and “outdoor area confirmed as a balcony” are different attributes. A loggia or terrace must not quietly be shoved into the same group based on personal feeling. If the original list does not allow these distinctions, limit the question accordingly.
For our calculation example the attribute is already clearly defined. Each record belongs to exactly one flat in a previously cleaned own selection. Duplicate listings, different observation times and similar problems would be upstream checks; here they are explicitly not solved by the proportion calculation.
Also determine the point in time to which the confirmed entry refers. An old statement may need to be rechecked if the object description has since changed. A neat table format does not render a missing confirmation superfluous.
Use three states instead of an empty field
The usable minimum distinction is: confirmed present, confirmed not present and unknown. “Unknown” can mean that the listing says nothing, a transfer failed or the source has not yet been checked. These causes should be noted separately if necessary.
A photo that shows no balcony does not alone confirm its absence. Likewise, a personal guess from an image is not automatically a reliable Yes. Record the actually evidenced state and refrain from additions based on expectation.
For queries, a separate list of unknown cases can be useful. It should help to clarify deliberately, not trigger an unwarranted mass collection of third‑party entries. Data and sources are used only to the extent permitted and clarified for the task at hand.
Name two different proportions correctly
Our fictional selection contains twenty flats. For nine the attribute is confirmed, for six it is confirmed not present, and for five it remains unknown. Thus fifteen cases are clarified with respect to this attribute.
Nine out of twenty is 45 percent of the entire selection. The precise wording here is: For 45 percent of our selected flats the attribute is confirmed. From this one cannot assert that only 45 percent actually possess this attribute.
Nine out of fifteen is 60 percent of the cases with a clarified entry. That is a different metric with a different denominator. It also does not automatically permit transferring the result to the five unknown flats. If certain objects are currently less often fully described, this clarified subgroup may systematically look different.
Illustrate the uncertainty with a simple bound
In the example none or all of the five unknown flats could have the attribute. Relative to the same twenty cases the actual proportion would then lie arithmetically between nine out of twenty and fourteen out of twenty, i.e. between 45 and 70 percent. These are logical bounds of the invented example, not a statistical confidence interval.
The range shows why checking unknown entries can be important for the concrete question. It does not say anything about how likely a particular value within the range is. The upper bound is not a forecast either.
It would be equally wrong to blanket-fill five unknown fields with the proportion of the clarified cases and then present the addition as observed data. Any modelled completion would require its own comprehensible method and would have to remain visibly separate from collected entries. For a small office report the honest unknown group is often sufficient.
Securing the presentation against misunderstandings
Show known yes, known no and unknown entries together. A table with counts can be clearer for a small selection than a pie chart in which the unknown cases disappear. If you highlight the share of clarified cases, its denominator should be shown immediately alongside it.
In the example a concise summary would be: “Of twenty selected flats, nine are recorded with and six without a confirmed balcony; five entries are open.” Only after that may the supplemental calculation follow. This way the reader understands what the number actually represents.
If you compare two time periods, also consider the proportion of unknown entries. A changed confirmed proportion may be related to improved collection. A substantive development should not be inferred from more complete forms alone.
Ask the appropriate follow-up question
Decide whether the data gap limits the intended statement too much. Then the affected cases are specifically clarified within the permissible framework or the statement is withdrawn. If only a limited description is intended, the open group can explicitly remain part of the result.
The procedure does not provide representative market knowledge. It makes a small analysis of your own reproducible: counted unit, confirmed status, unknown entries and denominator used remain visible. This turns a seemingly exact number into understandable information.