Seattle publishes several datasets that can look suitable for a housing supply chart. Their titles mention permits, issued work or built units, but the records do not all represent the same population. Choosing the wrong file can change the answer before any arithmetic begins. A reliable analysis starts by matching the dataset's inclusion rules to the exact housing question.
This guide compares the evidence needed for a Seattle permit count, a housing unit change and a completion measure. It does not publish a newly calculated trend or claim that actual city totals were independently reproduced here. The worked examples are fictional. The aim is to help a reader choose and document a defensible dataset before creating a chart that appears more certain than its definitions allow.
State the event and the quantity separately
Write the intended question in two parts. The event might be a permit issued during a calendar year or a project completed during that year. The quantity might be permit documents, units added, units removed or net unit change. A project can contribute differently to each measure. A chart should not use the event from one definition and the quantity from another without explaining the relationship.
For example, a permit issued for an alteration may involve no additional dwelling. A new project may add many units under one permit. A demolition can remove units. Counting rows and calling them homes ignores those differences. Keep the quantity in the table heading and the event in the date filter description so another person can understand both parts of the result.
Also define the geography. Seattle city is different from King County or the larger metropolitan area. A federal regional series can provide useful context, but it is not automatically a check on a city dataset. Preserve the exact source geography and identifier. Do not shorten all labels to Seattle and assume that the boundaries match.
Read the local dataset's inclusion rules
Seattle's Issued Building Permits view describes building permits issued or in progress within the city and provides fields for permit identifiers, dates, status and housing unit information. Its metadata distinguishes added units, removed units and net change. These field names make housing analysis possible, but their presence does not establish that every row is a completed project in the selected year.
Another city dataset, Residential Building Permits Issued and Final since 1990, has a more specific population. Its published description says it includes completed or currently issued permits and does not include permits issued but not completed that fall outside that population. The description explicitly warns that it is not comparable to statistics reporting permit issuance. Read that limitation before using it as a historical issuance series.
The same description says a permit can appear more than once when the project contains different housing unit types. Therefore, repeated permit identifiers are not automatically accidental duplicates. A row may represent a housing type within a project. Deleting every repeated identifier could remove real unit information. The correct aggregation rule depends on the measure and the dataset's documented row structure.
Keep current status separate from historical issuance
A dataset organized around current status can change its population as projects move through the process. A file downloaded today may not reproduce a count of everything ever issued in a past year. Before interpreting a historical dip, ask whether the file retains the relevant historical records or only a current subset. A date filter cannot restore records excluded by the dataset's population definition.
Create a source comparison note with columns for dataset name, row meaning, included statuses, relevant dates and update frequency. Add any explicit warning about comparability. This note should be written before the main extraction. It prevents a convenient source from being used for a question that its own documentation says it cannot answer.
If the research question changes from issuance to completed unit change, revise the chart title and method together. Do not keep a familiar permits headline while switching to a completion based dataset because it produces a cleaner series. The new measure may be useful, but it needs its own interpretation and should not be presented as an equivalent replacement.
Test the row structure with a fictional project
Imagine a fictional Seattle project with one permit identifier and two housing type rows. One row shows six units added in one category, and another shows two units added in another category. A project count might count the permit once. A units added calculation might sum the relevant quantities to eight. Treating both rows as duplicate permits and retaining only one could undercount the units.
Now add a fictional removal row showing three units removed. If the documented structure supports summing these rows, gross additions remain eight while net change becomes five. The result depends on the field definitions and relationship among rows. Do not subtract a removal twice if a net field already incorporates it. Preserve gross and net measures separately rather than combining them into a single ambiguous housing number.
Suppose a second permit relates to the same address but a different phase or work scope. Address alone cannot determine whether it should be merged with the first. Check the permit identifiers and documentation. A property can have several legitimate permits, and one permit can have several unit type records. Those two relationships require different handling in the extraction.
Select dates that match the question
Seattle's permit metadata distinguishes an application date from an issued date and describes current status as a process field. Use the date corresponding to the event you intend to count. An accepted application is not the same as an issued permit, and an issued date is not evidence of completion. If the dataset lacks a reliable date for the desired event, identify another source or narrow the question.
Use explicit period boundaries. Record whether a year means January through December or another interval. If the latest period is incomplete, label it as partial and compare it only with an equivalent period when appropriate. Do not place a partial year beside full annual values without a clear warning. The visual pattern can otherwise imply a decline caused only by the shorter observation window.
Keep missing dates visible. A blank is not zero and should not silently become the date the file was downloaded. Record how many records were excluded because the necessary event date was unavailable. If those exclusions materially affect the result, say so in the findings. A precise total built from incomplete event information needs a corresponding limitation.
Build an aggregation worksheet
Retain the original downloaded file before editing it. Create a working copy with the selected identifiers, date fields, status, housing type and unit quantities. Add columns explaining any exclusions or transformations. Keep numeric zero distinct from blank or unavailable. A zero may be a meaningful reported value, while a blank may indicate that the information was not supplied.
For permit counts, document how repeated identifiers are handled. For unit counts, document which rows and fields are summed and how added, removed and net values relate. Test a small selection manually against the source description. The purpose is to catch a misunderstanding of row meaning before it becomes a citywide total, not to claim that a small spot check validates every record.
Create a reconciliation line showing the number of raw rows, retained rows, distinct permits and total units under the chosen definition. These are separate outputs. A reviewer should be able to see why the counts differ. If a residual or inconsistency remains, retain it as an unresolved issue rather than forcing agreement through an undocumented adjustment.
Use Census statistics as a defined external context
The Census Building Permits Survey measures privately owned new housing units authorized under its published definitions. Its scope differs from a local net change measure that may include removals or alterations. Before comparing, list the construction scope, ownership coverage, geography, time period and revision status for each series. A numerical difference is not evidence of an error until those definitions have been aligned.
Do not add a Census new unit total to a Seattle local unit total as though they describe separate homes. They can overlap substantially while using different definitions. If both appear in a report, explain what each contributes. One might provide a standardized regional context while the other describes a local administrative measure, but neither should be used to fill the other's missing rows without a justified method.
A federal series may also distinguish reported values and estimates that include imputation. Preserve the source labels and footnotes. The local file's administrative detail does not make it automatically more complete for every purpose, just as the federal source's standardization does not answer every building specific question. Choose evidence according to the question rather than a blanket ranking of sources.
Treat revisions as information
Save the retrieval date and release information for every extract. When updating the analysis, compare changes to earlier periods as well as adding the newest period. A record may have been corrected, reclassified or moved through the status process. A changed historical value should not be described as newly constructed housing in the update month.
Keep an exception log for records requiring interpretation. Include the relevant identifier and the reason for the decision, without exposing unnecessary personal details. If a later release resolves an issue, update the log and explain the effect on the total. This makes the analysis maintainable rather than dependent on one person's memory of spreadsheet edits.
Before publication, verify that the chart matches the calculation table and that the title names both the event and quantity. Explain why the selected Seattle dataset fits the question and which comparisons remain inappropriate. A careful result may be narrower than the original headline idea, but it is more useful because readers can understand exactly what changed and reproduce the method.
A Seattle housing permit analysis is strongest when the dataset choice is part of the evidence. Distinct permit identifiers, repeated housing type rows, status populations and unit fields all matter. Reading those details first prevents a polished graph from turning current administrative records into an unsupported claim about historical authorizations, completed homes or future rental availability.