Planning guide for the San Francisco edition. Examples are hypothetical unless explicitly identified otherwise. Reviewed October 5, 2026.
A housing statistic is easier to trust when another reader can reproduce it. For the San Francisco edition, a useful source note identifies the original table, release, geography, variable and any calculation applied to the value.
A link to a general agency homepage is not enough to establish which observation was used. Preserve a direct source reference and the relevant identifiers. If the original file changes, the saved release information and a dated extraction record help explain what the article actually reported.
Describe transformations plainly. A subtraction, inflation adjustment or aggregation changes the meaning of the source value and should be visible. Do not label a derived figure as though it were an independently published official statistic.
This article provides an editorial method rather than new survey findings. It does not claim that a source note eliminates every error. It makes the evidence and calculation inspectable so that a correction can be specific and future updates do not silently change the basis of the article.
Your evidence worksheet
The six checks below are specific to this decision. The accompanying worksheet instructions explain how to preserve evidence and resolve an unanswered field. They are a proposed process, not a claim that Homzora has inspected a property, tested a product or collected a new local dataset.
Original source
Obtain: The agency and direct table or file. Why it matters: A homepage alone may not identify the observation.
Keep the original source and its date beside this entry. If two documents disagree, record both versions and ask the responsible party to resolve the difference. Do not select the more convenient figure merely because it improves the result. The working record should explain which version was accepted and why.
Release
Obtain: The exact dataset vintage and period. Why it matters: Keep publication and observation dates separate.
Write down what is known, what is assumed and what still needs an answer. These statuses should remain separate when the information is copied into a calculator or report. A blank field should stay visibly unresolved until evidence supports an entry; it should not silently become zero or a positive confirmation.
Geography
Obtain: The name and stable code. Why it matters: Names can be ambiguous or change.
Identify the person or source able to confirm this point. Ask a focused question that can produce a usable answer, then retain the response with the relevant record. If the answer is conditional, carry the condition into the decision rather than reducing it to an unconditional number or statement.
Variable
Obtain: The exact field and definition. Why it matters: Similar labels may describe different measures.
Test this item against the same scope and period used elsewhere in the worksheet. An answer for a different date, household, property or transaction may provide context without resolving this case. Explain any difference before using it, and do not force unlike records into one total without qualification.
Transformation
Obtain: Every calculation applied to the source. Why it matters: Label derived results as derived.
Consider what would change if this entry were revised after the initial decision. Identify the dependent payment, calculation or conclusion so that an update can be made consistently. Keep the previous value and reason for the correction in the working history rather than leaving two unexplained competing versions.
Revision record
Obtain: What changed and why. Why it matters: Preserve enough information to explain an earlier value.
Assign a completion condition and a review date. The condition should describe the evidence needed to close the question, not simply the act of sending an email or opening a file. If the evidence remains unavailable, state the limitation and decide whether the unresolved point prevents the next commitment.
A fictional source review exercise
This exercise concerns a reproducible source note for a housing statistic. It uses invented values solely to test editorial reasoning. It is not an additional Homzora dataset, a completed survey or a reported estimate for San Francisco. Imagine that a draft contains two monthly figures, $1,600 and $1,650, and proposes the sentence that the second location is more expensive for every renter.
The arithmetic difference is $50. That calculation alone does not support the proposed sentence. First identify what each figure measures, the population covered, the geographic boundaries and the reference period. If one is a listing price and the other a historical median, the comparison needs to be reformulated before any statistical question is considered.
Now suppose both figures are estimates of a matching measure from the same release, but they have published uncertainty. The writer must use the appropriate comparison method before making a significance claim. This exercise does not supply the information or perform the test required for that claim. It therefore cannot establish that the difference represents a reliable ranking.
Even a justified comparison of regional estimates would not establish the proposed statement about every renter. Households occupy different homes and face different agreements. The correct scope of the conclusion must follow the scope of the measure. A broad estimate can inform context without determining the price of an individual available apartment.
Rewrite the source note before rewriting the headline. Name the original source, table, variable, geography and period, then describe the calculation and its limitations. If one of those fields is missing, mark it as unresolved. Do not invent an identifier or infer a period from the date a webpage was updated.
Finally, give the material to another reviewer and ask what can be reproduced. The reviewer should be able to locate the inputs and identify the transformation. If the exercise instead concerns original survey work, the reviewer needs the actual design and release evidence. A methodology plan cannot be substituted for collected responses, and an illustrative calculation must remain labeled as illustrative wherever it is reused.
Write down the claim before choosing the number
Start with the sentence the research is intended to support. Identify the population, place, period and measure that the sentence requires. This prevents a convenient statistic from quietly changing the question. A figure can be accurate in its original context and still be unsuitable evidence for the claim being written.
Distinguish a descriptive statement from an inference. Reporting a published estimate is different from claiming that it explains a household outcome, predicts a future price or ranks every neighborhood. The latter claims require additional evidence and methods. A source citation does not automatically support every interpretation attached to the number.
Use the original source when available and keep its identifiers. A table number, variable code and geography code can be more useful for reproduction than a broad page title. Preserve the release and observation period as well as the date you accessed the material. Those dates answer different questions.
Check comparability before calculating a difference
Confirm that the measures describe sufficiently comparable populations and definitions. A city, a county and a metropolitan area are different geographic units. A current advertisement and a historical occupied housing estimate are different kinds of observations. Arithmetic does not resolve those differences; the text must explain them or avoid the comparison.
Check the reference periods. A multi year estimate should not be labeled as a reading from the final month of that period. A page modification date is not the observation date. If the intended question concerns change over time, consult the source’s comparison guidance before treating releases as directly comparable.
Inspect units and adjustments. Dollars, percentages, counts and rates cannot be substituted for one another. A median is not a mean, and a difference between two category medians is not necessarily the price of changing a feature in the same home. State any transformation in language a reader can follow.
Keep uncertainty attached to the result
Where the source publishes sampling uncertainty, retain the relevant information rather than presenting the point estimate as exact. A margin of error must belong to the same field, geography and release as the estimate. Do not borrow a more convenient margin from another table or suppress it because the headline becomes less decisive.
Use an appropriate method for statistical claims. The Census Bureau’s ACS comparison guidance is a starting point for understanding the requirements. A casual visual rule is not a substitute for a justified test, particularly when estimates involve related populations or periods. If the test has not been performed, do not imply that it has.
Sampling uncertainty is not the only limitation. Measurement, coverage, recruitment and missing information can affect what the data support. For an original survey, explain the actual design and response process. Do not apply the language of a representative sample to an open invitation merely because many people responded.
Make correction and reproduction possible
Keep a short calculation record with the original values and the transformation used. A reviewer should be able to reproduce the published result without guessing which spreadsheet cell or file version supplied it. If rounding is applied, preserve the unrounded calculation in the working record and describe the display convention.
Write the limitation near the finding it qualifies. A distant methodology page can provide detail, but the immediate wording should not overstate what the reader is seeing. Label hypothetical examples, derived figures and original observations differently. These categories serve different purposes and should remain distinguishable in both tables and prose.
Finish with the practical implication the evidence actually supports. Sometimes that implication is a better question or a reason not to rank two options. Research adds value when it clarifies a decision and its uncertainty. It does not need to produce a dramatic conclusion every time a new table is published.
Use the worksheet with Homzora resources
Open the related housing decision tool. Use the tool only for the inputs and calculations it supports; keep the additional evidence checks in your own record. Read the related San Francisco planning guide for surrounding context. Return to the San Francisco edition.
Sources and scope
The original worksheet, fictional examples and arithmetic are Homzora editorial material. The following primary sources support the general methods identified here. They do not verify the invented prices, provide a quote for this city or establish that a specific provider or property meets your needs.