One-line takeaway

Record the public source, match the geography, evidence company eligibility, and retain uncertainty before moving from housing context to a company-defined service-area count.

Source-backed signal

The U.S. Census Bureau's 2024 ACS release page says the 2020–2024 American Community Survey five-year estimates were released January 29, 2026. The estimates use data collected from January 1, 2020 through December 31, 2024. The Bureau makes them available through data.census.gov, its API, and the Summary File. It publishes all data tables down to census tracts and selected Detailed Tables down to block groups.

The Census Bureau's 2024 five-year API documentation warns that variables and their meanings can change over time. It also explains that annotation variables can carry important information about an estimate or margin of error, including a state in which an estimate is not applicable or not available.

The official DP04 API dictionary defines fields for selected housing characteristics. Those fields include estimates, margins of error, percentages, and annotation variables for categories such as housing occupancy, units in structure, year structure built, and house-heating fuel. Each field has a variable ID and a stated universe. The dictionary describes housing characteristics; it does not identify an HVAC contractor's serviceable addresses, customer records, equipment inventory, leads, membership offers, or members.

The ACS 2024 subject definitions define year structure built as the year the building was first constructed, not when it was remodeled, added to, or converted. That definition is about the structure. It does not provide the installation year, condition, ownership, or service eligibility of HVAC equipment.

The ACS 2024 accuracy documentation explains how margins of error and confidence bounds accompany estimates and notes that the meaning and natural limits of an estimate matter when constructing bounds. A planning record that copies an estimate but drops its margin of error or annotation loses part of the source evidence.

These sources establish the dataset's release, coverage, field definitions, and uncertainty conventions. They do not establish demand for an HVAC membership in any service area.

Interpretation

Interpretation, not a Census Bureau claim:

An HVAC operator can use an ACS housing estimate as one public-context input in a service-area worksheet. The estimate becomes more auditable when the worksheet preserves the dataset vintage, exact geography, table and variable ID, universe, estimate, margin of error, annotation, retrieval date, and source URL.

That public-context record should end at the evidence boundary. A housing unit is not automatically:

  1. Inside the contractor's actual service boundary.
  2. A serviceable address under the company's property, equipment, access, and capacity rules.
  3. Reachable through a permissioned marketing channel.
  4. Occupied by a person who wants service or a membership.
  5. A presented membership opportunity.
  6. An enrolled, active, or paying member.

Each transition needs a separate company-defined rule and an authorized first-party source. If the rule or source is missing, leave the next state unknown.

The same boundary applies to housing characteristics. Units-in-structure categories can describe the published housing universe; they do not decide which property types the contractor serves. Year structure built describes original construction under the ACS definition; it is not HVAC equipment age. Heating-fuel fields describe a survey characteristic; they do not identify the installed system at a particular address or establish service eligibility.

Do not multiply an ACS housing estimate by an assumed response, conversion, or membership rate and label the result a market opportunity. Without separate evidence for eligibility, reachability, offer presentation, and acceptance, that calculation is a scenario—not a measured funnel.

Operator lesson

Build the denominator in two layers and name them precisely.

Layer one is public context. Use a blank source card:

  • Source agency — U.S. Census Bureau or another explicitly named public source.
  • Dataset and vintage — for example, 2020–2024 ACS five-year estimates.
  • Collection period — preserve the period stated by the source.
  • Retrieval date — when the operator checked the source.
  • Geography type and identifier — county, place, tract, block group, or another supported geography.
  • Geography label — the source-provided name.
  • Table or profile — the exact source product, such as DP04.
  • Variable ID — the exact estimate field.
  • Variable label — copied from the current dictionary.
  • Universe — the population or housing universe the field describes.
  • Estimate — the published point estimate.
  • Margin of error — the paired source value, not an invented range.
  • Annotation — preserve returned source annotations and unavailable states.
  • Source URL — link to the release, dictionary, API response, or downloaded source file.
  • Known limitation — geography mismatch, vintage, sampling uncertainty, variable change, or another documented constraint.

Layer two is company-defined eligibility. Keep it separate:

  • Service-boundary version — the dated company definition, not a Census geography relabeled after the fact.
  • Geography crosswalk status — matched, partial overlap, unresolved, or not used.
  • Property eligibility rule — the company's current operational rule and owner.
  • Equipment eligibility rule — the company's current operational rule and owner.
  • Access or permission rule — what may be contacted and through which approved source.
  • Capacity constraint — a separately measured operational input, if used.
  • First-party evidence source — the authorized system or review that supports each exclusion or inclusion.
  • Unknown count — records or areas that cannot yet be classified.
  • Duplicate rule — how overlapping geographies or records are prevented from being counted twice.
  • Review date and owner — who maintains the rule and when it was last checked.

Never put customer addresses, contact details, payment records, or private FSM exports into a public denominator card. If an operator joins public geography to internal records, that implementation requires its own authorization, privacy, retention, access-control, and identity-resolution review.

Before moving a number from public housing context into a company-defined count, run a four-check evidence handoff:

  1. Source recorded — preserve the agency, dataset vintage, source geography, universe, variable, estimate, margin of error, annotation, URL, and checked date.
  2. Geography matched — show whether the public geography matches, partially overlaps, or does not represent the dated company service boundary.
  3. Eligibility evidenced — attach the separate authorized first-party rule and source for every company-defined inclusion or exclusion; a public housing field cannot supply this evidence.
  4. Uncertainty retained — keep margins of error, unknown overlap, exclusions, duplicates, unavailable values, and missing first-party evidence visible after any calculation.

If one check is open, label the resulting number as public context or an unresolved scenario. Do not advance it to serviceable, reachable, offered, or member status.

Use transparent labels for arithmetic:

  • Public housing context = the selected source estimate for the stated universe and geography.
  • Company-defined eligible count = first-party records that meet the dated eligibility rule.
  • Reachable count = eligible first-party records with a currently authorized contact path.
  • Membership opportunity count = reachable or served records with evidence that an offer was actually presented under the company's definition.
  • Member count = records with the company's separately defined enrollment, activation, or payment evidence.

Do not reuse one number across those labels.

Synthetic example only: a worksheet records a public housing estimate of 10,000 with a published margin of error of 400 for a fictional geography. The company separately documents 1,500 units outside its actual service boundary and 700 units that do not meet its fictional property rule. The point-estimate scenario is 10,000 − 1,500 − 700 = 7,800. Label 7,800 as a synthetic planning result, keep the source margin of error visible, and show that the two exclusions are synthetic inputs. Do not call 7,800 leads, reachable households, membership opportunities, or members. This arithmetic demonstrates the ledger; it is not a contractor result, forecast, or benchmark.

Practical playbook

  1. Write the planning question before retrieving data. Name whether you need public housing context, a serviceable-address count, a reachable audience, an offer count, or a member count.
  2. Choose one dated public dataset and preserve its vintage. Do not silently mix a current company boundary with an unlabeled historical estimate.
  3. Record the exact source geography and identifier. Mark partial overlap when the company's service boundary does not match the published geography.
  4. Copy the table, variable ID, label, and universe from the current dictionary. Recheck variable changes before reusing an old query.
  5. Store the estimate, paired margin of error, and annotation together. Treat unavailable or not-applicable values as evidence states.
  6. Keep year structure built separate from HVAC equipment age. Require a separately authorized equipment source for any equipment-level claim.
  7. Define company eligibility outside the public-source layer. Name the rule owner, evidence source, effective date, exclusions, unknowns, and duplicate treatment.
  8. Show every subtraction and its provenance. Do not hide exclusions or uncertainty inside one market-size number.
  9. Stop the calculation at the last evidenced state. Do not apply an assumed conversion rate and relabel a scenario as demand or revenue.
  10. Review the card when the dataset, geography, service boundary, eligibility rule, or source definition changes.
  11. Keep customer and FSM data in authorized systems. Publish only the blank field design, source definitions, and clearly synthetic examples.
  12. Report the next evidence gap beside the result: geography overlap, eligibility, reachability, offer presentation, acceptance, activation, or payment.
  13. Close the public-to-company handoff only after the source is recorded, geography is matched, eligibility is evidenced, and uncertainty is retained.

Email version

A housing estimate is not a membership prospect count.

The Census Bureau's 2020–2024 ACS five-year release can provide useful public context. But the source comes with a specific vintage, collection period, geography, universe, variable, estimate, margin of error, and sometimes an annotation.

Preserve all of those fields before doing arithmetic.

Then stop at the evidence boundary.

A housing unit is not automatically a serviceable address. A serviceable address is not automatically reachable. A reachable household is not proof that a membership offer was presented. An offer is not an enrolled, active, or paying member.

Keep two layers:

  • Public context: source, vintage, geography, variable, universe, estimate, margin of error, annotation, checked date, and limitation.
  • Company-defined eligibility: dated service boundary, property and equipment rules, authorized first-party source, exclusions, unknowns, duplicate rule, owner, and review date.

Before crossing from the public layer to the company layer, require four checks: source recorded, geography matched, eligibility evidenced, and uncertainty retained. If one is open, keep the number labeled as context or an unresolved scenario.

One especially important boundary: ACS year structure built means when the building was first constructed. It is not the installation year or condition of the HVAC equipment.

Use public housing data to make a denominator auditable—not to manufacture a lead list, demand estimate, or revenue forecast.

Which step in your service-area estimate is currently doing the most hidden work: geography, housing universe, eligibility, reachability, or membership intent?

LinkedIn post

A public housing estimate is useful context. It is not an HVAC membership prospect count.

Before using Census data in a service-area worksheet, preserve:

  • Dataset and vintage
  • Collection period
  • Exact geography and identifier
  • Table, variable, and universe
  • Estimate and margin of error
  • Annotation and checked date
  • Known limitation

Then draw a hard evidence boundary.

Cross it only when four checks pass:

  • Source recorded
  • Geography matched
  • Eligibility evidenced
  • Uncertainty retained

Housing unit ≠ serviceable address ≠ reachable household ≠ membership opportunity ≠ member.

Every transition needs a company-defined rule and a separate authorized source. If the evidence is missing, keep the next state unknown.

And do not use year structure built as HVAC equipment age. The ACS field describes when the building was first constructed, not when its equipment was installed.

The goal is not a bigger market number. It is a denominator whose source, uncertainty, exclusions, and evidence gaps can survive review.

Short post / thread starter

Housing unit ≠ serviceable address ≠ reachable household ≠ membership opportunity ≠ member. Cross the public-to-company boundary only when the source is recorded, geography matched, eligibility evidenced, and uncertainty retained. Stop at the last evidenced state.

Community-answer suggestions

  • Ask which public dataset vintage and exact geography produced the starting estimate.
  • Request the table, variable ID, label, universe, estimate, margin of error, annotation, and checked date.
  • Clarify whether the number means housing context, serviceable addresses, reachable households, presented offers, or members.
  • Mark partial geography overlap instead of treating a Census boundary as the company's service boundary.
  • Keep year structure built separate from HVAC equipment age, condition, ownership, and eligibility.
  • Ask which authorized first-party source supports each property or equipment exclusion.
  • Preserve unavailable, unknown, partial-overlap, duplicate, and not-applicable states.
  • Show each subtraction and identify whether it comes from a public estimate, a company rule, or a synthetic assumption.
  • Reject a conversion-rate multiplier when no measured eligibility, reachability, offer, or acceptance evidence exists.
  • Keep customer addresses, contact details, payments, and FSM records out of public artifacts.
  • Name the rule owner, review date, and next evidence gap before interpreting the result.
  • Ask whether all four handoff checks passed: source recorded, geography matched, eligibility evidenced, and uncertainty retained.

Sources

One useful decision at a time

Get the next issue

Reader question

Which step in your service-area estimate is currently doing the most hidden work: geography, housing universe, eligibility, reachability, or membership intent?