Methodology: How Our Calculators Compute

Methodology Principles

Five principles govern every calculator and every reference article on this site.

(1) Methodology is industry-standard. Every formula is derived from ACCA Manual J 8th Edition (load calculation), Manual S (equipment selection), Manual D (duct design), ASHRAE Handbook of Fundamentals (design conditions, psychrometrics), and AHRI 210/240-2023 (equipment rating standards).^[1] No proprietary inventions or unsourced shortcuts.

(2) Inputs use sensible defaults. Calculators open with default values (typically a 1,500 sq ft house, climate zone 4, average insulation) that produce a worked-example result immediately. The defaults are listed and justified on each calculator's methodology section; they are not picked to flatter the output.

(3) Formulas are shown, not hidden. The methodology section on every calculator page details the formula used. The TypeScript source code is in /lib/calculators/ and is the same code that produces every worked-example URL.

(4) Sources are cited. Every numeric constant, lookup table, or methodology choice cites the standard it comes from. The complete bibliography is at /sources/.

(5) Limits are stated. Every calculator's methodology section explicitly notes what the calculator does NOT account for (duct losses to unconditioned spaces, blower door measurements, custom internal gain schedules, etc.). The next section of this page goes through the limits across all calculators.

The Primary-Source Stack

The site's load-calculation and equipment-selection math comes from a small set of authoritative documents. Each calculator traces every formula and constant to one of these.

The full bibliography of cited documents — including peer-reviewed academic papers and government research lab publications that support specific claims — is at /sources/. Every article on the site has its own SOURCES list within the article, and every numeric claim links via inline citation to the relevant SOURCES entry.

Per-Calculator Methodology

Each calculator implements a specific subset of the primary-source stack. The Option C architecture (covered in the tools hub) means each calculator page renders both the live interactive widget and a worked-example default state with the math explained step-by-step.

The math implementations live in /lib/calculators/ as TypeScript modules. Each module exports a pure function (no side effects, deterministic output for given inputs) plus a TypeScript type for the input. The same function is called by the live calculator widget and by every worked-example URL — there is no separate "example calculation engine" that could drift from the live calculator.

The Heat Transfer Multiplier (HTM) approach used by both the BTU calculator and the Manual J calculator works as follows: for each envelope element (wall, ceiling, floor, window, infiltration), HTM = U-factor × design temperature difference, in BTU/h per square foot.

Multiplying HTM by the element's area produces the design heat loss (or gain) for that element. Summing across all elements produces the whole-house total. The math is identical in principle to the abbreviated method published in ACCA Manual J 8th Edition Appendix.^[1]

Accuracy Against ACCA Reference Cases

ACCA publishes reference cases used for software certification: fully-specified houses with envelope geometry, R-values, U-factors, infiltration rate, occupancy schedule, and location, paired with ACCA's published expected heating and cooling loads at the local design conditions.^[4] Software vendors run their tools against these cases and report deviation; ACCA approves software that lands within their tolerance bands.

We are not ACCA-approved and do not submit cases to ACCA. But we run our calculators against the publicly-available reference cases and report the deviation here.

The pattern that emerges: the calculator tends to read slightly high on heating load (mean +10% across the test set) and slightly high on cooling load (mean +14%), with wider error bands on older, leakier housing stock where simplified infiltration models break down. This bias toward over-estimation matches the conventional wisdom — planning-grade tools cannot make precise judgments about envelope quality and default to conservative assumptions.

The implication for users: if the Manual J calculator says "30,000 BTU/hr cooling," the real Manual J answer is most likely between 23,000 and 30,000 BTU/hr, with 26,000-27,000 most probable. That accuracy is more than enough to distinguish a 2.5-ton correct sizing from a 4-ton grossly oversized contractor quote. It is not enough to defend a specific tonnage selection in a permit submission.

The Manual J verification methodology article walks through each reference case individually with deviation analysis and failure-mode discussion.

Data Files and Open-Source Code

The data files backing the calculators are available in the project's public git repository at github.com/hvacloadcalc/hvacloadcalc.org.

The calculator code in /lib/calculators/ is similarly public — Manual J HTM tables, Manual S tolerance rules, NEEP CCASHP capacity-retention rules, and BTU/sqft baselines by climate are all in TypeScript with inline comments pointing to the underlying source documents. Anyone reproducing the calculations independently can do so from these files plus the cited standards.

The public availability is intentional. Educational reference content is more credible when the underlying code and data can be inspected. A homeowner running the same calculation independently and getting the same answer confirms the calculator does what we say it does. A researcher building a similar tool elsewhere can borrow the data files (with attribution) rather than re-encoding them from the original standards.

Known Limitations of Online Calculators

Every online HVAC calculator, including the ones on this site, operates under limitations that a Manual J performed by a credentialed party with site access does not face. We document the limitations explicitly so users can judge fit to purpose.

No site visit. A Manual J calculator cannot see the actual house. Wall thickness, insulation type, window orientation, sun exposure, basement coupling, and many other factors that affect load are estimated from defaults rather than measured.^[1]

No blower door data. Actual envelope air leakage is measured by blower-door test at 50 Pa pressure differential. Calculator inputs that ask the user to estimate infiltration ("average", "leaky", "tight") substitute a category for a measured value. The category-to-CFM mapping is the dominant source of accuracy error in older housing stock.

No duct system characterization. Ducts in unconditioned space leak conditioned air and exchange heat with the surrounding environment. A leaky attic duct system loses 20-30% of supply air to the attic regardless of how well the equipment is sized; a tight sealed-and-insulated duct system loses 3-8%. The calculator cannot measure this and defaults to an assumed loss.

No appliance audit. Internal heat gains (people, lights, appliances, cooking, hot water) depend on specific occupancy patterns and equipment. The Manual J defaults assume 1,200 BTU/hr sensible plus 200 BTU/hr latent per typical occupant, but actual values vary widely. A household with multiple gaming computers running 24/7 has very different internal gains than a household with two telecommuters working from quiet laptops.

No code interpretation. Local code requirements vary by jurisdiction. The calculator uses ASHRAE design temperatures, which is the right default for planning but may not match what a permit office in a specific jurisdiction requires.

No envelope quality verification. Wall R-values depend on how well the insulation was actually installed. The calculator assumes the labeled R-value is achieved; field measurements often show 70-85% of nameplate due to gaps, voids, and compression.

For high-stakes equipment selection (permits, rebates, high-cost installs), these limitations are why Manual J should be performed by a credentialed party using ACCA-approved software with site access. For planning-grade work (budget estimation, contractor quote evaluation, single-room sizing), the limitations are acceptable.^[4]

Update Cadence: When Standards Change

The standards we cite are updated by their respective bodies on a multi-year cycle. When a referenced standard changes, we update the relevant calculator and the methodology documentation.

Material changes are logged in the corrections page. The last-reviewed date on each calculator page reflects the most recent methodology review against current standards; a calculator labeled "Last reviewed 2026-05-30" has been checked against every referenced standard as of that date.

When a new standard is released, we typically wait 60-90 days before updating the calculator implementation to (a) allow the standard's errata period to elapse, (b) verify the changes against multiple manufacturer reactions, and (c) update internal tests. Critical safety-related changes (refrigerant regulation, CO standards) are implemented faster.

How to Report a Methodology Error

Methodology errors fall into three categories.

Numerical errors. A specific value in a data table or default disagrees with the cited primary source — e.g., the DOE recommended attic R-value for zone 4 is listed as R-49 but the DOE source actually says R-38 to R-60. These are clear bugs; report via the contact page with category "Correction" and we will fix and acknowledge.

Methodology disagreements. A reasonable practitioner disagrees with our implementation of an ACCA or ASHRAE procedure — e.g., the infiltration calculation uses one formula from Manual J 8th Edition while the practitioner argues a different formula from the same standard is more appropriate. These are valid topics for editorial discussion; report by email with rationale and we will evaluate and respond.

Out-of-date claims. A standard has been updated and our content reflects the previous version — e.g., NEEP releases v4.1 with different capacity-retention thresholds. Report by email with the link to the updated standard and we will refresh the affected pages.

We acknowledge all methodology error reports within 14 days. Material corrections are logged at /corrections/ with the original claim, the corrected claim, the date of correction, and (when applicable) the original reporter. Reporter names are published only with explicit consent.