Part 1: Method

Foreword

This is a private standard published by Trueinch. It is not an Indian Standard published by the Bureau of Indian Standards, and not an American National Standard.

The True Inch Standard for Homes is published as four documents:

Edition 1.0 is published for trial use until October 2028.

• During trial use the score is validated against panels of practising architects in each market (Annex C), and the text is open for public comment (Procedures §2).

• The next full review is due by October 2028.

• Comments go through trueinch.com/score.

How to read the requirements:

• "shall" is a requirement;

• "should" is a recommendation;

• "may" is a permission;

• "can" is a possibility;

• "must" is used only for a rule outside this standard, such as a law;

• notes and examples carry no requirements.

Introduction

Most people choose a home from a drawing. The True Inch Standard rates how well a home's design works for daily living: its space, its daylight, its sun, its air, its layout and its safety. It does this by measuring the home, built at true size from its drawing, with physical simulation at its real place and facing. The result is the True Inch Score, a number from 0 to 100, and the True Inch Report, which says what the number rests on and what can be done about it.

Intended uses:

• comparing homes;

• deciding before buying, renting or building;

• finding what is worth fixing, and what to ask.

Not intended as:

• a review against a building code;

• an inspection of construction or condition;

• a valuation;

• a certified measurement;

• an energy rating;

• a judgement of taste, or of beliefs such as Vastu.

Principles:

1. Measured, not guessed. Every number comes from a computation on the model of the home, with its method, its inputs and its status.

2. The score is computed. The plan, the inputs used and the edition determine the score. Nobody, human or model, sets a number by judgement.

3. Honest about what is unknown. What can't be measured is left unscored and becomes a question. What rests on an assumption says so. Where an unknown could move the score, the score is shown as a range.

4. Anchored and labelled. Every threshold names its source and its kind: a code, a standard, a recommendation, a convention, or a measure of our own.

5. Taste and belief are not scored.

6. Open and versioned. Every check, curve and weight is published, and every report names the edition that scored it.

7. Every finding leads to an action: ask, check, change, negotiate, or accept, with the reason.

1Scope

1.1 This Part specifies how a home's design, as shown in its floor plan, is rated under the True Inch Standard, and what a True Inch Report contains.

1.2 It applies together with one market Part. A score shall be stated with both, for example "True Inch Score 74 (TIS-IN 1.0)", where TIS-IN 1.0 means Part 2-IN edition 1.0 with Part 1 edition 1.0.

1.3 The object rated is one dwelling as drawn: one or more floor plans of a single home. Each market Part lists the building types it covers.

1.4 The following are outside the rating:

• construction quality, materials, finishes and condition;

• structure and building services;

• legal status, approvals and title;

• location, neighbourhood, common areas and amenities;

• price and value;

• energy use;

• outside noise and air quality;

• the view out;

• shading by surrounding buildings and trees, unless declared (edition 1.0 assumes an open sky; 8.4).

2References

Each market Part lists the codes and standards it uses, with dates.

This Part draws on the following. They are informative: no requirement depends on them.

• OECD/JRC, Handbook on Constructing Composite Indicators (2008), on normalisation, weighting, aggregation and sensitivity analysis.

• ISO/IEC Directives Part 2, on the structure and verbal forms of standards.

• IPCC guidance on the consistent treatment of uncertainties, for the word "likely".

• EN 17037:2018+A1:2021, Daylight in buildings, for the names of its three levels (minimum, medium, high).

3Terms and definitions

4Inputs, defaults and evidence

4.1 The model. The model shall be built from the plan. Every printed dimension shall be checked against it, and the result kept as a receipt.

4.2 Inputs.

• The place is required. It gives the latitude and longitude, the time zone, the climate and the typical year.

• The other inputs are optional and have defaults:

  • the facing (true north);

  • the building type;

  • the floor level;

  • the surroundings;

  • the household;

  • the person's decision (buying, renting, owning or professional).

• The decision orders the report (10.7). It never changes the score.

4.3 Defaults. Each market Part lists its defaults. Every default a score used shall be listed in the report as assumed.

4.4 Answers from the owner. An answer from the owner replaces a default, and is marked owner in the evidence. The report shall show how many of the inputs used came from the owner.

4.5 The typical year gives the climate zone, the usual wind, the weight of welcome against unwelcome sun, the market's own weightings, and (in edition 1.1) annual daylight. Where it can't be had, the market Part's fallback tables apply, and the report says so.

5Measurement

5.1 Determinism. Each check's quantity shall be computed from the model under the conditions its market Part states. The same model, inputs and edition shall give the same quantities.

5.2 Validated engines. The engines shall be validated against references:

The method page shall link the current bench results.

5.3 Common conventions. Each market Part may vary these, saying so.

• Daylight is computed on a work plane 0.75 m above the floor:

  • on a 0.25 m grid, inset 0.2 m from the walls;

  • with furniture removed;

  • under the CIE overcast sky for daylight factors.

  Lux "for half the daylight hours" is the daylight factor × the site's median diffuse horizontal illuminance (the EN 17037 method).

• A room is in sun at a moment when at least 0.5 m² of its floor, walls or furniture below 2 m is in direct sun. Smaller patches are shown as glimpses and are not counted.

• Air comes from an airflow network:

  • wind pressure from the AIVC coefficient tables for the stated terrain;

  • the stack effect, with the inside 2 K above the outside;

  • each window open by its type's openable share;

  • doors inside open, unless the market Part says otherwise.

• The breeze comes from the indoor 3D airflow field (isothermal) in the usual wind:

  • read where people are: over a bedroom's bed at a lying head's height, 0.65 m; over a living room's seats, a dining table or a desk at a seated head's, 1.1 m; across the floor at 1.1 m in a room with none of them;

  • cooling effect by ASHRAE 55's elevated air speed method.

• Smells: a source runs for 20 minutes in the kitchen, or in a WC, in usual conditions.

  • A room's share is its peak concentration ÷ the source room's peak.

  • A room has cleared once it falls below a tenth of the source room's peak.

• Routes are walked on a 50 mm grid:

  • a person walking forwards needs 600 mm, and sideways 450 mm;

  • a way is easy at 900 mm or more, tight at 600–900 mm, sideways at 450–600 mm, and blocked below that.

• Fit uses the market's furniture at real sizes (its Annex D). A piece fits when it and its clearances lie inside the room without overlapping walls, door swings or other pieces.

5.4 Tolerances. Where a result would change level within the tolerance of the dimensions it rests on, both outcomes shall be carried into the likely range (8.2).

Note: edition 1.0 carries the unknowns of 8.2 into the range. A measure's own tolerance is carried where its market Part says how: a stair's width and going within its drawing's precision of the code's value are taken as meeting it (Part 2, SF-3).

6Checks, levels and curves

6.1 Kinds of check. There are four:

6.2 A graded check states:

• its quantity and unit;

• what it applies to: rooms, routes or the home;

• its anchors (zero, minimum, medium and high, or the reverse);

• its scale (logarithmic or linear);

• the source and kind of each anchor.

6.3 The curve. Let g(x) be ln x on a logarithmic scale, and x on a linear one. For a check where more is better, with anchors x₀ < x_min < x_med < x_high:

• p(x) = 0, when x ≤ x₀;

• p(x) = 50 · [g(x) − g(x₀)] / [g(x_min) − g(x₀)], when x₀ < x < x_min;

• p(x) = 50 + 25 · [g(x) − g(x_min)] / [g(x_med) − g(x_min)], when x_min ≤ x < x_med;

• p(x) = 75 + 25 · [g(x) − g(x_med)] / [g(x_high) − g(x_med)], when x_med ≤ x < x_high;

• p(x) = 100, when x ≥ x_high. There is no credit above high.

Rules for the anchors and the scale:

• Less is better (heat, smell, distance): the curve is mirrored, with x_high < x_med < x_min < x₀.

• Logarithmic scale for ratio quantities: lux, air changes, hours, lengths.

• Linear scale for shares and counts.

• x₀ on a logarithmic scale shall be positive. It is one tenth of x_min unless the check says otherwise.

• Where too much is as bad as too little, the check shall either use two curves and take the lower, or be paired with a counter-check in another part, named in both. Daylight is paired with unwelcome sun, and breeze with draught.

6.4 Corpus anchors.

• When used: where no standard sets levels, the anchors are fixed from the reference set.

• How set: the share of rooms below minimum should match the share below minimum on the standard-anchored checks in the same set, within 10 percentage points.

  • The standard-anchored checks are the graded checks some anchor of which a code, a model code, a standard or a recommendation sets. Their share is pooled: each room a check judges counts once, and a check judged on the home counts each home once. So does the corpus-anchored check's own share.

  • A provisional curve already within 10 points is kept. One that isn't has its minimum moved to where the shares match, to the anchor's own rounding, and keeps its shape: its zero, medium and high keep their ratios to the minimum, a rounding step apart at least (a share's zero is never past 100%).

  • A set that can't fix a curve says so. Where as many of the set's homes have none of what is measured as the minimum would leave above it, within 10 points (most homes have no floor only for passing at all), matching would make having any of it the fault rather than how much: the curve stays provisional, and the market Part's Annex A says why.

• How published: as absolute values in the market Part's Annex A, with the reference set, the shares and the date they were fixed.

• When they change: only with a new edition, never from live reports.

Note: this is deliberately lenient. A 50 means adequate, not typical.

6.5 Gates.

• Passing has no effect.

• Each failure subtracts a penalty from the check's part:

  P = min(25, 10 + 50·s), where s = shortfall ÷ required value. For a rule with no quantity, s = 0.5.

• The total of gate penalties in one part shall not exceed 50.

• Every gate failure is a finding. It needs attention, unless s is under 0.05, when it is worth knowing.

• A gate whose code allows an alternative the plan can't show (electric light instead of windows, or a fan instead of a window) shall be worded as "below the natural minimum, confirm the alternative". It shall not be worded as unlawful.

6.6 Checks with several tests.

• A room's level is the highest level whose test, and every lower test, passes.

• Between levels, points follow the curve of the quantity that decides the next level.

• Where a check has only discrete levels, the points are 100, 75 or 50 for the level reached. A room that misses the minimum only by a clearance or margin of 150 mm or less gets 25; otherwise 0.

7Aggregation

7.1 Rooms to a check.

• A check's points for a home are the mean of its applicable rooms' points, each room weighted by its room hours × its occupants (the market Part's Annex C).

• Where the worst habitable room scores below 50 on a check, it shall be listed as a factor.

7.2 Checks to a part.

P_part = clip( Σ w_c·p_c / Σ w_c − Σ penalties, 0, 100 )

• The sums run over the applicable checks that were scored.

• The w_c are the market Part's check weights within the part.

7.3 Parts to the total.

T₀ = exp( Σ W_k · c_k · ln(max(P_k, 1)) / Σ W_k · c_k )

• This is the weighted geometric mean over the scored parts.

• Each part counts by its weight times its coverage, c_k (7.6): a part scored on a third of its applicable checks carries a third of its weight. So one scored check can't swing the total as if it were the whole part, and a part that is mostly unknown waits for its questions to be answered.

• The default part weights are:

  Space	Daylight	Sun	Air	Layout	Safety & access
  20	15	15	20	20	10

• A market Part may change them only with reasons given in its rationale annex.

Note: with a geometric mean, a weak part pulls the total down more than a strong part pushes it up. A home that is dark can't buy its score back with space. Each weight works as an elasticity: 10% more Air raises the total by about 2%.

7.4 Caps.

• With one critical issue: T₁ = min(T₀, 69). With two or more: T₁ = min(T₀, 54).

• Each market Part lists its critical issues.

• The report shall show the score before the cap as a factor ("capped from 78").

7.5 Balance.

• If any scored part is below 50, T₂ = min(T₁, 84): Excellent requires every part at 50 or above.

• Otherwise T₂ = T₁.

7.6 Coverage.

C = Σ W_k · c_k / Σ W_k

• c_k is the share of part k's applicable check weight that was scored.

• If C < 0.80, no total shall be given. The report shows the parts it could score, names the checks that are missing, and asks the questions that would score them.

• A drawing with no room people live in (no bedroom, living room or kitchen: a site plan, an elevation, a barn) has nothing to score: no check applies, and the report says so.

• Otherwise the total is given, with "scored on C%".

7.7 Rounding and factors.

• The total and the parts are given as whole numbers, rounding halves up.

• Points lost by part i are:

  (100 − T₀) · W_i · c_i · ln(100/P_i) / Σ_j W_j · c_j · ln(100/P_j)

  A part at 100 loses nothing.

• Within a part, points lost are shared among the checks in proportion to w_c·(100 − p_c), with gate penalties counted as their own factors.

• Within a check, points are shared among the rooms in proportion to room weight × (100 − p_room).

• Caps and the balance rule are factors of their own.

• Displayed factors are rounded by largest remainder, so they add up to the points lost as displayed.

8Uncertainty

8.1 Three things, reported separately:

• the likely range: what is unknown about the home;

• reading confidence: how sure the reading of the drawing is;

• the method's own uncertainty: its weights and curves. This is reported on the method page only, never per report, because the edition defines the score.

8.2 The likely range. The score shall be computed over the unknowns.

• Discrete unknowns are enumerated. The facing:

  • ±22.5° when the owner gave one of eight directions;

  • ±2σ when it was read from the plan's arrow.

• Continuous unknowns are sampled with 32 Latin-hypercube draws, the same draws for every home (a fixed seed), each scored at every facing enumerated:

  • sill and head heights, each ±150 mm where not printed, a home's windows together, as one builder sets them;

  • the window type, where none is drawn, among the market's common types: the default type in half the draws, the others sharing the rest (Part 2 §4.2);

  • the terrain class, where nobody has said: the default in half the draws.

• The fast engines work out each draw:

  • the airflow network, again for each draw;

  • the daylight factor, scaled by each room's glass in the draw: the average daylight factor is in proportion to the glazed area (BRE's formula, BS 8206-2), and the light model is made once;

  • the sun, each window's sunlit glass and each room's sunlit floor scaled by its glass in the same way.

• The slow breeze field is held at its nominal value, with its bench error added: its seated speeds are scaled by a lognormal factor, a factor of two either way at one standard deviation (its bench finds it indicative).

• What a draw sets stays assumed. A check that asks about an assumption rather than failing on it (principle 3) asks in every draw, as it does of the home.

• The reported values:

  • the headline is the score at the stated defaults: the middle of the heights drawn, the commonest of the types and the terrains;

  • the likely range is the draws' 17th–83rd percentile;

  • the very likely range, shown in the details, is the 5th–95th.

• When it is shown: the likely range shall be shown when it is 3 points wide or more. When it crosses a band edge, both bands are named ("Good to Very good").

• Why the headline isn't the draws' median. A median is no home anyone can be shown: its factors would not add up to it (7.7), and every finding and figure the report shows is of the home at its stated assumptions, which the report lists. The headline is that home; the range shows how far, and which way, what isn't known could move it.

8.3 Unknown facing. When nothing tells the facing, no single total shall be headlined. The report shows the score for each of N, E, S and W, and asks the facing first.

8.4 Surroundings. Edition 1.0 assumes an open sky. This is not part of the range. The report shall say that the sun and daylight shown are the most the home can receive.

8.5 Reading confidence.

At low confidence, the report shall lead with the questions that would raise it.

8.6 Checking the ranges.

• When an owner later answers a question, the new score is recorded against the old likely range.

• The method page shall publish how often the new score falls inside it. The target is at least 66%.

• Re-reads of the same sheet shall be published as a variability figure.

9Bands and standing

9.1 Bands:

9.2 Standing. Standing is stated as "in the top X% of the N homes of its kind in the reference set", in steps of 5%.

• Its kind means its market and its bedroom group (studio or 1, 2, 3, 4 or more), where that group has at least 20 homes; otherwise its market alone.

• The reference set is frozen with each edition and published: its count, sources and places.

10The report

10.1 What a True Inch Report shall contain:

1. The score:

   • the score and its band;

   • the likely range (where 8.2 requires it);

   • reading confidence and coverage;

   • the edition, the date, the report ID and the plan fingerprint.

2. The six parts, each with its points and its main factor.

3. The market's size check.

4. The findings, each with its weight, its evidence and its action.

5. The improvements, each with its simulated before and after, a cost range with its source and date, and the score after it, added up step by step.

6. The questions to ask and the checks for a visit.

7. The inputs and defaults used, each marked owner or assumed.

8. What a plan can't tell.

9. The standards used, each with its kind.

10. Any lens the owner turned on, marked as not scored.

10.2 Every number in a report shall cite the evidence that produced it.

10.3 Words.

• A report shall not say "complies", "safe", "healthy", "quiet", "guaranteed", "certified" or "Vastu-compliant".

• "Meets" may be used only of a gate that passed, naming it: "meets NBC Part 3's minimum".

• Kinds. A report shall name each threshold's kind.

• The drawing. It shall say "the drawing shows" of the built home.

10.4 The disclaimer. The report shall carry once:

An analysis of the drawing under the True Inch Standard. Not a building-code review, an inspection, a valuation or a certified measurement.

10.5 The certificate. A shared report shall carry its report ID, plan fingerprint, date, edition, score, band and range, and a link and QR code to a public page that verifies them.

10.6 Improvements.

• An improvement shall be simulated by the engines before and after, and scored again.

• It shall be offered only if it raises the total and lowers no part by more than it raises another.

• Improvements are listed in order of benefit for their cost.

• Each says when it can be done:

  • ask the builder before it is built;

  • at any time;

  • needs approval (from a housing society or an authority).

10.7 The reader's decision orders the report. It never changes the score or the findings.

11Conformity

11.1 A score may be called a True Inch Score under an edition only if:

• the implementation that computed it passes that edition's conformance plans (the market Part's Annex E);

• it does so within 1 point on the total and 2 points on each part.

11.2 Trueinch's implementation is the reference implementation, and its version is printed in each report.

Annex A (normative): Symbols

Annex B (informative): A worked example

Four homes with part weights 20/15/15/20/20/10:

What the geometric mean does.

• Under an arithmetic mean, home D's space buys back its dark rooms, and it would be Very good.

• Under the geometric mean it is Good, below the balanced home C.

• For home A, 10 more points of Daylight add 2.9 to the total; 10 more of Space add only 1.2. The improvements that lead the report are, rightly, the ones that fix the weakest part.

A's 44.7 points lost, by part (7.7):

Annex C (informative): Validation, as pre-registered

The protocol shall be published before any architect sees a plan.

Plans. 30 per market, plus 10 more held back to confirm, chosen across:

• bedroom count;

• house or flat;

• climate;

• the quintile of their predicted score.

Each market's set includes four plans with known flaws.

Architects. Five or six per market, practising, with their conflicts declared, and paid a fixed honorarium that doesn't depend on the results. Fewer than five per market makes the result a pilot, and it is called one.

What they see. The drawing, the place, the facing and the floor. They see neither the score nor the 3D model.

The task. For each pair of plans:

1. which home works better for daily living, ignoring taste, size and price;

2. which is better on each of the six parts;

3. how confident they are;

4. one line of why.

Each architect also judges 20 common pairs, 8 repeated pairs with the sides swapped, and places 15 plans on the five bands. That is about 70 judgements each.

Targets.

Weights.

• Each architect spends 100 points across the six parts, each framed as the step from minimum to high.

• Weights are also inferred from their pairwise choices.

• A weight changes only if all three hold:

  1. the panel's median differs from it by more than 5 points;

  2. the inferred interval excludes it;

  3. the change improves agreement on the held-back plans.

Variants tested:

V2 stands unless another variant does better on the held-back plans by more than the bootstrap noise.

What is published. The protocol, the panel's make-up, the anonymised judgements, the code, every statistic with its confidence interval, and the five largest disagreements per market with what was done about each.

Annex D (informative): Why this method

Absolute levels, not a reference home.

• HERS's reference home has the same geometry as the home being rated, so it cancels exactly what this standard rates.

• Absolute levels (lux, sun hours, air changes) measured under each site's own climate keep the meaning of a score the same everywhere.

• The standing line gives the relative view.

Three named levels follow EN 17037: minimum, medium, high.

Logarithmic curves, flat at the top. For lux, air and hours, returns diminish, as in Lighthouse's curves and SAP's log branch. No credit is given above the high level.

The geometric mean across parts, caps and a balance rule follow the composite-indicator handbook's advice for parts that complement each other, BREEAM's minimum standards, LEED's prerequisites and WELL's floors.

The likely range and confidence follow IPCC's calibrated words and Spiegelhalter's levels of uncertainty. A range is given where an unknown can be bounded. A question is asked where it can't, as with an unknown facing.

Bibliography

• OECD/JRC (2008), Handbook on Constructing Composite Indicators: Methodology and User Guide.

• CEN (2018, A1 2021), EN 17037, Daylight in buildings.

• ISO/IEC Directives Part 2, Principles and rules for the structure and drafting of ISO and IEC documents.

• IPCC (2010), Guidance Note for Lead Authors on Consistent Treatment of Uncertainties.

• Google, Lighthouse performance scoring (developer.chrome.com).

• BRE, BREEAM technical manuals (category weightings, minimum standards).

• Research notes behind this Part: K: method; the research.