Choice Bundle Normalization¶
Normalization procedure — instantiates Revealed Preference Validation Against Indifference Curves
Re-expresses every option as a common-unit bundle of attributes and prices, so trade-offs made on different occasions can be compared on the same footing.
Choices made on different days are rarely comparable as they stand: pack sizes differ, prices move, promotions distort, and the same product appears bundled one week and unbundled the next. Choice Bundle Normalization is the rescaling step that fixes this — it maps every option onto a fixed set of attribute axes and converts its price and constraints into a common unit, so that a switch from one option to another can be read as a genuine trade-off rather than an artifact of incomparable packaging. Its defining move is representational, not inferential: it builds the commensurable coordinate system in which preferences will later be inferred, and takes no position yet on what the chooser actually preferred.
Example¶
A telecom analyst wants to learn how customers trade data allowance against price. The raw plans defy comparison: one is $40 for 10 GB and 300 voice minutes; another $55 for "unlimited" data throttled after 40 GB, with 500 minutes and a three-month promotional rate; a third bundles a handset subsidy into the monthly fee. Choice Bundle Normalization dissolves the packaging. It defines canonical axes — effective dollars per GB, dollars per included minute, and a promotion-amortized monthly cost over the contract — and re-expresses each plan as a point in that space, decomposing the handset subsidy out of the recurring fee and converting the throttled "unlimited" tier into an effective usable allowance.
Once every plan lives on the same axes, a customer's move from the 10 GB plan to the throttled tier reads as a clean statement — this much extra effective data for that much extra normalized cost — that a downstream estimator can turn into a substitution rate. Before normalization the same move was apples-to-oranges; after it, it is a trade-off.
How it works¶
- Fix the attribute schema. Choose the canonical axes on which all options will be described (allowance, quality, price-per-unit), coarse enough to be stable across occasions.
- Map raw features onto the axes. Translate each option's idiosyncratic spec sheet into a vector on those axes, decomposing combos into their component attributes.
- Convert price to a numéraire. Express every price and budget in one common unit — per-unit, inflation-adjusted, or promotion-amortized — so cost is comparable across dates and packages.
- Reconcile units across occasions. Align quantities, currencies, and time bases so that two choices months apart sit in the same coordinate frame.
Tuning parameters¶
- Schema granularity — few coarse attributes versus many fine ones. Finer schemas capture more of what varies but multiply missing-data problems and can invent distinctions the chooser ignored.
- Numéraire choice — which unit becomes "price." A poorly chosen numéraire can manufacture apparent kinks in the trade-off surface where none exist.
- Temporal/promo adjustment — amortize promotional pricing over the contract, or take face value. Amortizing reveals the true long-run trade-off but assumes the chooser reasoned that way.
- Bundle decomposition rules — how a combined offer is split into attribute contributions when its parts are never priced separately.
- Missing-attribute imputation — how absent spec values are filled before an option can be placed on the axes.
When it helps, and when it misleads¶
Its strength is that it makes cross-occasion comparison possible at all, and it surfaces hidden equivalences — two superficially different plans that occupy nearly the same point, so a switch between them reveals almost no preference. Without this step, trade-off estimation is comparing quantities that were never on the same scale.
Its failure mode is that normalization is never neutral: the attribute schema and numéraire are modeling choices that can create or destroy apparent trade-offs. A hedonic decomposition that assumes attributes are separable — that the value of data doesn't depend on how many minutes come with it — will smooth over real interaction effects and hand the estimator a distorted surface.[1] The classic misuse is treating the normalized bundle as ground truth rather than as one lens, then reading structure in the data that is really structure in the schema. The guarding discipline is to test how sensitive the downstream trade-offs are to the schema and numéraire, and to keep the raw, un-normalized record available for re-cutting.
How it implements the components¶
Choice Bundle Normalization realizes the archetype's commensurability layer — turning heterogeneous options into comparable coordinates:
attribute_bundle_model— its primary output: the fixed-axis, common-unit representation of every option as a bundle of attributes.constraint_and_budget_context— it rescales prices and budgets into the shared numéraire so cost is comparable across occasions.
It does not reconstruct which options were available or log the raw selection (feasible_option_set, observed_choice_record) — that is Budget Set Reconstruction, whose output this step re-expresses — and it computes no trade-off rates (revealed_substitution_trace), which is Marginal Substitution Estimator.
Related¶
- Instantiates: Revealed Preference Validation Against Indifference Curves — this builds the comparable coordinate space the rest of the pipeline reasons in.
- Consumes: Budget Set Reconstruction supplies the raw options and context that this step re-expresses in common units.
- Sibling mechanisms: Budget Set Reconstruction · Marginal Substitution Estimator · Indifference Region Visualization · Dominance Violation Scan · Revealed Preference Consistency Matrix · Preference Reversal Probe · Choice Architecture Confound Audit · Stated vs Revealed Gap Report · Ethical Preference Inference Review
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The mechanism converts heterogeneous options into a common attribute-and-price schema and computes comparable effective quantities, producing an analytically normalized basis for trade-off comparison.
Nearest alternative: Intervention, Treatment & Transformation — It rewrites the options' representation, but the target options themselves are not treated; the load-bearing operation is a common-unit calculation used for comparison.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Economics & Finance
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Consumer and revealed-preference analysis established converting prices and attributes to comparable bundles before inferring tradeoffs.
Review resolution: Consumer and hedonic analysis supplies the complete recognizable lineage: represent heterogeneous goods as comparable attribute bundles before inferring tradeoffs. Coordinate mathematics and scaling are enabling techniques, not independent mechanism origins.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Hedonic pricing, formalized by Sherwin Rosen (1974), decomposes a good's price into the implicit prices of its underlying attributes. It is the standard way to place heterogeneous products on common axes — and its separability assumptions are exactly what a normalization schema inherits and must be stress-tested for. withdrawn registry ↩