# Brightbank inputs and evidence decisions

All values below are authored demonstration assumptions. They were not fitted to real soda purchases or generated by the app's paid GPT analysis. The creative was AI-generated and visually reviewed. The numerical feature scores were assigned manually after inspecting the supplied material.

## Product and acquisition

One modeled purchase is one twelve-can pack. Each can is 12 fl oz with a fictional 35 g added sugar. The baseline price is $28.80, shipping is $6.95, and the example tax assumption is 7% of merchandise after discount plus shipping. Tax is excluded from revenue, while processor fees apply to the total charged including tax. The model is not a tax service.

Product cost is $10.80 per pack. The $8.50 merchant shipping-cost field includes assumed outbound shipping, pick/pack and packing material. Processor cost is 2.9% plus $0.30. These are invented inputs, not a supplier quote or current payment-provider price. Overhead, spoilage, returns, repurchase, subscriptions and lifetime value are excluded.

One Instagram feed impression per consumer; assumed $9.50 CPM. No targeting auction, ad delivery optimization, click fraud or previous media spend. The art and body copy contain no price, so each price scenario uses the same acquisition probabilities. No real media was purchased.

The baseline has a visible promo field, a fictional BRIGHT10 code for 10% off merchandise, and an assumed 25% coupon-search success probability. The automatic-discount scenario applies the same discount without searching. The mock carts accept BRIGHT10 and pass the simulated discount to checkout. Guest flows offer simulated wallet and PayPal selections. No payment provider is contacted.

## Population

| Input | Value | Interpretation |
|---|---:|---|
| Population | 10,000,000 | Individually evaluated IDs, not a scaled sample |
| Seed | 20260915 | Common preferences and random streams across journeys |
| Structurally uninterested | 65% | Assumed nonmarket for this product; zero purchase eligibility |
| Active need | 28% | Conditional on the other 35%; 9.8% overall in expectation |
| Offline only | 18% | Assumed unwillingness to buy this product online |
| Median budget | $32 | Generic modeled available purchase budget |
| Mobile share | 87% | Assumed device composition |
| Brand familiarity | 2% | Fictional early brand exposure |
| Deliberation | 0.42 | Generic research tendency |
| Price sensitivity | 1.25 | Model multiplier, not empirical elasticity |
| Distrust | 0.55 | Generic merchant-reassurance requirement |

There is no direct trait for sugar avoidance, dietary restrictions, orange/vanilla preference, age or health. Structural noninterest is an aggregate assumption and cannot identify the reason an actual person would reject this drink. No sensitive demographic profiles or real customer records are used.

## Manual baseline scores

Scores are 0 to 1. The evidence supports descriptions of the materials; it does not establish the score's behavioral effect.

| Feature | Score | Manual rationale |
|---|---:|---|
| Ad clarity | .88 | Product, flavor, sweetener and pack size are explicit |
| Ad attention | .82 | Large readable headline, strong orange contrast and can focal point |
| Ad credibility | .72 | Specific ingredient positioning, but unverified fictional claims |
| Ad benefit | .70 | Flavor and occasion implied; no health or performance claim |
| Ad offer | .62 | Pack size and CTA shown; price and delivered cost omitted |
| Query match | .50 | Neutral stored value; Instagram does not use search-query matching |
| Page match | .94 | Same flavor, packaging, pack size and claims as ad |
| Page clarity | .76 | Price and add-to-cart action are explicit |
| Product information | .35 | Baseline lacks full ingredients and nutrition |
| Answer access | .30 | Baseline directs questions to support without accessible details |
| Product proof | .25 | Product copy only; no fabricated reviews or test results |
| Merchant trust | .57 | Basic return statement, limited support detail |
| Page usability | .78 | Plain mobile-friendly purchase layout; not a measured usability score |
| Cart clarity | .66 | Pack price shown; shipping and tax are deferred |
| Information access | .35 | Cart links to the limited page |
| Coupon prominence | .82 | Visible promo field can prompt a search |
| Upsell pressure | .15 | Low residual modeled distraction; no added upsell product |

Complete-PDP scenarios change information to .90, answer access to .90, proof to .40, trust to .72 and cart information access to .90. The HTML/PDF add ingredients, full nutrition, storage, delivery, return terms and a support control. Proof remains modest because there are no customer reviews or verified product demonstrations. These assumed score changes are visible in the configuration JSON.

## Checkout and interventions

| Step | Baseline fields / seconds / error | Shorter flow fields / seconds / error |
|---|---|---|
| Account/contact | 4 / 16 / 1.2% | 4 / 8 / 0.3% |
| Address | 7 / 24 / 3.5% | 5 / 12 / .875% |
| Shipping | 1 / 7 / .8% | 1 / 3 / .2% |
| Payment | 5 / 25 / 2.5% | 4 / 12 / .625% |
| Review | 1 / 8 / .5% | 1 / 3 / .125% |

Guest/wallets changes payment access and the account gate. The combined scenario also reduces contact fields to two. The supplied forms reflect these controls, but waiting times and failure rates are hypothetical, not telemetry from the mockups. Baseline card-flow required fields match the declared counts. Selecting a simulated wallet or PayPal skips card entry in the mockup; the engine uses generic stage friction and does not reproduce every payment-specific field interaction.

Free shipping changes only the customer shipping charge; merchant fulfillment cost remains $8.50. The final combined scenario additionally discloses $6.95 shipping on the PDP. The six-price experiment changes only pack price on the complete-PDP/guest/shorter-forms scenario. The supplied PDF describes the $28.80 anchor product; no price-point-specific visual analysis is claimed.

All runs use the unchanged `consumer-1.0.0` authored stage coefficients: ad -4.1, page -1.7, cart 1.1, checkout 1.6. No fabricated observed funnel was used to calibrate them. The 14-day generic return input is retained across comparisons, but the engine does not specifically model perishable-food returns.

## Provenance and reproducibility

`scripts/brightbank-fixture.mjs` is the source of the configurations and HTML. `scripts/brightbank-study.mjs` records actual computed outputs. `evidence-manifest.json` hashes every source asset; `evidence-map.json` identifies the materials for each scenario. Paid GPT analysis was not executed because no provider key was supplied. The missing-key guard was tested in the API and browser.