How FlexiDealsHQ scores deals

Every product we track gets a FlexiScore from 0 to 100 and a verdict. The score is a weighted sum of five components, each scored 0-100 from our own price observations. The weights below are read from the same configuration file the scoring service uses (version 2.0.0), so this page cannot say one thing while the code does another.

The formula

FlexiScore = clip(
    0.35 × authenticity_score
  +    0.25 × historical_percentile_score
  +    0.20 × forecast_score
  +    0.10 × cross-retailer_score
  +    0.10 × quality_score
, 0, 100)

Verdict thresholds

Rules are evaluated in this order; the first that matches wins. The scoring service and this page run the same cascade.

1. Early read: fewer than 14 days or 30 observations of history. The score is still shown, the verdict is not definitive.
2. Buy now: FlexiScore ≥ 85 and forecast confidence ≥ 70%.
3. Wait: our forecast predicts a drop of more than 5% within 14 days. A credible predicted drop outranks "Good deal": we will not tell you to buy today when the price is about to fall. It never outranks "Buy now", which carries its own forecast-confidence gate.
4. Good deal: FlexiScore ≥ 70.
5. Wait: FlexiScore ≥ 45.
6. Skip: FlexiScore below 45.

Authenticity and the former-price test

A retailer's "was" price counts only if it was the prevailing price for at least 30% of the last 90 days in our own observations. Otherwise the page labels it "Retailer's list price" and the authenticity component is scored down. This mirrors the FTC Guides Against Deceptive Pricing (16 CFR Part 233) and California's 90-day prevailing-price rule (Cal. Bus. & Prof. Code 17501).

The same test written for a shopper rather than for a scoring service, with the checks you can run yourself on a retailer's page: How to spot a fake discount.

Historical percentile

Where today's best price sits in the last 90 days of our tracking: at the 90-day low the component scores near 100; at the 90-day high it scores near 0. Percentiles are always computed over our own daily observations, never over a retailer's claimed history.

Forecast

A model trained on our price history and the U.S. retail calendar estimates the expected price move over the next 14 days with a confidence. An expected drop lowers the forecast component; low confidence keeps a high score from becoming "Buy now". A successor launch expected within 45 days leans the verdict toward Wait.

Cross-retailer

How the offer compares with every other retailer we track for the same product, condition and configuration. Carrier installment, activation and trade-in prices are shown as promotions and never scored.

Quality

Data completeness: number of retailers, number of observations, specification coverage, and image and category presence.

How we substantiate claims

Substantiation records are kept for 4 years. "Lowest price in 90 days" always means in our tracking, and every price carries the timestamp at which we observed it.

Commission-blind by construction

Mutate all commission rates and cookie windows; assert identical scores, verdicts, and rankings. No affiliate field is reachable from the scoring code; a continuous-integration test mutates every commission rate and cookie window and asserts identical scores, verdicts and rankings, so a build that broke the rule could not ship. We will publish an audit summary here each quarter once the site carries live pricing; none has run yet, because no verdict has yet been produced from real price data.

Cashback portals (Rakuten, TopCashback, Capital One Shopping) are shown in the true-cost calculation even though their cookie can override ours. We accept that revenue loss for accuracy.

Generated explanations

Explanations are data-first: every number is a slot filled from the record, and the connective text is reviewed under our editorial policy (2% sampled weekly, all flagged text reviewed). Each generated section is labeled "Generated from FlexiDealsHQ price data and reviewed under our editorial policy."

The same rules, written for shoppers

This page is the specification. The pages below are the hand-written explanation: what the rules mean when you are standing in front of a retailer's product page deciding whether to buy. All of them are bylined and reviewed by a person, and none of them quotes a current price, a discount or a saving for any product.

• How to read a price history — What the shape of a price chart tells you, what it leaves out, why the 90-day window matters, and how to tell a real floor from a one-day blip.
• How to spot a fake discount — Inflated list prices, permanent sales and pre-event markups: what U.S. law requires of a "was" price, and how to check one yourself.
• The real price: membership, shipping, tax and card rewards — The gap between the listed price and what leaves your account, how memberships and store cards change the comparison, and why we show cashback portals that cost us revenue.
• When prices actually drop — The U.S. retail calendar as it really behaves: which events move which categories, why tax-free weekends rarely help with electronics, and why a successor launch beats any sale.

There is also one buying guide per tracked category, covering what matters in the specification, what is marketing noise, when that category genuinely discounts, and what a real discount looks like in it.

• Buying a refrigerator: fit first, features last
• How to buy a TV without overpaying
• Laptop buying guide: the specs that decide how long it lasts
• Noise-cancelling headphones: what the specs mean
• Robot vacuums: which specs predict a machine you keep using
• Smartphone buying guide: timing beats specification

All guides and explainers.

Change log