DealFlipAI Methodology & Scoring Reference

What this page is

This is the reference specification for how DealFlipAI turns a Facebook Marketplace listing into a 0–100 deal score: the factor weights, the band cutoffs, which valuation logic each category routes to, which risk signals are checked, and what the system explicitly cannot do.

It exists because “AI-powered” is not a methodology. If you are going to act on a score — or cite one — you should be able to see what produced it.

For narrative explanations rather than the specification, see how the deal score works, how accurate the valuations are, and how risk detection works.

Score factor weights

The base score is a weighted blend of five factors. These weights sum to 100%.

DealFlipAI deal score factor weights
Factor	Weight	What it measures
Price vs estimated value	50%	How far below the estimated resale value the asking price sits. The single largest factor, but capped — see the score bands below.
Condition signals	15%	Condition parsed from the listing text: like new, good, fair, for parts, damage wording, and repair language.
Listing quality	15%	How completely the seller described the item: description length, specifics, and whether key identifying details are present.
Category-specific factors	10%	Signals that only matter in one category — mileage and title status for vehicles, model and storage for electronics, brand and material for furniture.
Valuation confidence	10%	How much usable data the valuation had. A confident estimate on a clearly identified item counts for more than a guess.

Risk penalties are applied after this blend, and some listings are removed before scoring entirely.

Score bands

DealFlipAI score band meanings
Score	Rating	What it means
80–100	Excellent	Strong opportunity if the listing verifies in person. Worth messaging first and inspecting promptly.
65–79	Great	Promising, but usually needs negotiation or cleaner comparable pricing before the margin is real.
50–64	Good	Possible margin, but not urgent. Fine to watch rather than chase.
0–49	Fair	Usually too thin, too risky, or too uncertain to justify the time and travel.
Filtered	Filtered	Removed before scoring. Applied to listings whose price or wording indicates the listed price is not the real purchase price (down-payment and finance-lead patterns), parts-only items presented as working, and similar cases.

A bigger discount does not raise the score indefinitely. Past a threshold, an extreme discount stops reading as a bargain and starts reading as a problem — a misdescribed item, a down-payment listing, or a scam — and the risk checks begin reducing the score instead.

Category valuation routing

Listings are routed to a category-specific valuation service, because the attributes that drive price are different in each category. Routing uses the Marketplace category when available, and falls back to keyword detection on the title and description.

Valuation service per category
Service	Primary signals	Category labels routed here
Vehicles	Year, make, model, trim, mileage, title status, and condition wording. Title branding and dealer-listing patterns are weighted heavily.	17
Electronics	Exact model and generation, storage tier, carrier lock status, included accessories, and screen or battery condition wording.	15
Phones	Model and generation, storage, lock status, battery health when stated, and cosmetic condition.	11
Furniture	Brand, material, dimensions, damage wording, and pickup difficulty, which is a real cost in this category.	10
Appliances	Type, capacity, age, working status, and whether installation or removal is the buyer's problem.	9
Tools	Brand and battery platform, corded vs cordless, whether batteries and chargers are included, and working condition.	12
Sports	Type, size, brand tier, and wear, which drives most of the value spread.	19
Collectibles	Exact item identification, completeness, edition, and condition grading, where small differences move the price a lot.	16
Generic	Fallback for listings that do not route to a specialised service. Uses broader comparable pricing and produces a lower confidence level.	9

Risk signals checked

These are prompts to verify, not accusations against a seller. A flagged listing may be completely genuine.

Implausible discount: Asking price far below the estimated value, at a magnitude that historically indicates a non-genuine listing rather than a bargain.
Known scam price bands: Prices that fall in ranges commonly used in high-value electronics scam listings.
Down-payment and finance-lead wording: Listings where the advertised number is a deposit or monthly payment rather than the purchase price.
Parts-only wording presented as working: Repair, salvage, and 'for parts' language that contradicts the rest of the listing.
Thin or evasive descriptions: Missing the identifying detail a genuine seller would normally include for that category.
Seller account age: Very new seller accounts, where that data is available.
Seller trust score: Low aggregate trust signals on the seller, where available.

What DealFlipAI cannot do

Published deliberately. A methodology page that only lists strengths is marketing.

Valuations are estimates, not appraisals. Every estimate is published with a confidence level, and low confidence means exactly that.
DealFlipAI cannot physically inspect an item, verify ownership, confirm authenticity, or test that something powers on.
Marketplace listings are incomplete and change constantly. Sellers omit damage, reuse photos, mistype prices, and delete posts.
A risk flag is a prompt to verify, not an accusation. Estate sales, moving sales, and motivated sellers create genuine bargains that look statistically odd.
Accuracy is weakest where comparable pricing is thin, the category is unusual, or the seller wrote a vague title.
Scores do not account for your travel time, storage, cash position, or how fast you personally can resell. Those are yours to weigh.

Metro coverage

The scanner has built-in coverage for these 37 US metro areas. Custom searches can also be run against other locations, with a radius you set.

Atlanta
Austin
Baltimore
Boston
Charlotte
Chicago
Cleveland
Columbus
Dallas
Denver
Detroit
Houston
Indianapolis
Kansas City
Las Vegas
Los Angeles
Memphis
Miami
Milwaukee
Minneapolis
Nashville
New Orleans
New York
Orlando
Philadelphia
Phoenix
Portland
Sacramento
Salt Lake City
San Antonio
San Diego
San Francisco
San Jose
Seattle
St Louis
Tampa
Washington DC

Not seeing your area? Custom location searches are not limited to this list — these are the metros with pre-configured coordinates.

Methodology questions

How does DealFlipAI calculate a deal score?

The score is a weighted blend of five factors: price versus estimated resale value (50%), condition signals (15%), listing quality (15%), category-specific factors (10%), and valuation confidence (10%). Risk penalties are then applied, and some listings are filtered out entirely before scoring.

Why does a huge discount not automatically produce a high score?

Because beyond a certain point a large discount stops indicating a bargain and starts indicating a problem — a misdescribed item, a down-payment listing, or a scam. Past that threshold the discount factor stops increasing the score and the risk checks start reducing it.

Where does DealFlipAI's valuation data come from?

Listings are routed to a category-specific valuation service that weighs the attributes that actually drive price in that category, compared against resale context for similar items. Confidence reflects how cleanly the item could be identified from the listing.

Which cities does DealFlipAI cover?

The scanner has built-in coverage for 37 US metro areas, and custom searches can be run against other locations. The full metro list is published on this page.

Can DealFlipAI guarantee a profit?

No. Scores rank opportunity based on information available in the listing. Actual profit depends on real condition, negotiation, fees, repair costs, travel, storage, and the final sale price. DealFlipAI publishes confidence levels and limitations rather than presenting estimates as certainties.