Deals: where the edge actually is

Research brief · 18 Sep 2026 · companion files: aggregators.md, aggregators.csv, deal-heuristic.md

One-line answer: don't build or hunt for one aggregator. Amazon is already solved by price-history trackers, Marketplace and Craigslist are a speed problem you can rent a fix for at $5–10 a month, and the only real gap is NYC tickets, where nobody joins up Theatr, CrowdVolt, DICE, RA, TKTS and rush.

The read

"Best deals across Facebook Marketplace, Craigslist and Amazon" sounds like one problem. It is three, because the price gets set three different ways, and what counts as a good deal changes with it.

On Amazon the question is whether the discount is real. Usually it isn't. Keepa and CamelCamelCamel answer that for free, and Slickdeals pays editors to check price history before anything hits its front page. There is nothing to add.

On Marketplace and Craigslist the question is whether you see the listing before someone else does. Facebook's own saved-search alerts are reported to lag by two to four hours. Underpriced goods in New York don't last that long. Cross-site search engines like SearchTempest don't fix this; minute-level alert tools do.

On tickets the question is what the seat is worth right now, because it's worth zero at curtain. Theatr and CrowdVolt are both good, for different reasons. Theatr caps price at what the seller paid. CrowdVolt lets you post a bid and wait. Neither tells you whether a cheaper route to the same seat exists tonight.

What is actually going on

MechanismVenuesControlling constraintReference priceYour edge
Retail with historyAmazon, open-box, refurbFake anchors90-day median and all-time lowPatience plus an alert
P2P localFB Marketplace, Craigslist, OfferUp, NextdoorLatency, then trusteBay sold compsBeing first with comps already done
PerishableTheatr, CrowdVolt, DICE, RA, TKTS, rush, GametimeTime decay vs sell-out riskCheapest listed alternative nowStanding bids and preset price alerts

The hidden structure: every good deal is a gap against a reference price you trust, and every bad deal is a gap against a reference the seller picked. List price, face value and "retails for" are all seller-picked.

Evidence map

Known · source of record

Specialist and first-person · suggests

Critical lane

Unknown

The heuristic, short form

Significant = cost far enough below a trusted reference that it survives hassle and risk, on something you already wanted.

  1. Vetoes first: Zelle or deposit to a stranger, off-platform links, unescrowed tickets, too-cheap phones from new accounts.
  2. Reference price R: 90-day median (retail), eBay sold comps (P2P), cheapest listed alternative right now (tickets). Never list price, never face value.
  3. Cost C: price, shipping, travel time at $25 an hour, fix-up. Buyer fees are left out on purpose.
  4. D = (R − C) / R, cut for weak comps and for the chance of losing the money.
  5. Act-now bars: retail 20% and at the all-time low. P2P 40%. Tickets 30%, or anything under the alternative on a sold-out show.
  6. Gates: saves at least $25 and twice your hassle time; was on your list beforehand.

Full version with eight worked examples is in deal-heuristic.md. The thresholds are my inference. The structure is borrowed from how SeatGeek, Slickdeals editors and flippers already do it.

Counterargument

The best serious objection: for a personal buyer, all of this is more work than it saves. If you buy a handful of things a year, Slickdeals alerts plus TodayTix rush gets you 80% of the value, and a scoring system mostly makes you shop more. The want-list gate is there because of this. If after a month the log shows you buying things that weren't on the list, drop the system and keep the alerts.

Second objection: Theatr's "below price paid" is already a good enough rule. It isn't, on nights when TKTS or rush has the same show. It is on sold-out shows.

What I would do

  1. Retail: install Keepa, set drop alerts at your act-now price for the five things you want. Add Slickdeals keyword alerts. Done, free.
  2. P2P: one Flipify watchlist for the single category you care most about, comps run in advance. Keep native Craigslist and Nextdoor alerts for the rest. Don't pay Swoopa prices unless you're reselling.
  3. Tickets: this is where to spend attention. Theatr price-point alerts per show. Standing CrowdVolt bids at your act-now price. Join DICE and RA waitlists the day a show sells out. Check TKTS and TickPick before accepting any Theatr alert.
  4. If you build: build the ticket layer only. A small NYC feed that takes a show or event, pulls the cheapest route across those venues, and scores it. Don't scrape Marketplace; Meta's terms forbid it and the paid tools already carry that risk for you.

What would change my mind

Next move

Time box: 30 minutes this week. Set three alerts: one Keepa, one Theatr price-point, one CrowdVolt bid. Start the decision log from the bottom of deal-heuristic.md.

Success: three alerts live, first row logged. After 20 rows, tune the thresholds or drop them.

Source trail