Keelway
Solution · Fraud Detection

Stop booking the chameleon carrier.

Every inbound quote gets scored against FMCSA authority, Highway identity, inbox-level signals, and your own carrier history — server-side, before the coordinator sees the row. Chameleon carriers, fake-MC rebrokers, phishing, and rate-con interception caught at the front door.

<600ms
Cold-MC fraud check, server-side
~3–7%
Of inbound quotes flagged, calibrated
>95%
Rate-con extraction accuracy
$700M+
Annual US freight fraud (reported)
Illustrative · industry estimates

Freight fraud doesn't look like the movies. It looks like a clean email from dispatch@coldchian-logistics.com — one letter off — quoting your Memphis-to-Dallas reefer at $1,650 when the lane usually runs $2,100. The MC resolves. The DOT looks fine. The insurance is on file. A coordinator at 4:47 PM Friday afternoon books it, the load disappears, and on Monday morning your shipper is on the phone asking where their freight is.

The signals were all there. The domain was registered eleven days ago. The MC was issued in March. The previous fourteen replies on the thread were on a different corporate domain. The lane rate was two standard deviations under the rolling average. Any one of those is noise. Together they're a fingerprint.

That's what Keelway's fraud detection is. Composite signal scoring, server-side, on every inbound quote — before the coordinator sees the row.

01 · The four patterns

What we catch — and how.

Chameleon carriers

A previously-bad MC rebrands under fresh authority — same operator, same dispatch number, sometimes the same address with a comma reordered. We match operator names, addresses, phone fingerprints, and dispatch-email domains across the revoked-and-replaced pool and surface the prior identity before you book.

Fake-MC rebrokers

Fresh authority (< 90 days), zero insurance history, bidding $200 below market on a premium spot lane that doesn't match the declared equipment or operating radius. Standalone, any one signal is noise. Together they're the signature of a rebroker — they don't have a truck; they're going to book it and re-broker it to a real (or fake) carrier.

Email-level phishing

Domain spoofing (cold-chain-logistics.com vs cold-chian-logistics.com), look-alike TLDs (.co vs .com), sudden signature drift on a 14-message thread, free-mail address replying to a thread that was on a corporate domain. We check SPF/DKIM/DMARC on every inbound and flag the drift.

Rate-con interception

A downstream party intercepts the rate confirmation and resends it with the remittance bank-account number swapped. We hash banking details against your prior pays to the same MC — when the routing/account on a rate con doesn't match the last six months of pays to that carrier, the rate con is held for review before settlement.

02 · How the score is built

Six signal sources. One composite score.

No single signal kills a quote. The model weights the composite, and the weighting calibrates per-brokerage from your override history.

03 · The 90-second pre-book check

What a coordinator sees, before they click accept.

Below is a real-shape inbox row for an inbound quote, with the trust score and underlying signal breakdown — the kind of row that lands in the coordinator's ranked list within 600 ms of the email hitting the inbox.

From: dispatch@coldchian-logistics.com · 4:47 PM
Trust 14 · Hard-flagged
Lane bid
MEM → DAL · Reefer · $1,650
MC reported
997412 (issued 2026-03-04 · 87 days)
Domain signal
Look-alike: differs from coldchain-logistics.com by 1 char
Thread drift
14 prior replies on coldchain-logistics.com
Lane rate vs history
−22% vs your 20-load rolling avg ($2,115)
Suggested action
Suppress · reply not surfaced in ranked list

Illustrative example using a synthetic MC + spoofed domain.

04 · The workflow

What happens when a quote gets flagged.

Soft-flag

Quote surfaces with red badge + reason

Coordinator sees the row with a one-line reason ('MC under 30 days old', 'signature differs from 14 prior threads'). Accept, decline, or override with typed reason. Override history feeds per-brokerage tuning.
Hard-block

Suppressed from the ranked list

Does not surface in the main ranked list. Tucked behind a 'Show suppressed (N)' affordance with the same signal breakdown one click away. Almost never overridden.
Banking alert

Rate-con + settlement instruction holds

A routing/account number on a rate con that doesn't match the last 6 months of pays to that MC fires a hold-for-review alert before the wire goes out.
Tenant learning

Per-brokerage score calibration

Every override (accept of a soft-flag, reject of a clean) feeds back into your tenant's score weighting. Model gets better at your book over the first two weeks.
Audit trail

Every score, every signal, every override logged

Logged and exportable. SOC 2 audit-friendly. If a shipper asks 'why did you book this carrier', the answer is one report away.
Network signal

Cross-brokerage chameleon propagation

Confirmed-fraud signals (anonymized) propagate to the rest of the Keelway network within minutes. Same chameleon hitting three brokers gets caught at brokers two and three.
05 · Compared to the alternatives

Why this is different from a Highway subscription, a Carrier411 lookup, or the 90-second check.

ApproachWhat it catchesWhat it missesCoordinator cost
Manual 90-second checkFMCSA authority, insurance, BASIC scoresEmail signals, chameleon match, rate anomaly~90s × every quote
Highway / Carrier411 standaloneCarrier identity, chameleon flags, monitoringPer-quote scoring, email signals, rate anomaly, integration with the ranked listmanual lookup per quote
Trust-and-prayNothingEverything~$5K–$50K per missed
Keelway Fraud DetectionComposite: FMCSA + Highway + email signals + cross-brokerage + internal history + rate anomalyPure-voice-channel fraud (a real human dispatcher calling the broker) — those still need a human ear~0s per quote

Vendor capability descriptions illustrative; verify direct before sizing.

Frequently asked questions

How is this different from your Carrier Trust Score?+

Trust Score is the output — a 0–100 number you see next to every inbound quote. Fraud Detection is the full pipeline that produces it: FMCSA authority + insurance + crash data, Highway identity signals, MC-DOT mismatch checks, domain and signature analysis of the email itself, chameleon-pattern detection across our cross-brokerage signal set, and your own internal carrier history. Trust Score is what you look at; Fraud Detection is what generates it. Same product line — different surface area.

What specific fraud patterns does it catch?+

Four families. (1) Chameleon carriers — an authority that recently rebranded from a previously-bad MC, including operator-name and address overlap. (2) Fake-MC rebrokers — a freshly minted authority bidding aggressively on premium spot lanes that don't match its declared equipment or operating radius. (3) Email-level phishing — domain spoofing, look-alike top-level domains, sudden signature drift on an established thread, replies coming from a free-mail address when the prior thread was on a corporate domain. (4) Rate-con interception — a downstream party intercepts a rate con and replaces the remittance instructions; we hash and check banking details against prior pays.

How fast does the check run?+

Server-side, before the coordinator sees the inbox row. Average end-to-end is under 600 ms for a fresh-cold MC (no cache, full FMCSA pull, Highway lookup, email-signal pass). Cached carriers re-score on every new quote in under 80 ms. The coordinator doesn't wait — they open the inbox and the ranked list already has scores attached.

What's the false-positive rate?+

On a calibrated book, ~3–7% of inbound quotes get flagged after the first two weeks. About 1.5% are hard-blocked (definite fraud — fake MC, domain spoof, hard chameleon match) and the rest are soft-flagged for coordinator review. Per-brokerage false-positive rate stays under 1.0% on the soft-flag tier — meaning fewer than one in a hundred coordinator-reviewed flags turns out to be a legitimate carrier with a quirky signature. We tune per-brokerage.

Do you integrate with Highway, Carrier411, MyCarrierPackets?+

Yes. Highway is our primary identity layer — every inbound email gets a Highway lookup if the MC resolves. Carrier411 (RMIS), MyCarrierPackets, Truckstop SmartWay, and SaferWatch are pulled where the broker has accounts; we federate the signals rather than replace them. If you don't have those subscriptions, Keelway's FMCSA-direct pull plus our internal chameleon-pattern model covers the bulk of the surface.

What happens after a quote is flagged?+

Soft-flagged quotes get a row in the ranked list with a red badge, a one-line reason ('MC under 30 days old', 'signature differs from 14 prior threads', 'authority address overlaps with revoked MC 4xxxxx'), and the underlying signal breakdown one click away. Hard-blocked quotes don't surface in the main list — they're behind a 'Show suppressed (3)' affordance with the same breakdown. Coordinators can override with a typed reason; overrides feed back into the per-brokerage tuning.

Is this just FMCSA data, or something more?+

FMCSA data is the floor, not the ceiling. The composite signal includes: FMCSA operating-authority status + insurance on file + crash history + BASIC scores + out-of-service rate; Highway identity if the broker has the integration; cross-brokerage chameleon patterns (we anonymize the inputs); email-level signals (domain age, MX record, SPF/DKIM/DMARC pass, signature drift, free-mail vs corporate sender); rate anomaly vs your own lane history; and your internal carrier history (have you booked this MC before, what was the on-time, any claims). Any single signal is rarely a kill — the model weights the composite.

What does freight fraud actually cost?+

Industry estimates put total freight fraud at $700M–$1B+ annually in the US — and that's just the reported piece. Per-incident: a single double-brokering event runs $5K–$50K in stolen freight value, plus 20–60 hours of coordinator time chasing the missing load, plus reputational damage with the shipper. A chameleon-carrier booking that results in stolen cargo regularly clears six figures. The break-even on prevention is one stopped incident per quarter for most mid-sized brokerages.

How does this interact with our existing carrier-onboarding process?+

Cleanly. If you onboard carriers through MyCarrierPackets or a similar packet system, Keelway treats packet-completed carriers as priors (higher base trust). Brand-new carriers without a packet still get scored on FMCSA + Highway + email signals and surface in the ranked list with appropriate confidence. The onboarding flow doesn't change — fraud detection runs alongside it on every inbound quote, not just first-time bookings.

Pricing?+

Bundled into the per-load workflow — the same $1 per load that covers email triage, ranking, and check-calls covers full fraud detection on every inbound quote against that load. No per-quote, per-flag, or per-MC charge. Highway / Carrier411 / RMIS subscriptions, if used, are passed through at cost. See the pricing page for the full breakdown.

Stop the next chameleon at the front door

Score every inbound quote — before the coordinator sees the row.

Book a demo

Related