Controlled synthetic-buyer audits

See the price you were never meant to compare.

Paste a URL. Twenty-four synthetic buyers check it from every angle — geography, device, cookies, referrer, language — and return evidence, with statistics and receipts.

Public-web only · Sample audit completes in ~60 seconds

sample audit  UA182 · JFK → LHRbaseline  $498highest  $642spread  +$144 · +29%driver  location · 62%
01 ]The problemsame url · different price

Same URL.
Different buyer.
Different price.

Location, device, language, cookies, referrer, and session quietly change what a server decides to charge. It is real, widespread, and — done by hand — almost impossible to prove.

Android · Chrome
rural Iowa · VPN
$498
USD 498.00 · baseline
Δ+$144+29%
iPhone · Safari
Manhattan · direct
$642
USD 642.00 · exposed

same seat · same flight · same date — measured, not inferred

02 ]The mechanism∂price / ∂vector
Controlled experiment

Vary one vector. Hold the rest constant.

Jacobi changes a single buyer-context dimension at a time against a fixed baseline, so any price movement can be attributed — not guessed. The Jacobian isolates the driver.

Six vectors — geography, device, browser language, cookies, referrer, and session — are tested in controlled waves. Only differences that clear a Welch t-test are reported as evidence.

Buyer-context matrix6 vectors · 1 varied
VectorState∂P / ∂V
Geography
ip-resolved city
varied
62%
Device
user-agent · platform
held
21%
Cookies
aged vs fresh
held
10%
Referrer
direct vs comparison
held
7%
Browser language
Accept-Language
held
4%
Session
visit history
held
3%
One vector varies; the rest are held constant. Only gaps that clear a Welch t-test are reported.
03 ]The evidencereceipts · hash-sealed
Raw evidence

Receipts a regulator can read.

Every probe stores the on-page price, native currency, request context, and timestamp — hash-sealed into one reproducible audit trail.

Audit ReceiptJACOBI · LOG #A7F2-22 · v3
targetua182.example-air.com/booking · UA182 · JFK → LHRcaptured2026-06-27 14:02:11 UTCprobes24 synthetic buyers · 6 vectors · 3 network tiers · public-web only
#LocationIP typeAgentNativeUSDVerdict
01US · ManhattanRESIDENTIALiOS · SafariUSD 642.00$642DEVIANT
02AE · DubaiMOBILEiOS · SafariAED 2,356.0$641DEVIANT
03JP · TokyoRESIDENTIALmacOS · SafariJPY 91,200$612NORMAL
04GB · LondonRESIDENTIALWin · EdgeGBP 472.00$596NORMAL
05IN · BangaloreDATACENTERWin · FirefoxINR 42,600$512NORMAL
06US · Rural IowaDATACENTERAndroid · ChromeUSD 498.00$498CLEAN
Price Exploitation Index
1.29
Observed spread
+$144
Dominant driver
location
Progressive
sha256: 9f2c1ab7e4d0c8b3a611f0e2…7d54 · sealed · reproducible
04 ]The verdictbaseline → exposed
At a glance

The whole audit, in one readout.

Baseline, highest observed, the delta, statistical confidence, and the dominant driver — resolved into a single readout you can hand off or export as an audit-ready report.

ua182.example-air.com/booking · UA182 · JFK → LHR Complete · 24 / 24
Baseline
$498
Highest
$642
Delta
+$144
PEI
1.29
iPhone · Manhattan · directtop
$642
Safari · Tokyo · direct
$612
Edge · London · direct
$596
Firefox · Bangalore · VPN
$512
Android · rural Iowa · VPNbaseline
$498
Progressive
+$144
29% over baseline · per ticket
Location
62%
Device
21%
Cookies
10%
Referrer
7%
Welch t-testp < 0.01
95% CI[ +$96, +$192 ]
ResultSIGNIFICANT
05 ]Why it holds uppublic-web only
Defensibility

Built to be defended.

01

Native-currency capture

Prices recorded in the real on-page currency first, with a USD-normalized comparison alongside.

02

Timestamped evidence

Each observation is time-stamped and ordered, so any result can be reproduced and defended later.

03

Real vs inferred, separated

Directly observed agents are never mixed with inferred ones. Inferred values never appear as raw evidence.

04

Controlled probe accounting

Agent counts, network tiers, and blocked requests are reported in full — no silent gaps in the sample.

05

Public-web only

Jacobi reads exactly what any buyer could see. No logins, no private data, nothing behind a wall.

06

Audit-ready export

Summary, figures, and a complete evidence appendix, sealed into a research-grade PDF.

Built forCompliance / Pricing / MAP & brand protection / Market intelligence
06 ]Run an auditone url · ~60s

Stop guessing. Start auditing.

24 synthetic buyers · 6 context vectors · one URL