diff --git a/references/worked_example.md b/references/worked_example.md
new file mode 100644
index 0000000..ee278f8
--- /dev/null
+++ b/references/worked_example.md
@@ -0,0 +1,94 @@
+# Worked example — John Wayne Airport Shuttle (S01027 × Passent JWA001-071425)
+
+Use this to check your own output. If your numbers for this deal differ from these,
+something was misread — usually a discount, a section header counted as a line, or a
+recurring line classified as one-time.
+
+## Inputs
+
+**Ad proposal** (Passent Media, 14-Jul-2026): 64,000 monthly on-boardings, 6-minute average
+journey, 24 fifteen-second slots, 1,536,000 monthly impressions at an assumed 65% blended
+fill. Three revenue lines — TransitAds.co local at $12,000/mo, programmatic ad-floor at
+$2,765/mo ($12.00 CPM, 15% fill), national direct buy at $7,373/mo ($24.00 CPM, 20% fill).
+**Gross total $22,138/mo, $265,651/yr.** A 2028 LA Games block worth $334,200 over 2.5
+months is captured but disabled.
+
+**Odoo S01027**: 17 money lines across Hardware, Installation, Professional Services,
+Licensing & Support, and Optional Support. Quote total $84,334.45.
+
+**Pairing check**: the proposal's PROJECT/REF is "SNA JOHN WAYNE ARPT" and S01027 is billed
+to John Wayne Airport-Orange County — same agency, same site. The proposal's impression math
+assumes eleven onboard screens; S01027 carries eleven M27V vehicle displays. Both documents
+describe an onboard/mobile deployment. This pair reconciles, so the model may be built. These
+revenue figures belong to this deployment alone and must not be carried to any other quote.
+
+## Classification
+
+Four lines are recurring: `LIC-MPTV-Pro-Base` ($5,489/yr, $2,189/yr cost),
+`API - Tripshot` ($1,125/yr, $0 cost), `LIC-ADV-FM` ($0 — the vehicle for the revenue share
+itself, not a billable line), and `SUP-MP-FM` ($7,150/yr, $2,200/yr cost). Everything else
+is one-time.
+
+MPTV Professional at $199/unit/year across 11 units is $182.42/month — the figure Andy
+thinks of as "roughly $20 per controller per month."
+
+Two lines carry zero margin legitimately: `DBE-Dispatch` ($4,000) and `PS-Travel` ($4,000)
+are pass-throughs at full cost. Several hardware and cabling lines carry 100% margin
+because no standard cost is set in Odoo — worth flagging, since it makes COGS optimistic.
+
+## Expected outputs
+
+| Figure | Value |
+|---|---|
+| One-time COGS target (yellow) | $45,365.01 |
+| One-time sale price target (orange) | $70,570.45 |
+| One-time MSRP target (green) | $87,857.00 |
+| Recurring lines, annual net | $13,764.00 |
+| Recurring cost carried monthly | $365.75/mo |
+| Blended discount on one-time scope | 19.7% |
+
+Cross-check: $70,570.45 one-time + $13,764.00 recurring = $84,334.45, the printed quote
+total.
+
+With a 12-month ramp starting at 25%, a projection start of 2027-02, and 12% money:
+
+**Model 2A — MPM financed (MPM 30 / Passent 30 / Reseller 10 / End User 30)**
+
+Principal $45,365.01 at 12% APR, retired in month 13 with $3,888.81 of interest paid. No
+negative amortization: MPM's month-1 share of $1,660 covers the $454 of first-month
+interest comfortably.
+
+| Milestone | Target | Reached |
+|---|---|---|
+| COGS | $45,365.01 | month 13 |
+| Sale price | $70,570.45 | month 17 |
+| MSRP | $87,857.00 | month 20 |
+
+60-month distribution: MPM $368,598, Passent $368,598, Reseller $122,866, End User $368,598.
+
+**Model 2B — same split, revenue only.** Identical distribution figures, no targets, no cost
+basis sheet. The check that matters is that nothing in this file discloses margin.
+
+**Model 2C — customer financed (MPM 5 / Reseller 2.5 / Passent 22.5 / End User 70)**
+
+60-month distribution: MPM $61,433, Reseller $30,716, Passent $276,448, End User $860,061.
+The End User's 70% share, net of the $1,147/month they pay for licensing at invoiced price,
+recovers the $87,857 MSRP they paid in **month 11**.
+
+## How to read it
+
+Both structures work on this deal, and the interesting comparison is what MPM trades away.
+Financing costs $45,365 of cash for thirteen months plus $3,889 of interest, and returns
+$368,598 over five years. Selling outright books the hardware immediately and returns
+$61,433 — a $307,165 difference for capital that is fully recovered by month 13. On these
+assumptions financing is clearly the better use of the money, which is precisely why the
+assumptions deserve scrutiny rather than celebration.
+
+Everything rests on a 65% fill rate the proposal itself warns may not hold early. Halving
+`monthly_gross` is the stress test worth running: COGS recovery slips from month 13 to 23,
+MSRP from month 20 to 38, and interest paid rises from $3,889 to $6,696. More telling is the
+first month, where MPM's net of $464 barely clears the $454 of accruing interest. The deal
+does not tip into negative amortization, but it comes within ten dollars a month of doing
+so — meaning a slower ramp, a deeper fill miss, or a higher rate would. If a deal ever does
+tip, the script flags it in red on the Summary; treat that as a reason to renegotiate the
+share or the rate, not a cosmetic warning.