diff --git a/docs/finance/revenue-model.md b/docs/finance/revenue-model.md
index afc6a9e..4dd3241 100644
--- a/docs/finance/revenue-model.md
+++ b/docs/finance/revenue-model.md
@@ -519,6 +519,29 @@ What it changes:
   serve two ResurRec plants; if it does, ResurRec has its own metal path
   and the stamper-supply offset shrinks.
 
+**Against CNX's launch plant** (one Duplex system and one Classic, figures
+from [scale](../scale.md#production-capacity); same 60 to 70 %
+effectiveness and 300 days; estimates):
+
+| | ResurRec Japan (3 AD12, as photographed) | CNX launch plant (2 Duplex presses + 1 Classic) |
+|---|---|---|
+| Rated, records/h | ~415 to 450 (24 to 26 s cycle) | ~248 to 288 (2 × 60 to 80 + 128) |
+| Effective, records/h | ~250 to 315 | ~150 to 200 |
+| One 8-hour shift, records/day | ~2,000 to 2,500 | ~1,200 to 1,600 |
+| One shift, records/yr | ~0.60 to 0.76M | ~0.36 to 0.48M |
+| Two shifts (CNX's plan), records/yr | ~1.2 to 1.5M | ~0.71 to 0.97M at ~20 % specials, 0.58 to 0.90M at 60 % |
+| Press operators per shift | 1 (Pheenix: one per four presses) | 2 (one per Duplex cell, plus the Classic's attendant; our reading) |
+| Specials | splatter kit, semi-automatic mode, rate unpublished | Duplex with EX 35 colour extruder, 120 to 160/h rated |
+| Steam at full rate | ~500 to 540 kg/h (1.2 kg per cycle, vendor figure) | 450 to 540 kg/h |
+
+On black and solid colours the Japan floor out-produces the launch plant
+by about 1.6 times per hour, so its one shift matches CNX's two. On the
+60 % specials CNX plans for, the comparison turns on the AD12's
+splatter-mode rate, which is the number to ask Pheenix for. The steam
+load is about the same, so per record the AD12s use about a third less
+steam (~1.2 kg against ~1.8 to 1.9 kg at rated output; vendor figures and
+our sizing).
+
 Open questions: where the Japan plant is, since when, how many presses
 and which, whether it plates, and whether Bangkok keeps its orders or
 volume work moves to Japan.