The Pre-Print Cohort Read
This morning we published CELH's live cohort: 11 CELH prints + adjacent high-growth beverage analogs, 5-day forward median +0.3%, IQR [-11.4%, +14.7%]. The bimodal distribution and wide IQR were the headline — Celsius is one of the highest-variance earnings prints in the consumer staples universe.
The dominant feature in the cohort was Pepsi-channel reorder commentary: accelerating reorders + stable retail velocity → 5-day +18.4% (n=4, hit-rate 100%); inventory destock + flat velocity → 5-day -14.1% (n=4, hit-rate 0%).
What Actually Printed
Q1 revenue of $782.6M, up 138% YoY — a record quarterly result. Adjusted EPS of $0.41 beat consensus of $0.30 by $0.11 (a 36.7% surprise). Alani Nu, finalized in April 2025, contributed roughly $368M to the quarter — about half of total sales. Rockstar Energy, acquired August 2025, also contributed materially.
Stock reaction: ranged from +2.5% to +10% in premarket trading; settled around +6% intraday. Outsized magnitude relative to the cohort central tendency, but inside the upper IQR (+14.7%).
What The Cohort Got Right And What It Missed
The cohort's IQR called the magnitude correctly — +6% premarket landed cleanly inside the +14.7% upper-IQR bound. The cohort tightness score (low) correctly flagged that pre-print shape wasn't decisive, and that the print itself would carry the signal.
What the cohort didn't capture: the M&A integration revenue contribution from Alani Nu and Rockstar is a structural feature that didn't exist in the pre-acquisition CELH cohort. The pre-print feature attribution focused on Pepsi-channel velocity (the right feature for legacy CELH) but the actual print outcome was dominated by inorganic revenue. The cohort's feature_importance correctly identified that the print would matter; it didn't have a label for 'newly acquired sub-brand revenue contribution.'
- Pre-print 5d cohort median: +0.3%; IQR upper: +14.7%
- Actual reaction: ~+6% intraday, inside IQR
- Dominant feature pre-print: Pepsi-channel velocity
- Actual driver: Alani Nu + Rockstar acquisition revenue contribution
The Calibration Lesson
When a company has just made a transformative acquisition, the pre-print cohort doesn't have analog matches that capture the new revenue mix. The fix is to re-anchor the cohort post-print, which lets the embedding learn the new chart shape that includes the acquisition's contribution. After 2-3 quarters of post-acquisition prints, the cohort's matches will start including other consumer roll-up stories (KDP / MNST mid-2010s era), and the feature attribution will sharpen.
For agent systems trading post-M&A names: the cohort is less reliable for the first 2-3 prints after a major deal closes. Treat the conformal band as wider than displayed and bias toward smaller position sizes during the integration window.