GLP-1 Cravings Protocol: Evening Snacking Dropped 8/10 → 2/10 in 11 Weeks

A 42-year-old marketing director in London was three months into semaglutide when the appetite reduction stopped. Not everywhere — just after 8:30pm, when the drug's satiety effect apparently clocked off with her. Evening snacking was running at a daily 8/10 by her own log. In 11 weeks it dropped to a twice-weekly 2/10. The lever wasn't discipline. It was mapping the shape of the craving.

The presenting state

Daytime appetite was suppressed as advertised. Evenings were a different story: predictable 9pm carbohydrate pull, mild irritability by 10, and by 10:30 either surrender or three hours of chewing gum. Weight loss had plateaued at week 10. She was calling it "willpower failure." It wasn't.

GLP-1s primarily suppress homeostatic hunger. They are far less effective on hedonic eating — food seeking driven by reward, stress and habit (Farr et al., Mol Metab 2016). The evening pattern was pure hedonic, timed to a cortisol dip and a specific emotional cue (end-of-workday decompression). The drug was never going to touch it.

The protocol

Nothing was banned. Everything was logged — because a craving you can see the shape of is one you can pre-empt, and a craving you white-knuckle is one that wins.

Craving map — every evening pull logged in the Journal with: time, intensity 0–10, food type, preceding emotion, preceding activity. Four weeks of data, no interventions.
Pattern surfaced by week 3 — 82% of episodes clustered 8:45–9:30pm, 71% followed a Slack close or a difficult meeting recap in her head, and 88% went to salt-fat combinations, not sweet.
Pre-empt sequence at 8:15pm — 400 ml water with electrolytes, 15 g whey protein, 10 minutes of the Soft Evening Unwind practice from the Kokorology library.
Protein floor 1.6 g/kg distributed to a 4pm meal, not left to a 7pm dinner — front-loaded satiety survives the 9pm dip (Leidy et al., 2015).
Salt-fat swap on standby — a small tin of olives + hard cheese available in the kitchen when the pre-empt didn't hold. Two of these a week were built into the plan, not treated as failure.
Sleep window 22:30 — protected. Late-night snacking is the tell of a delayed circadian sleep window (Baron et al., Obesity 2011).

What changed

Evening craving average: 8.2/10 → 2.1/10 by week 11.
Frequency: daily → twice weekly.
Weight loss un-plateaued: −1.4 kg across weeks 8–11 after being flat for three.
Sleep onset latency: 31 min → 14 min (Journal sleep log).
Self-reported "willpower failure" language: gone from her weekly check-ins by week 6.

Why this worked

The craving didn't get smaller — it got understood. Once she could see it was a 9pm cortisol/emotion signature, not stomach hunger, the response stopped being moral (bad, weak, ate again) and became procedural (pre-empt at 8:15). Hedonic eating drops when it's met with a specific pattern interruption at the specific trigger point, not with a general instruction to resist (Boswell & Kober, Obes Rev 2016).

The drug did the daytime. The protocol did the evening.

Sources

Farr et al., 2016 — GLP-1 and hedonic vs homeostatic eating, Mol Metab.
Boswell & Kober, 2016 — Food cue reactivity and eating behaviour, Obes Rev.
Leidy et al., 2015 — Protein distribution and satiety, Am J Clin Nutr.
Baron et al., 2011 — Circadian timing and late eating, Obesity.