Pattern detection explained
Once you have 10+ published posts per channel, distro surfaces patterns automatically.
Individual post-mortems are per-post. Pattern detection is cross-post — it looks at your last 20+ published pieces and finds structural or topical clusters that correlate with high or low performance.
What counts as a "pattern"
Three types:
- Structural — hook type, length, presence/absence of numbers, use of first person, etc. "Your top 5 LinkedIn posts all open with a concrete number."
- Topical — recurring themes. "Pricing / billing posts consistently outperform feature announcements on X."
- Temporal — time-of-day or day-of-week effects. "Tuesday 8am LinkedIn posts average 2.1× your weekday median."
How many posts do you need?
Minimum 10 published pieces per channel for pattern detection to turn on. Below that, the signal is too noisy to say anything useful. "Pattern" status in the UI shows a progress bar until you hit the threshold.
Where patterns show up
- Today view — top 2 patterns get a line each in the morning brief
- Repurpose — the prompt includes the top 3 structural patterns for each platform
- Post-mortem — cross-references the pattern list when explaining a win or loss
False patterns
Small samples produce spurious patterns. We show a confidence level with each pattern (high / medium / low) based on sample size and effect size. Low-confidence patterns are dim and don't flow into downstream prompts. As data accumulates, patterns re-evaluate.