Headlines Analytics Console
Per-publication: top-performing topics
Per-pub × topic × lift

Top topics by lift, per publication

For each publication, the topics that exceed the publication's median by the largest margin. Lift = (topic median − pub median) / pub median.
PubTop topicnLift Verdict
News Miami Herald Politics120 +34% p<0.05
News Miami Herald Crime95 −1% no diff
News Miami Herald Sports—Local78 −50% p<0.01
L&E US Weekly Celebrity Health32 +110% p<0.001
L&E US Weekly Royals18 +32% p<0.10
L&E Woman's World Recipes—Comfort41 +78% p<0.05
Per-publication: top-performing headline formulas
Per-pub × formula × lift

Top formulas by lift, per publication

For each publication, the headline formulas that exceed the publication's median PVs by the largest margin. Canonical formula list lives in formula_taxonomy.py (shared by classifier + grader + variant Worker). Baseline: the publication's overall median across all headlines.
PubFormulanLift Verdict
Per-vertical: which formulas + sites work best
Mind/Body Everyday Living Experiences · 3 verticals × 8 formulas × 8 sites

Vertical × formula × site—the cube

Mind/Body n=389 · median lift +28%
Number-lead × Woman's World
+85% · n=28 · p<0.001
Curiosity-gap × US Weekly
+42% · n=19 · p<0.05
What-to-know × Miami Herald
+12% · n=22 · p<0.10
Everyday Living n=510 · median lift +14%
Number-lead × Woman's World
+62% · n=41 · p<0.001
Direct-declarative × KC Star
+9% · n=37 · no diff
Experiences n=287 · median lift +6%
Question × Sacramento Bee
+38% · n=14 · p<0.10
"Here's" × US Weekly
+11% · n=18 · no diff
Trends over time, per publication
Quarterly median PVs by publication

Per-publication PV trajectory

Median page-views per publication, broken into calendar quarters. Trend = (latest quarter ÷ first quarter) − 1. Lets you see which pubs are climbing and which are slipping—not Apple News specifically, total reach across all distribution.
Pub Quarters Trend
Sweet-spot finder: pub × topic × formula
Three-way combinations that beat baseline · weather added as a topic

Pub × topic × formula combinations with measurable lift

Lift can hide inside an interaction even when the publication or topic alone looks flat. Example: politics at Miami Herald may sit at +0% overall, but politics + quote-lead at Miami Herald may run +60%—the formula unlocks that topic at that pub. This tile surfaces those three-way combos with n ≥ 5 so they don't go unassigned.
TopicPubFormulanLift Verdict
Bottom-Performer Patterns · what to avoid
3 anti-patterns · n=187 · p<0.01 each

Patterns underperforming consistently

Combinations whose median PVs sit ≤ 0.5× the publication median, with n ≥ 10. Use as a "what to avoid" reference at assignment time.
Anti-pattern Scope n Median PV ratio
Team Performance · per-author percentile vs publication median
Tracker-based · n=12 active writers

Author percentile rankings · last 30 days

WriterPubVerticalMedian PVsPub %ilen articles
RBMiami Herald48954.2%703

Names anonymized in brief; live site uses tracker AUTHOR field directly.

Data: Snowflake TRACKER_ENRICHED · Tarrow XLSX · the Tracker sheet · grader signals Docs data-keywords ↗