Blog
Notes from the team building SignalGrid.
Modeling decisions, engineering tradeoffs, and the occasional unpopular opinion about how this category should work.
- 9 min readDraft
Why we score 40 KPIs per film, not 200 — calibration-honesty in catalogue analysis
Our 200-parameter catalog measures screenplays. Our catalogue scores films we have shipped and seen the receipts for. Conflating the two is the easiest way to lie with confidence.
Priya Ramanathan · Head of ModelingRead - 7 min readDraft
The case for cross-vertical entity graphs
Most signal-intelligence tools resolve entities inside their own vertical and stop. We argue the interesting predictions live across the seams.
Priya Ramanathan · Head of ModelingRead - 6 min readDraft
Why we don't trust survey-based Q-scores
Annual phone surveys of a few hundred respondents cannot keep up with how fast public attention now moves. Here is what we use instead.
Daniel Okafor · Lead, Celebrity EquityRead - 9 min readDraft
How we built the kynetranews ingest layer
Forty-nine RSS feeds, six social platforms, four search products, and one knowledge graph. Notes on what we learned shipping it.
Marcus Yi · Engineering Lead, IngestRead - 8 min readDraft
Calibration matters more than accuracy in flop prediction
A 90 percent accurate flop predictor that is over-confident is worse than an 80 percent accurate one that knows when it does not know.
Priya Ramanathan · Head of ModelingRead