DV360 Custom Bidding Skill (dv360-custom-bidding)

Validates, simulates, calibrates, and compiles Google Display & Video 360 (DV360) Custom Bidding scripts and multi-objective funnel algorithms:

• dv360-bidkit (Open-Core): Fast AST-based linter for Google Custom Bidding grammar, 2026 signal registry verification, pre-flight simulation without eval()/exec(), Google S2 readiness checks (500 advertiser / 50 line item positives), deterministic RuleSet compilation, and customer inventory policy management.
• dv360-bidkit-pro (Commercial): Display Active View quality proxy fitting via automated Empirical Bayes shrinkage, client funnel evaluation and mass-balanced optimal weighting (bidkit-pro weights), multi-objective Pareto trade-off scenario grids, customer inventory impact audits, and complete production campaign bundle export (script.py, rules.json, configuration.json, calibration.csv, diagnostics.json, audit.html, manifest.json).

[!IMPORTANT] Division of Responsibilities:

• The Engine (dv360_bidkit / dv360_bidkit_pro) executes all calculations: Never write custom AST parsers, heuristic viewability formulas, or custom evaluation loops. Always invoke lint_script(), compile_script(), evaluate_script(), fit_display_quality_proxy(), or evaluate_and_calibrate_funnel_weights().
• The Agent formats inputs and guides decisions: Inspect client reports, define column mappings, run pre-flight simulations, audit proxy signals, check readiness diagnostics, and guide traders on Pareto trade-offs.
• Model Training vs. Targeting Distinction: In DV360 custom bidding, return None excludes an impression from machine-learning model training; it does NOT block the ad from serving. Blocking requires line-item targeting lists. Never claim custom bidding scripts enforce publisher blacklists.

Pro Licensing & Availability