nx_ad_h2h.nx

nx_ad_h2h.nx -- LIB: MEASURED HEAD-TO-HEAD vs representative REAL incumbent baselines -- the evidence the census requires for an EXCEEDS grade (the "beats" bit), computed by the organ, NEVER self-scored. Each comparator measures OUR value and a real incumbent baseline on a NAMED, objective, lower-is- better axis and returns 1 ONLY if we strictly beat it. HONEST RESTRAINT (feedback-no-wave-measured-exceed + feedback-show-raw-output-no-overclaim): there is DELIBERATELY no comparator for viewability (incumbents measure it too, e.g. Google Active View) or bot-detection DEPTH (incumbents do SIVT via the per-user fingerprinting we REFUSE) -- those axes stay MEASURED, not EXCEEDS. We only claim a beat where it is structural and objective. Cited incumbent baselines: DELIVERY standard programmatic display loads a third-party <script src> (Google Publisher Tag, securepubads.g.doubleclick.net/tag/js/gpt.js) [srch_adfraud.raw] TRANSPARENCY programmatic supply has reseller hops -- ads.txt exists to authorize them (DIRECT vs RESELLER); a DIRECT single-sponsor model references zero external seller domains [srch_adstxt.raw] SIGNUP CPM = "cost per thousand impressions": billed per impression regardless of outcome; ours bills $0 on a miss (nx_ad_bill returns 0 for an UNDER verdict) [srch_cpm.raw] license_tier: ORIGINAL