A map of the vendor and data-layer architecture that grew up around one structural fact: streaming platforms and Big Tech gatekeepers increasingly withhold granular subscriber, viewership, and engagement data, so studios, advertisers, and analysts pay a stack of specialized vendors to reconstruct ground truth. It documents market structure; it does not make recommendations.
Point-in-time rendering. Source documents written July 6 to July 14, 2026; eight-layer integration pass July 8, 2026.
Layer 0 sits apart as the boundary condition: not a vendor layer but the reason the seven below exist. The vendor stack reads top to bottom, from the billing relationship itself down to the identity substrate every other layer matches through. Where the underlying map records a disagreement it could not resolve, the open question is shown in place rather than silently settled. Click any layer to expand it.
Eight platforms were rostered above the ecosystem as gatekeepers: each withholds some piece of viewership, subscriber, or engagement telemetry that the seven vendor layers below exist to reconstruct, and several run their own clean rooms monetizing the very data they withhold from studios and licensors. The withholding takes two forms: Amazon, Apple, and Roku withhold granular economics and monetize the data themselves, while Fox, Google/YouTube, and Disney control the rights and carriage architecture that creates the opacity. The layer's deep-research pass reclassified Paramount Skydance/WBD as a vendor subject rather than a gatekeeper, and Fox agreed in June 2026 to acquire Roku outright, so the active gatekeeper set is effectively seven, converging on a combined Fox-Roku entity by H1 2027. Netflix, long the acknowledged exception, is converging with the pack as its ad tier scales. This layer is not a peer of the seven below; it is the demand source they respond to.
The flow documented in this layer is predominantly one-way: gatekeepers withhold or control, and the vendor layers respond. What flows back from vendors to gatekeepers (payments, data licenses, measurement contracts) is comparatively undocumented in the source passes and remains an open question.
The billing relationship itself
The ground-truth stratum of the ecosystem: built from aggregated and anonymized card and bank feeds and digital invoice tracking, it isolates the billing relationship itself. Four derived metrics carry the layer's institutional value: cohort survival curves, plan-mix migration, cross-platform wallet mapping, and serial-churner tracking. Its distinguishing legal exposure is financial-transaction consent rather than device consent, which places it directly in the path of CFPB Section 1033 and California's Delete Act.
The map records the Disney+/Hulu/Max bundle case study, a Wallet-layer retention measurement, with non-overlapping numbers: one reading gives the pricing architecture and a 38% discount versus taking the services separately, the other a 59% twelve-month survival rate for 2024 sign-ups, roughly 28 percentage points above the standalone services. Neither source gives both numbers; they are held as two separate claims pending verification, not one reconciled data point.
The hardware and telemetry substrate of streaming measurement: chipset-level Automatic Content Recognition embedded in smart TVs and set-top boxes, capturing screen-level viewership regardless of whether the source is linear broadcast, a CTV app, a DVR, or a direct-to-consumer streamer. The data is the raw material of transactable TV-advertising currency, and its consent regime is device-level opt-in, a separate regime from the Wallet layer, with the Video Privacy Protection Act as the dominant statutory exposure. Attention verification (EDO, TVision) is folded in here as a narrow niche riding directly on ACR infrastructure rather than a structurally distinct layer.
Nielsen's 2025-2026 crisis is told two different ways in the source passes: one adversarial and scandal-framed (a delayed Gauge report and a Video Advertising Bureau accusation of "indefensible manipulation" ahead of the 2026 upfronts), one procedural (MRC priority fixes, a customer-requested delay to August 31, 2026, accreditation reaffirmed with the 2025 audit left open). Both cite the same underlying fact, adults 25-54 impressions down roughly 10% year over year. The map holds them as two partial views of one ongoing saga pending primary-source reconciliation; the layer file reads them as two tracks of a single story.
Independently verifies that programmatic advertising actually reaches a human, in a suitable environment, without fraud: not who watched, but whether the ad-serve was real. It stands between advertisers and the publishers and platforms selling them inventory, and its non-negotiable value proposition is independence, since a platform paid to transact media cannot credibly grade that media's quality. Following Integral Ad Science's take-private, DoubleVerify is the sole remaining public proxy in the layer.
These vendors hold no authorized platform telemetry or hardware-level ACR: they ingest social media activity and sentiment, search query velocity, piracy downloads, and fan wiki engagement, then harmonize those touchpoints into synthetic demand indices that proxy actual viewership. The core financial metric is the implied cost-per-demand-point: an estimated production budget divided by a title's registered demand-expression footprint. As platforms withhold precise metrics, these indices function as the primary quantitative baseline for carriage licensing, contract renewals, and catalog valuation.
On a specific, named, post-2025 deal or dispute where demand data directly moved a backend compensation outcome, the source passes fall back on historical precedent only (the Tommy Lee Jones v. William Morris Agency arbitration); contemporary agency negotiations remain closely guarded, and this stays the layer's one genuinely open question.
Sits at the top of the streaming conversion funnel, upstream of every transaction: search navigation, watchlist additions, and app-store download velocity, all captured before any billing relationship exists. Its analytical value peaks during tentpole launch windows, when tracking upstream intent against ad spend lets analysts reverse-engineer a platform's customer acquisition cost efficiency well before quarterly disclosure. Historically the least documented segment of the map, largely because its vendors present as consumer utilities while their actual enterprise architectures are B2B data licensing and advertising engines.
The sub-vertical handling what standard SVOD transaction panels cannot: synchronous low-latency delivery, betting-handle correlation, and regional blackout compliance at scale. It is structurally unique within the stack because it runs on exclusive, multi-year league data licensing rather than consumer transaction consent, which is why its vendors hold pricing power no panel vendor has. As broadcast rights costs escalate, the data layer is consolidating into fewer, deeper, longer deals, increasingly laced with direct league equity in the vendors.
The NBA-Sportradar rights end date is carried two ways in the corpus: two research passes independently find the deal running through the 2030-31 season, while earlier project framing says 2032, most likely conflating the NBA's date with MLB's separate, confirmed 2032 end date. Both readings are held pending a direct check.
The substrate under every other layer: the infrastructure that lets Wallet-layer transaction data get matched to Glass-layer viewership data without either party exchanging raw personally identifiable information. Whoever owns the match keys and the matching environment collects a toll on every data collaboration that crosses it. The layer's defining 2026 event is an ownership change: LiveRamp, the last major independent operator of that toll booth, agreed in May 2026 to be acquired by Publicis Groupe, and the neutral-intermediary model the open-web data economy was built on is being replaced by holding-company ownership on one side and retail walled gardens on the other.
The identity substrate everything else matches through