Blog · CTV

CTV arbitrage can stay under 1% of measured impressions while the share of ranked resellers jumps sixfold.

October 29, 2026 · RTBlint

On October 29, 2026, Pixalate released Q3 2026 Connected TV Seller Trust Index 2.0 rankings across Roku, Amazon Fire TV, Apple TV, Samsung TV, and LG Smart TV in the United States. Trade coverage led with the channel comparison: less than 1% of ranked U.S. CTV inventory was resold rather than sold publisher-direct, versus 21.4% on web and 16.4% on mobile in the Web and Mobile STI releases the same day. That headline is true for the aggregate arbitraged inventory ratio Pixalate publishes. It is not the same statement as saying the CTV supply path got simpler this quarter.

Two denominators, one press day

Pixalate measures each seller with Supply Chain Object analysis on predominantly open-auction, buy-side traffic. The Arbitraged Inventory Ratio is the percentage of a seller's measured impressions that did not flow through an authorized-direct path, meaning the share sold as a reseller or arbitrage intermediary according to Pixalate's SCO work. At the channel level for Q3 2026, ranked U.S. CTV inventory shows 0.8% in that bucket, roughly flat against 0.7% in Q2 and down from 4.1% in Q1.

The same release documents a second cut: how many ranked seller listings behave as primarily resellers, defined as selling more than half of measured inventory through resale rather than authorized-direct paths. That count rose to 7 of 118 listings, about 6%, from 2 of 195 listings, about 1%, in Q2, while still below the 8% share recorded in Q1. Samsung TV had the highest platform concentration, with 3 of 14 ranked sellers in Q3 classified that way. No ranked seller on LG Smart TV resold more than half of measured inventory in the index tables Pixalate published.

GlobeNewswire and the Pixalate blog posts that shipped October 29 pair the low CTV arbitrage percentage with platform leaderboards where Google AdExchange ranks first on all five CTV platforms, with Magnite, OpenX, Sharethrough, Xandr Monetize, and Amazon Publisher Services appearing in top-three slots by platform. Those tables also disclose per-seller SIVT ratios, GIVT ratios, and Likely Fraudulent Bundle ID Traffic Ratios that do not move in lockstep with a 0% arbitraged inventory ratio on a given row. A buyer who stops diligence when arbitrage rounds to zero on the headline misses the bundle and invalid-traffic columns sitting beside it.

What authorized-direct means on the wire

Pixalate contrasts what sellers claim, via app-ads.txt DIRECT lines detected across apps with open programmatic activity, with verification of how impressions actually flowed. OpenRTB 2.x carries that story in source.ext.schain nodes and in the seller identity fields downstream systems use to reconcile app-ads.txt. An authorized-direct path is not a vibe check on the exchange brand; it is a claim that the hop you are buying matches the publisher's declared DIRECT authorization for that app and seller ID.

When web STI 2.0 reports 21.4% of global authorized web inventory resold in Q3, up more than six percentage points from 15.3% in Q2, the industry gets a clear arbitrage alarm on the open web. CTV's 0.8% U.S. figure sounds like the opposite problem until you notice the seller-listing reseller share climbing while aggregate impression weight stays low. That pattern is compatible with a smaller set of resellers moving fewer measured impressions through ranked paths, while more SSP names on the STI roster flip into primarily-reseller status on specific platforms. It is also compatible with CTV inventory concentrating on exchanges whose measured paths look direct even when bundle-level fraud metrics stay elevated.

Fraudulent Bundle ID Traffic Ratio appears explicitly in the Q3 CTV tables. Pixalate defines it as the percentage of a seller's ads sold using likely fraudulent bundle IDs in the bid stream. RTBlint covered bundle mimicry mechanics separately; the STI release puts numbers next to arbitrage on the same row. Google AdExchange leads several platform tables with 0% arbitraged inventory ratio while still carrying risk labels such as High SIVT or High GIVT and non-zero fraudulent bundle ratios on Roku, Fire TV, Apple TV, and Samsung TV in the press tables reproduced October 29.

Why exchanges feel both stories at once

Supply teams read STI as a quarterly scoreboard. Operations teams live in request logs where a single app can send conflicting app.bundle values, incomplete schain nodes, or seller IDs that do not match the app-ads.txt row marketing emailed to buyers. The index summarizes months of observation; the bid request is the object you can reject before spend clears. When Q3 headlines say CTV is the cleanest channel on arbitrage percentage, sales narratives compress that into buyer-facing decks. Fraud and supply-path engineers still see bundle warnings and clawback research on the same publishers from earlier Pixalate products.

Clawback warnings and STI arbitrage metrics answer different questions. Clawback labels reflect dispute history on publishers in MRT; STI arbitrage ratios reflect SCO-shaped path analysis on measured auctions. An exchange can pass impressions with readable schain while the publisher behind the app accumulates off-exchange dispute signals. Neither metric replaces validating the JSON your bidder receives on Monday morning.

The October 29 web STI release adds context for cross-channel buyers. Mobile authorized inventory resold at 16.4% globally in Q3, also up quarter over quarter. Treating CTV as uniformly safer because 0.8% is smaller than 21.4% ignores that the CTV index counts different apps, platforms, and seller lists, and that the reseller-listing count on CTV moved faster than the aggregate arbitrage percentage this quarter. Unified dashboards that only chart channel-level arbitrage percentages will mis-rank where to spend engineering time.

Fields to read before you trust the headline

Pixalate publishes DIRECT Publisher Penetration as the count of apps with open programmatic ad activity that list a seller as DIRECT in app-ads.txt. High penetration with a 0% arbitraged ratio still leaves bundle and IVT columns to interpret. A request can present a DIRECT-authorized seller while app.bundle or device fields disagree with store records buyers trust.

// Illustrative schain fragment (OpenRTB 2.x)
"source": {
  "ext": {
    "schain": {
      "complete": 1,
      "nodes": [
        { "asi": "publisher-ssp.example", "sid": "12345", "hp": 1 },
        { "asi": "exchange.example", "sid": "67890", "hp": 1 }
      ]
    }
  }
}

STI verification asks whether measured impressions matched authorized-direct paths for those nodes. RTBlint asks whether the request you are about to route is internally consistent: bundle format, required video object fields, and schain completeness flags buyers configured in their adapters. A seller can score well on Pixalate's quarterly arbitrage ratio while still sending requests your integration marks invalid via BidResponse.nbr code 2.

Platform splits matter because Samsung TV showed the highest share of ranked sellers that are primarily resellers in Q3, while LG tables showed no seller above half resold inventory. Platform-specific Prebid and SSP configs that assume one CTV hygiene policy across all living-room OS families will misallocate blocks. Read STI platform tables as separate markets, not as one CTV bucket.

What to do this week

Separate metrics in QBR decks: show arbitraged inventory ratio and primarily-reseller listing share side by side; do not merge them into one CTV cleanliness bullet.
Pull platform tables: compare Samsung, Roku, and Fire TV seller rows for fraudulent bundle ratios even when arbitraged inventory ratio prints 0%.
Reconcile app-ads.txt: when DIRECT penetration counts rise in STI, spot-check apps that started listing your seller ID against live app.publisher.id and schain seller nodes in production traffic.
Keep clawback research separate: follow MRT dispute labels without assuming STI arbitrage scores substitute for that diligence.
Validate requests: paste production CTV bid requests into the bid request tester after any STI-driven policy change so blocks target shape failures, not vendor rank alone.

The honest limit

Pixalate's indexes reflect buy-side open auction traffic Pixalate can measure and opinions stated in the October 29 releases. They are not exhaustive of private marketplace paths or every exchange's seller graph. RTBlint does not compute arbitrage or call MRT. It checks whether the OpenRTB request is structurally coherent before you spend cycles blaming demand for unfilled CTV pods.

The Q3 story worth acting on is not that CTV arbitrage disappeared. It is that two published cuts, impression-weighted arbitrage percentage and reseller-heavy seller listings, diverged while web arbitrage jumped to 21.4%. Supply engineers should optimize for the bid request and both STI cuts, not for the friendliest headline on release day.

Sources

Percentages and seller table descriptions are from Pixalate's October 29, 2026 CTV and Web Seller Trust Index 2.0 releases. OpenRTB field names follow IAB Tech Lab conventions. RTBlint is independent and not affiliated with IAB, IAB Tech Lab, or Pixalate.