rtblintAlpha

Blog · CTV

CTV pod bidding grows up: duration floors, guaranteed deals, and live sports

· rtblint

Pod bidding has been in OpenRTB 2.6 since 2022, but 2026 is the year it stops being optional. The Trade Desk has told publishers it wants long-form video ad breaks delivered as pods across its entire supply path by August 31, 2026, and IAB Tech Lab spent the spring adding live-content signals to the spec because a live sports break cannot be treated like an on-demand one.

Why now: live sports breaks the slot-by-slot model

Selling a commercial break one slot at a time works fine for on-demand video, where a viewer's ad break is assembled once, at their own pace. Live sports removes that slack. Millions of concurrent viewers hit the same break at once, and if every slot in every break is its own bid request, the request volume multiplies with the audience. FreeWheel has described the shift to pod-level bidding as a way to cut that multiplication down to something an exchange can actually clear in the seconds available during a live event, rather than a slot-by-slot auction for every viewer.

That is the forcing function behind the fields OpenRTB 2.6 added for pods: imp.video.podid (and imp.audio.podid) group impressions that belong to the same break, podseq marks whether a pod is the first, last, or any pod in the content stream, and slotinpod (replacing the deprecated video.sequence) marks position within the pod. The field that matters most for live TV specifically is rqddurs: an array of the exact durations a creative may run. It exists because live breaks cannot absorb padding or trimming the way a stored asset can, dead air or a truncated end card is visible to the whole audience at once. rqddurs is mutually exclusive with minduration/maxduration; a request sets one pair or the other, never both.

Structured pods, dynamic pods, and where the money is

OpenRTB 2.6 supports two pod shapes. A structured pod is one imp per slot, all sharing a podid, for breaks with a fixed, known geometry. A dynamic pod is a single imp carrying poddur (total seconds available in the break) and maxseq (maximum number of ads that may fill it), letting bidders return creatives of different lengths and letting the ad server assemble the break from whatever wins. Trade press on 2026 rollouts describes the industry as being in a bridge period: dynamic pods are what is actually shipping now, with structured pods, where a buyer can reserve a specific position in the break, described as the next step once dynamic podding is running end to end.

Pricing follows the same split, and it shipped in two stages. A structured slot prices like any other impression, with an ordinary bidfloor. A dynamic pod cannot: a 15-second spot and a 60-second spot filling the same poddur are not the same product. The 2022 base release (2.6-202204) already covered the open-auction side of that with mincpmpersec on imp.video and imp.audio, a floor priced per second of duration rather than per impression. The 2.6-202309 dated snapshot then extended duration-aware pricing two ways. First, durfloors: an array of DurFloors objects, each with mindur, maxdur, and a bidfloor, at imp.video.durfloors, imp.audio.durfloors, and Deal.durfloors, letting a floor vary across several duration bands instead of one flat per-second rate. Second, two deal-level fields: Deal.guaranteed, a boolean marking a deal as a guaranteed buy rather than a floor to bid against (the field is spelled out in full, not abbreviated to guar), and Deal.mincpmpersec, the deal-scoped counterpart to the imp-level field, which floors per-second pricing for the guaranteed or negotiated slice of a pod rather than its open-auction slice. Full field paths and defaults are on the bid floors page.

What shipped this spring: live-content signals

In April 2026, IAB Tech Lab's Programmatic Supply Chain Commit Group opened public comment (through May 28) on a further OpenRTB and AdCOM change aimed squarely at live sports: two new Content object attributes, content.realtime and content.firstbroadcast, plus an updated description for the existing content.livestream field. The goal is to let a seller distinguish content that is happening at this exact moment from content that is merely streamed live but pre-recorded, and from a program in its first broadcast window versus a rerun, since buyers price all three differently. The same proposal updates the ${AUCTION_PRICE} macro description and introduces two new ones, ${AUCTION_DISCOUNT_PCT} and ${AUCTION_DISCOUNT_CPM}, for communicating discounting on win notices. As of this writing the comment window has closed; whether these attributes land in the next dated snapshot is worth tracking if you sell live sports inventory.

What is actually verifiable about adoption

It is easy to overstate how far pod bidding has spread, so here is what is publicly on record rather than an estimate. The Trade Desk has stated a target of transitioning its full publisher supply path to pod bidding by the end of August 2026, and FreeWheel and Index Exchange have both described 2026 as the year demand partners roll out podded requests at scale. There is at least one named, quantified result: Index Exchange reported that DSP StackAdapt, after adopting its ad podding for streaming TV in mid-2025, saw a 71% average reduction in queries per second, a 25% increase in win rate at peak times, and a 39% drop in creative duplication errors, consolidating what had been separate bid requests for each 15- or 30-second slot in a 90-second break into single podded transactions.

That is one SSP-DSP pairing's case study, not an industry-wide adoption rate, and no source found for this article publishes a measured percentage of total CTV volume running through pod fields today. Treat pod bidding in mid-2026 as a large, named-partner rollout in progress with at least one public efficiency result behind it, not a completed industry transition, and verify against your own partners' documentation before assuming a given DSP or SSP sends or expects pod fields on a given deal.

What to check this week

If you run CTV inventory, the failure mode to watch for is a pod field landing in the wrong place, an ext block, or an old field like video.sequence still in use instead of slotinpod. Paste a sample request into the bid request tester to check pod and duration-floor fields against the current 2.6 snapshot, or run the same checks in CI with the CLI. The full field-by-field rundown, including rwdd and ssai, is on the CTV protocol page; floor mechanics and the Deal.guaranteed/Deal.mincpmpersec paths are on the bid floors page.

Sources