Publisher Yield

Unified auction vs waterfall: which CTV ad serving model earns more?

The difference between unified auction and waterfall ad serving can mean a 20–40% difference in publisher yield on the same inventory. Waterfall sequentially offers impressions to demand sources — if the first buyer passes, the second gets a chance, then the third. Unified auction lets all demand sources bid simultaneously, awarding the impression to the highest bidder. The logic for why unified auctions earn more is almost trivially simple: more competition at once produces better prices. But CTV's technical constraints mean the transition is not as clean as it was in display, and India's market has specific nuances.

How waterfall ad serving works — and why it underperforms

In a waterfall setup, a publisher configures a priority list of demand sources in their ad server. When an ad slot becomes available:

  1. The ad server calls Demand Source A (e.g., the preferred direct deal buyer or top SSP) with a price threshold.
  2. If Demand Source A fills the impression at or above the threshold, the auction ends. The impression is sold.
  3. If Demand Source A does not fill — because no buyer met the threshold or no demand was available — the ad server calls Demand Source B.
  4. This continues down the waterfall until an impression fills or all sources are exhausted (resulting in an unfilled slot or house ad).

The waterfall has a fatal flaw: Demand Source D might have been willing to pay ₹500 CPM, but by the time the ad server gets to it, the impression has already sold to Demand Source B at ₹200. Demand Source B happened to be higher in the waterfall, not because it pays more, but because someone put it there historically — often based on stale assumptions or relationships.

The result: the publisher consistently undersells inventory. The highest-value buyer is not always at the top of the waterfall, but the waterfall architecture rewards position, not price.

How unified auction works

In a unified auction (also called an open auction, first-price auction, or header bidding-equivalent for server-side CTV), all demand sources are called simultaneously. Every DSP and demand source submits its bid for the impression at the same time. The highest bid wins. The impression goes to the buyer who actually values it most, not whoever is highest in an arbitrarily ordered list.

This creates genuine price competition. Buyers who want your inventory must bid competitively because they know other buyers are bidding at the same time. Floor prices become the floor of a competitive auction, not a sequential acceptance threshold. The result is systematically higher CPMs on the same inventory.

First price vs second price in unified auctions

Most CTV unified auctions are now first-price: the winning buyer pays what they bid, not a penny above the second-highest bid. This is important for publishers to understand because it changes how buyers bid. In a first-price auction, buyers shade their bids downward (they do not want to overpay). This can dampen the CPM uplift from moving to unified auction if buyers significantly shade their bids. Publishers can partially counteract this with well-calibrated floor prices that establish a credible reserve.

Has CTV actually moved to unified auction?

In the US and Europe, the display industry largely completed its transition to unified/header bidding between 2016 and 2020. CTV has been slower, for reasons rooted in the technology:

  • App-based delivery: CTV runs in apps, not browsers. There is no HTML header to insert JavaScript. The technical implementation of unified auction on CTV requires server-side architecture or SDK-based solutions — both more complex than web header bidding.
  • VAST and SSAI constraints: CTV ad serving typically uses VAST tags or SSAI stitching. Neither was designed for parallel auction dynamics. Adapting these protocols for unified auctions requires infrastructure investment.
  • SSP consolidation: Fewer SSPs serve CTV meaningfully compared to display. A unified auction with two participants is only marginally better than a waterfall with two steps.

The honest picture: large US CTV publishers (Peacock, Paramount+, Tubi) have largely moved to unified auction or header bidding equivalents. Mid-market and smaller publishers — and most non-US markets including India — are still on waterfall or hybrid setups.

What India CTV publishers are actually using

India's CTV programmatic ecosystem is earlier-stage than the US. The dominant monetisation model for large India platforms (JioCinema, Hotstar) is direct IO — brand deals sold by sales teams, not programmatic auctions. Programmatic CTV in India is primarily handled through a small number of SSPs (PubMatic, Magnite, and Google Ad Manager being the most active).

For most India CTV publishers currently running programmatic at meaningful scale:

  • Waterfall setups with two to three SSPs are common — partly because SSP choice is limited, partly because the engineering investment in unified auction has not felt urgent given programmatic's secondary role
  • Google Ad Manager is frequently used as the primary ad server, with Open Bidding (GAM's version of unified auction) available but inconsistently activated for CTV inventory
  • True server-side header bidding (Prebid Server, OpenWrap OTT) is present at some larger publishers but is not standard practice

This means most India CTV publishers are leaving yield on the table today. The yield gap is harder to close when programmatic demand is thin — but the architecture still matters, and moving toward unified auction positions publishers well as India CTV programmatic demand grows.

The hybrid approach: what many publishers actually do

Pure waterfall is increasingly rare even in India. Pure unified auction is not yet standard. Most publishers operate a hybrid:

  • Direct IO deals and PMP deals fill first (these are negotiated outside the auction and are reserved)
  • Remaining inventory goes to a programmatic layer, which may use Open Bidding or a limited parallel call setup
  • Unfilled programmatic impressions fall to a backfill partner or house ads

This hybrid model is pragmatic but leaves the programmatic layer — often the largest volume tier — optimised suboptimally. The cleanest improvement India publishers can make today is ensuring the programmatic layer runs unified auction (via Google Open Bidding or an SSP-level unified auction setup) rather than sequential waterfall.

Which earns more? The data

Global studies on the shift from waterfall to unified auction in display found yield uplifts of 20–50% on average. CTV-specific data is less published, but SSPs running head-to-head tests report similar directional findings — 15–35% higher CPMs on comparable inventory when moving from waterfall to unified auction, primarily because genuinely competitive buyers who were buried in the waterfall can now win impressions.

For India CTV publishers, the uplift will be lower than global averages in absolute CPM terms (because the programmatic demand pool is smaller), but the percentage improvement from correctly implementing unified auction can still be material. On content categories with genuine advertiser demand — live sport, premium originals — the uplift should be significant.

Practical steps for India publishers transitioning from waterfall

  1. Audit your current ad serving setup — map which demand sources are where in your current waterfall or priority hierarchy
  2. Confirm which SSPs you work with have unified auction or Open Bidding support for CTV in India
  3. If using Google Ad Manager: enable Open Bidding for your CTV ad units and verify it is configured correctly for video/CTV formats
  4. If using PubMatic or Magnite: discuss OpenWrap OTT or equivalent server-side unified auction setup
  5. Run a comparison period — measure effective CPM per available impression on waterfall vs unified auction on comparable inventory segments
  6. Move direct and PMP demand out of the waterfall entirely — these should be reserved deals that do not compete with programmatic