A soft reserve is a price threshold in an advertising auction that changes what the winner pays, not whether a bid is accepted. It sits above the hard reserve, the conventional minimum beneath which no bid qualifies at all. A bid below the soft reserve still wins if nothing higher arrives, and is charged its own number. A bid above it pays the greater of the runner-up bid and the soft reserve. Practitioners say soft floor; platform interfaces and the auction theory literature use soft reserve for the same setting.

The mechanism exists because second-price auctions pay sellers badly when only one buyer turns up. The winner is charged one increment above the next bid, so an impression drawing a single $4.00 offer and no competition clears at the floor, however far beneath $4.00 that floor sits. A soft reserve raises the price in exactly that situation and nowhere else. It is a rule about discounting, not eligibility.

How the pricing rule works

Two thresholds produce three outcomes. Write the hard reserve as r, the soft reserve as s, the winning bid as b1 and the runner-up as b2, with s at or above r. If b2 reaches s, the auction settles at b2, as an ordinary second-price auction would. If b2 falls below s but b1 clears it, the price is s. If b1 is itself below s, the winner pays b1, the first-price rule. Set s equal to r and the auction reverts to second price; set s above every plausible bid and it becomes first price in practice.

Microsoft's documentation for Monetize, the supply-side platform formerly known as Xandr and before that AppNexus, supplies the worked example. With a soft floor of $1.25 and bids of $1.50 and $1.00, a plain second-price auction charges $1.01. The soft floor lifts that to $1.25, capturing 24 cents the pricing rule would otherwise have handed to the buyer. Under the same setting, where the highest bid is $1.00, the winner pays $1.00, because a soft floor never rejects.

Where the setting lives

In the Xandr platform API the two values are fields on a yield management floor object. The field hard_floor is required at creation; soft_floor is optional, must be at least the hard floor, and is documented as the CPM floor used as the soft reserve, the lowest price at which price reduction will occur. A priority value between 1 and 10, defaulting to 5, resolves overlapping rules. A calculation type of gross or net decides whether the number is measured against the buyer's bid or the publisher's payout, shifting the effective level by the whole intermediary fee. Floors set here supersede placement-level reserve prices.

Nothing equivalent exists in OpenRTB, the IAB Tech Lab specification that carries bid requests between sellers and buyers. The request exposes one floor, imp.bidfloor, alongside an auction type field, at, where 1 means first price and 2 means second price plus. A bidder receiving a request marked second price cannot learn from the protocol whether a soft reserve applies, or at what level. Researchers Shuai Yuan, Jun Wang and Xiaoxue Zhao, analysing exchange logs in 2013, observed that advertisers are often unaware a soft floor exists, and inferred its use from impressions where the price paid equalled the bid submitted. Roughly 40% of their dataset fell into that category.

Origin and evolution

Soft floors emerged as exchanges industrialised second-price selling in the early 2010s. AppNexus patent filings call the setting a shadow bid price establishing a floor for price reduction only, selectable by content category, geography, placement, size or buyer. Robert Zeithammer of UCLA circulated the first formal analysis in 2017 under the title The Futility of Soft Floors in Auctions, published two years later in Management Science under the neutral title Soft Floors in Auctions.

The sharpest contemporary account came from Ari Paparo, then chief executive of the demand-side technology firm Beeswax, writing in AdExchanger on February 27, 2015. Because a single-bidder auction settles at the floor, he argued, publishers gain by raising floors, and the soft floor simply let sellers "get more money when there's only one bid". He urged sellers to abandon it for first-price auctions.

AppNexus, Index Exchange and OpenX led a switch to first price through 2017 and 2018. Google announced on March 6, 2019 that Ad Manager display and video inventory would move to a unified first-price auction, beginning full rollout the week of September 5, and retired existing floor rules for unified pricing rules capped at 100 per network. A first-price auction performs no bid reduction, so nothing remains for a soft reserve to floor. The setting turned vestigial across the open exchange within three years.

Why it matters to the marketing community

Soft floors are among the reasons the buy side stopped trusting second price. Soft floors, dynamically raised floors and per-buyer pricing let a supply-side platform capture part of the discount the mechanism was meant to hand to buyers, who responded by shading bids in a format where shading was theoretically pointless.

The lineage runs into today's buy-side products. Rubicon Project chief executive Michael Barrett described the exchange's Estimated Market Rate as a repurposed soft-floor product running the same logic in reverse: where a soft floor stopped a price falling, the same machinery aimed at buyers pushes a price up. Since an open auction seller quotes no rate and simply lets a floor stand in for the rate card, a soft reserve is a second, hidden entry on that card.

Contested evidence

Whether soft floors raise revenue remains unsettled. Zeithammer concluded that with symmetric bidders soft floors have no revenue effect at all, because a symmetric equilibrium exists in strictly monotonic bidding strategies and standard revenue-equivalence arguments apply. His 2017 working paper went further: soft floors cannot increase the auctioneer's revenue, and can reduce it once bidders adapt.

Field and laboratory work disagrees. Forsch and co-authors ran a large-scale field experiment in 2017 finding that even low soft floors increase revenue, a divergence Zeithammer attributed to bidders lacking time to adjust. Six economists including Dirk Bergemann of Yale, Peter Cramton and Axel Ockenfels published a fuller account as Cowles Foundation Discussion Paper 2512 in April 2026, updating a version circulated a year earlier. Their argument is behavioural: bidders dislike losing at a price below their own valuation, and a soft reserve sells them entry into a regret-free second-price contest.

Their sessions at the Cologne Laboratory for Economic Research in April 2023, with 244 subjects producing 11,712 bids, found a soft floor of 52 raising revenue 14% over a first-price auction with no reserve, effects ranging from 6% at a floor of 46 to 17% at 58, and efficiency steady near 97% throughout. Bidders accepted the opening price far more readily than predicted, 46% of bids against a forecast 24%.

Limitations and criticism

The central objection is disclosure. Because no OpenRTB field carries the value, a soft reserve operates inside the exchange, visible only in aggregate to anyone bidding into it. That opacity turned floors into an antitrust and standards question rather than a yield question.

Formal rulemaking has not reached the term. The Media Rating Council issued draft Digital Advertising Auction Transparency Standards in September 2025; IAB Tech Lab's Programmatic Auction Definitions, finalised on June 26, 2026 as a 12-step workflow with 15 defined terms, covers auctioneer, bid price, buying platform and final auction. Soft reserve is not among them.

Header bidding closed the door structurally rather than by rule. Prebid's Price Floors Module offers enforcement settings, a skip rate and dynamic floor providers, but no soft floor concept: a wrapper resolving at first price has no discount to withhold.

Disambiguation

A hard reserve, or hard floor, is the minimum bid a seller will accept. Bids beneath it are discarded and the impression may go unsold. A soft reserve never rejects a bid.

Bid shading is the buy-side mirror image: submitting less than an advertiser's valuation in a first-price auction. Soft reserves raise clearing prices from the sell side; shading lowers them from the buy side.

Dynamic floors, sometimes marketed as reserve price optimisation, adjust the hard reserve algorithmically by impression, geography or buyer. They change eligibility; a soft reserve leaves eligibility alone.

A deal ask price is the agreed rate on a private marketplace deal, carried in the same bidfloor field but flagged as a price rather than a minimum. It overrides yield management floors entirely.

Recent developments

Xandr remains the clearest surviving implementation, both fields present in API documentation revised on October 22, 2025. Elsewhere the industry attacks the same problem from the opposite side. Index Exchange introduced per-impression dynamic take rates in October 2025, cutting its own margin so bids falling marginally short of a floor still clear, and reporting a 45% rise in impressions served for The Guardian. That inverts the soft floor logic: instead of lifting the price a buyer pays, the intermediary lowers the price it keeps.

Sell-side pricing control has been partly restored: Google removed unified pricing rules from Ad Manager in December 2025 under antitrust pressure, returning the ability to set different floors for individual buyers. The 2019 rules had stripped out differential floors per exchange, buyer or advertiser, a change the Virginia court found increased AdX impressions and revenue while reducing both for third-party exchanges. Per-buyer floors are hard floors, but they restore by another route the price discrimination soft reserves were built to deliver.

Timeline

  • Early 2010s: Advertising exchanges introduce soft floors alongside conventional reserve prices
  • 2013: Yuan, Wang and Zhao document soft floor use in exchange logs, estimating 40% of impressions clear at effective first price
  • February 27, 2015: Ari Paparo publishes a critique of soft floors and waterfalling in AdExchanger
  • 2017: Robert Zeithammer circulates The Futility of Soft Floors in Auctions
  • 2017: Forsch and co-authors report a field experiment finding low soft floors raise revenue
  • 2017 to 2018: AppNexus, Index Exchange and OpenX switch display inventory to first-price auctions
  • 2019: Zeithammer's analysis appears in Management Science, volume 65, issue 9
  • March 6, 2019: Google announces the move of Ad Manager inventory to a unified first-price auction
  • September 5, 2019: Google begins full rollout, retiring existing floor rules for unified pricing rules
  • April 2023: Cologne laboratory sessions collect 11,712 bids across four soft-floor auction formats
  • September 2025: Media Rating Council issues draft Digital Advertising Auction Transparency Standards
  • October 22, 2025: Xandr platform API documentation carrying the soft_floor field is revised
  • December 2025: Google removes unified pricing rules, restoring buyer-specific floors in Ad Manager
  • April 2026: Cowles Foundation publishes discussion paper 2512 on soft-floor auctions and regret
  • June 26, 2026: IAB Tech Lab publishes final Programmatic Auction Definitions, without a soft reserve entry

Summary

Who. Supply-side platforms and exchanges operate soft reserves as auctioneers, with publishers setting the values. Demand-side platforms and their advertisers pay the resulting prices, usually without visibility into the setting. Microsoft's Monetize platform maintains the best-documented implementation. Academic economists at UCLA, Yale, Cologne and Maryland supply the competing revenue analyses.

What. A price threshold above the hard reserve that governs bid reduction rather than bid eligibility. Above the soft reserve the auction charges the greater of the runner-up bid and the threshold; below it, the winner pays its own bid.

When. Introduced by exchanges in the early 2010s, documented in research from 2013, criticised publicly from 2015, and largely superseded by the 2017 to 2019 industry move to first-price auctions. The economic case for the format was reopened in a 2026 Cowles Foundation paper.

Where. Inside exchange and supply-side platform pricing engines, configured through yield management rules. The value is not carried in OpenRTB, so it does not travel to bidders in the bid request.

Why. Second-price auctions clear at the floor whenever a single bidder competes, which is common in real-time bidding. A soft reserve lets sellers raise those clearing prices without forfeiting the sale, and the resulting opacity is one of the reasons buyers, standards bodies and courts began scrutinising auction rules.