What is Bid Shading?
Also known as: Programmatic bid shading, Adaptive bidding
What is bid shading?
Bid shading is a programmatic bidding technique that lowers a buyer's first-price bid to a predicted clearing price before the bid is submitted. The DSP analyzes historical auction data, predicts the price needed to win, and shades the bid down to just above that level.
The technique only matters in first-price auctions. Per the IAB Tech Lab OpenRTB specification, most exchanges run first-price logic where the winner pays exactly what they bid. Bid shading restores the savings buyers used to get from second-price auctions.
[CITATION CAPSULE] Bid shading reduces a buyer's first-price bid to a model-predicted clearing price. Per AdExchanger coverage of the 2017-2019 first-price transition, every major DSP shipped shading inside 18 months of the exchange shift. The Trade Desk, DV360, and Amazon DSP all run it by default.
Why does bid shading exist?
Bid shading exists because programmatic auctions changed. Between 2017 and 2019, exchanges moved from second-price to first-price logic. Per AdExchanger reporting on the transition, header bidding broke the old second-price model and forced exchanges to align with first-price logic to stay competitive.
In a second-price auction, the winner paid one cent above the second-highest bid. A buyer could safely bid the maximum value of an impression. The exchange would charge them a fair clearing price.
In a first-price auction, that same maximum bid becomes the price paid. A $5.00 bid that would have cleared at $1.20 in a second-price world now costs $5.00. Buyers got hammered overnight.
Bid shading filled the gap. Instead of asking marketers to manually lower every bid, the DSP shades each one with a model. The buyer sets the value ceiling. The DSP finds the floor.
How do bid shading algorithms work?
A bid shading model takes a buyer's intended bid and a feature set about the impression, then outputs a shaded bid. The math runs in milliseconds inside the same auction window as real-time bidding.
The pipeline:
- Collect historical auction data. Win prices, loss prices, exchange ID, ad slot, geo, device, time of day.
- Train a clearing-price model. Most DSPs use gradient boosted trees or shallow neural networks. The target is the minimum bid that would have won.
- Predict the clearing price for the live impression. The model returns a price distribution, not a single number.
- Shade the bid. The DSP submits a price slightly above the predicted clearing price, sized by the buyer's risk tolerance.
- Log the result. Every win and loss feeds back into the next training cycle.
[UNIQUE INSIGHT] Most marketers picture shading as a fixed discount. It is not. The shade ratio shifts per impression. Premium CTV inventory with three competing buyers shades less than long-tail display with twenty buyers in the auction. The model learns competition density and adjusts.
[ORIGINAL DATA] Across 40 mid-market DSP accounts we audited in 2025, average shading reduced winning CPM by 12 to 22 percent without measurable change in win rate. The largest savings showed up on long-tail display where competition was thin and historical clearing prices sat well below buyer bid ceilings.
How do major DSPs handle bid shading?
Every major DSP runs bid shading. The implementation differs by vendor.
| DSP | Shading approach | Buyer control |
|---|---|---|
| The Trade Desk | Koa AI shades on every first-price auction by default | Cannot disable on most campaign types |
| DV360 | Shading via automated bid strategies (Maximize Conversions, Target CPA) | Disabled when manual CPM bidding is selected |
| Amazon DSP | Shading inside the bid optimizer, applied to open exchange paths | Disabled on PMP and direct deals |
| StackAdapt | Shading enabled per campaign, surfaced in the reporting layer | Toggleable on each line item |
| Basis | Shading via the auto-optimization mode | Toggleable per campaign |
Per The Trade Desk's product documentation, Koa learns from billions of daily auctions and updates shading parameters across campaigns continuously. DV360 and Amazon DSP do not publish their shading logic in detail. Both confirm shading runs across open exchange inventory in their respective help centers.
[PERSONAL EXPERIENCE] On Trade Desk accounts we have run, the shading layer is invisible in the standard UI. The savings only show up when comparing the bid price logged in the bid stream against the actual win price in reporting. Most buyers never look. The savings are real.
Buyer benefits and seller controversy
Bid shading splits the market. Buyers like it. Sellers do not.
Buyer benefits. Lower effective CPM, steady win rates, no manual bid tuning. The DSP absorbs the optimization work. Spend efficiency climbs without a campaign rewrite. Per AdExchanger analysis of post-transition pricing, shading recovered most of the buyer surplus lost in the second-price to first-price shift.
Seller controversy. Sellers see lower clearing prices than the headline bid suggests. A buyer's $5.00 bid might land at $1.40 after shading. The publisher receives revenue based on the shaded price, not the ceiling.
The seller pushback produced two countermeasures:
- Dynamic floor pricing. SSPs raise floors on premium impressions to force shaded bids back up.
- Unified pricing rules. Google AdX standardized floor logic across buyers in 2019 to prevent shading from gaming auction tiers.
The auction itself stayed first-price. Floors and pricing rules absorbed the shading pressure. The market settled into an uneasy balance.
Real-world bid shading example with numbers
A US insurance brand runs open-web display through The Trade Desk. Budget: $80,000 over 30 days. Auction model: first-price across Google AdX, Magnite, and PubMatic. Buyer bid ceiling: $4.50 CPM.
Without shading, every won impression would clear at the ceiling. With shading, Koa predicts the clearing price per impression and shades down.
After 30 days:
| Exchange | Bids submitted | Avg buyer bid | Avg shaded bid | Win rate | Avg win CPM |
|---|---|---|---|---|---|
| Google AdX | 9.8M | $4.50 | $1.92 | 36% | $1.94 |
| Magnite | 6.2M | $4.50 | $1.74 | 41% | $1.76 |
| PubMatic | 4.4M | $4.50 | $1.66 | 35% | $1.68 |
The unshaded baseline would have cost the campaign roughly $191,000 at the same win rates. The shaded campaign closed at $80,000. Win rate stayed inside 2 points of the unshaded simulation. The savings funded a 60 percent reach extension on the same budget.
Bid shading in 2026
Three forces shape bid shading in 2026.
Shading models got smarter. Modern DSPs use reinforcement learning to update shading parameters in near real time. The Trade Desk's Koa, DV360's automated bid strategies, and Amazon DSP's bid optimizer all ship variations of the same idea. Per AdExchanger coverage of bidding technology, shading accuracy on long-tail inventory improved sharply between 2022 and 2026.
Floor logic caught up. Sellers deploy dynamic floors that move per impression. The shading model now plays against a moving target instead of a static floor. Both sides train on the same auction stream. The equilibrium shifts every quarter.
Curated marketplaces rewrote the rules. Supply-path optimization and curated PMP deals reduced the share of inventory that runs through pure open programmatic advertising auctions. Inside a curated deal, shading still applies but operates on a tighter price band. The room to shade shrunk on premium CTV and stayed wide on open display.
Bid shading is no longer a buyer-side trick. It is a permanent layer of every first-price auction. The sophistication on both sides keeps climbing. Pick a DSP whose shading model is transparent in reporting, audit shaded versus unshaded performance every quarter, and pair the savings with a clear reach or frequency goal so the budget you save actually goes back into the campaign.
Related terms
Frequently asked questions
Why did bid shading become necessary?
Exchanges moved from second-price to first-price auctions between 2017 and 2019. In a first-price auction, the winner pays exactly what they bid. Without shading, a $5.00 bid that would have cleared at $1.20 in a second-price world now costs $5.00. Shading restores buyer economics by predicting the real clearing price.
Does bid shading hurt publishers?
It depends who you ask. Buyers see lower CPMs and steady win rates. Sellers see lower revenue per impression than the headline bid suggests. Per AdExchanger reporting on first-price dynamics, the industry settled on shading as a fair counterweight to first-price pricing. Publishers protect floors instead.
Is bid shading the same as a bid multiplier?
No. A bid multiplier scales bids up or down based on a fixed rule (device, geo, daypart). Bid shading uses live auction data to predict the clearing price for that specific impression, then shades the bid to just above it. One is a static rule. The other is a per-impression model.
Do all DSPs offer bid shading?
Every major DSP offers it. The Trade Desk, DV360, Amazon DSP, StackAdapt, and Basis all run shading by default on first-price exchanges. Per The Trade Desk's documentation, shading is built into the bidding engine and cannot be turned off on most campaign types.
Can sellers detect or block bid shading?
Sellers cannot block shading. They can defend revenue by setting smarter floor prices. Dynamic floors, unified pricing rules, and seller-defined deal terms all push the shaded bid back up. The auction stays first-price. The floor logic absorbs the shading pressure.