Real-Time Bidding (RTB)

The Auction Sequence

A user loads a page. Before content finishes rendering:

1. The publisher's ad server identifies an available impression and passes it to an SSP
2. The SSP constructs a bid request containing page URL, ad slot dimensions, device, geography and any available audience identifiers
3. The request is broadcast to connected DSPs
4. Each DSP evaluates it against active campaigns, applies targeting rules and bidding models, and returns a bid or declines
5. The highest eligible bid wins and the creative is served

The whole sequence typically completes within one hundred milliseconds. Bidders that respond too slowly are simply excluded.

First-Price and Second-Price Auctions

Programmatic display historically ran second-price auctions — the winner paid one cent above the second-highest bid. Bidding your true maximum was rational because you rarely paid it.

The market has largely moved to first-price auctions, where the winner pays exactly what they bid. Bidding your true maximum now means paying it every time, which changes bidding strategy fundamentally.

Bid Shading

Bid shading is the response to first-price auctions. Rather than bidding the maximum you are willing to pay, the DSP predicts the minimum bid likely to win and bids slightly above that.

Done well it maintains win rate while reducing cost per impression. Done badly it loses valuable inventory to competitors. Most major DSPs apply it automatically; the quality of the underlying prediction is a genuine differentiator between platforms.

What Determines Whether You Win

Bid price is the primary factor, but not the only one. Also relevant: whether your creative meets the publisher's technical and category requirements, whether you can identify the user well enough to value the impression, latency of your bid response, and any publisher floor price.

Low win rates with high bids usually indicate a technical or eligibility problem rather than a pricing one — creative rejection, failed targeting match, or timeout.

Why Impressions Cost What They Cost

The same ad slot varies enormously in price depending on what is known about the user. An impression with rich audience data on a premium publisher may clear at many times the price of an anonymous impression on long-tail inventory.

As third-party identifiers decline, the spread between identified and anonymous inventory has widened, which is the commercial driver behind first-party data and contextual targeting investment.

Where the Latency Budget Actually Goes

The whole auction completes in roughly the time it takes a page element to render, and that budget is the constraint shaping every decision in the chain.

The exchange sets a timeout — typically in the low hundreds of milliseconds — and a bidder that does not respond in time is simply excluded. Not penalised, not logged as a loss you can analyse: absent. This produces a failure mode buyers routinely misdiagnose as poor inventory or weak targeting.

What consumes the budget: the network round trip, which is physics and is why bidders run infrastructure close to the exchanges; the bidder's own decisioning, including any model inference; and any external lookup — a user data call, a frequency check, a brand-safety score.

The practical consequences. Every enrichment you add to the bid path costs win rate, so the question for each is whether its lift exceeds the impressions it loses. Timeout rate is a metric worth watching and almost nobody does; a bidder timing out on a share of requests has a capacity problem wearing a performance costume. And geography matters — bidding into a distant exchange means part of the budget is spent before any decision is made.

Reading a Win-Rate Problem Correctly

Low win rate is the most common complaint and has several unrelated causes that look identical in a dashboard.

Symptom	Likely cause
Few bid requests reaching you at all	A supply problem, not a bidding one. Check your integrations and targeting breadth before your bids
Bidding often, winning rarely, clearing price far above your bid	Genuinely outbid. The inventory is more valuable to someone else
Bidding often, winning rarely, clearing price close to your bid	A bid-shading or pacing issue. Small increases may move a lot
Winning almost everything you bid on	Usually bad news. You are probably overpaying, or the inventory nobody else wants
High timeout rate	Infrastructure. No bid adjustment fixes it

The line worth remembering: an unusually high win rate is a finding to investigate, not a result to celebrate. In a competitive auction, winning most of what you bid on means either your bids are too high or you are the only buyer interested — and the second is rarely a compliment to the inventory.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.