Bidding Agent

A bidding agent is a program that bids in an auction in your place. You define in advance what a thing is worth to you at most and what goal you’re pursuing. After that, the software takes over: it observes the auction, calculates, and places bids. The advantage lies in speed. Many auctions on the internet don’t last days but a few milliseconds, and there a human simply can’t keep up anymore. The best-known application is the trading of advertising space on websites and in apps.

Why no one bids manually anymore

Every time you open a news website, the space for the ad banners on it is auctioned off. This auction starts while the page is loading and is decided after about 100 milliseconds. In this time, several interested parties must place a bid. A human alone would need longer just to read the question. Automated bids here are therefore not a convenience but the precondition for the market to function at all.

On top of that comes the sheer volume. A larger company participates in billions of such mini-auctions every day. Each individual one may bring in or cost a fraction of a cent. Only the mass adds up to relevant sums. An agent that bids on average one percent smarter than the competition can thereby change entire annual results.

The term is economically interesting because a large part of the revenues of Google, Meta, or Amazon comes from exactly such automated auctions. Whoever operates these systems collects a cut with every awarded bid. And whoever buys advertising depends on the quality of their agent. That’s why bidding agents regularly appear in quarterly reports and in antitrust proceedings.

From maximum bid to learned prediction

You know the simplest form from eBay. You name your maximum amount, and the system raises your bid only as far as necessary to stay ahead. That is pure rule-following, without any capacity to learn. Still, it’s already a bidding agent, because it acts independently on your behalf.

Modern agents work differently. They first estimate how valuable the opportunity on offer is to them. In the case of advertising, that means: how likely is it that exactly this person will click on exactly this ad and then buy something? This estimate comes from a model trained on millions of past cases. From the estimated value and its own budget, the agent then calculates its bid.

The auction format matters. In a second-price auction, the winner only pays the amount of the second-best bid. Then it pays off to honestly bid your own value. In a first-price auction, the winner pays their own bid, so the agent deliberately bids somewhat below it. A common misconception is equating bidding agents with high-frequency trading on the stock market. The technology is similar, but on the stock exchange there is no auctioneer distributing awards.

Ad space, electricity, and package prices

Most often you encounter bidding agents invisibly: in every ad you see online. The technical term for this is real-time bidding. In the tools from Google or Meta, the agents are usually called Smart Bidding or Automated Bidding. There, a merchant only sets a goal, such as as many sales as possible per euro spent.

Outside of advertising, you find the same idea on electricity exchanges. There, programs offer quantities of electricity every quarter hour and adjust prices to weather forecasts. Freight forwarders and delivery services also distribute orders via automated auctions. In news articles the term often comes up when regulators examine whether such systems are artificially driving up prices.

For you as a user, this means above all one thing: prices on the internet are rarely fixed. They arise at the moment you open a page. If you read in the future about advertising revenue or dynamic pricing, there is almost always a bidding agent behind it.