Real-Time Bidding (RTB) and Dynamic Pricing

based on the paper: Display Advertising with Real-Time Bidding (RTB) and Behavioural Targeting by Jun Wang.

1. What is Real-Time Bidding (RTB)?

RTB is a way of selling and buying online ad impressions one at a time, in real time, via auctions.
When a user loads a page/app:
1. An ad request is sent to an ad exchange.
2. Multiple demand-side platforms (DSPs) bid for the impression.
3. The highest bidder wins, and their ad is shown.
4. This all happens in ~100 ms.

Key players:

Publisher: owns the site/app; sells impressions.
Ad exchange / SSP: runs the auction (the RTB marketplace).
Advertiser / DSP: bids for impressions to show ads to users.

2. Auction Mechanics in RTB

Most RTB systems use variants of a second-price auction (or modified second-price):

Each bidder submits a bid (e.g. cost-per-mille, CPM).
Bids: $b_1 \ge b_2 \ge \dots \ge b_K$
Winner: highest bid $b_1$
Price paid (classic second price): approximately the second-highest bid $b_2$ (sometimes with floors and other adjustments)

Why second-price?

Encourages bidders to bid closer to their true value, because they pay (roughly) the next highest price, not their own.

3. Where Dynamic Pricing Comes In

Dynamic pricing in RTB is mainly a publisher-side concept:

The publisher (or exchange) adjusts prices and contract terms over time and across impressions to maximise revenue under uncertainty, instead of using a single fixed price.

Dynamic pricing shows up in three main areas:

Dynamic reserve prices in RTB auctions
Programmatic guaranteed / direct deals with time-varying prices
Ad options / First Look deals priced off future RTB uncertainty

4. Dynamic Reserve Prices (Floor Prices)

4.1 What is a reserve price?

A reserve (floor) price $\alpha$ is the minimum price at which the publisher is willing to sell the impression.
In a second-price auction with a reserve:
- If $b_1 \ge \alpha$ , the highest bidder can still win.
- If $b_1 < \alpha$ , the impression is not sold in this auction (or is routed elsewhere).

More precisely, given $b_1 \ge b_2 \ge \dots \ge b_K$ :

Neutral case: $b_1 > b_2 \ge \alpha$ $b_{1} > b_{2} \geq α$
- Winner: highest bidder
- Price: $b_2$ (same as no reserve; floor is inactive)
Desirable (binding) case: $b_1 \ge \alpha > b_2$ $b_{1} \geq α > b_{2}$
- Winner: highest bidder
- Price: $\alpha$ (higher than $b*2$ ; floor _increases* revenue)
Bad case: $\alpha > b_1$ $α > b_{1}$
- No winner, no revenue from this impression

4.2 Dynamic aspect

The key is that the reserve price is not fixed. It can:

Change by time of day, user segment, ad format/placement, or inventory scarcity.
Be updated using statistical models of bid distributions or online learning/bandit algorithms.

In practice:

The publisher/exchange tests different reserve prices.
Observes revenue and fill rate:
- Higher floors → higher price per impression but potentially more unsold inventory.
- Lower floors → more impressions sold but lower prices.
Uses algorithms to optimise the trade-off.

5. Dynamic Pricing via Programmatic Guaranteed

RTB is the spot market. On top of that, publishers also sell Programmatic Guaranteed / Programmatic Direct deals:

Contracts to deliver a fixed number of impressions in the future
At a pre-agreed CPM, often sold in advance

Dynamic pricing here means the guaranteed price is time- and state-dependent, based on:

Forecasted future RTB prices
Forecasted inventory supply
Observed demand for guaranteed deals

Logic:

If future RTB competition is expected to be weak → guaranteed prices may be lower to lock in revenue.
If future RTB competition is expected to be strong → guaranteed prices can be higher, since advertisers will pay extra to secure inventory.

Publishers dynamically decide:

How many impressions to allocate to guaranteed contracts vs leaving for RTB.
What guaranteed CPM to charge at each time.

This is a dynamic pricing + allocation problem.

6. Ad Options and First Look: Dynamic Pricing via Option Theory

To manage risk and flexibility, publishers offer ad options or First Look deals.

6.1 Ad options

Advertiser pays an upfront option price for the right (not obligation) to buy future impressions/clicks:
- At a fixed strike price (CPM/CPC)
- During a future time window
The underlying is the future RTB clearing price.

Dynamic pricing tasks:

Model the stochastic process of future RTB prices.
Use option-pricing ideas (Black–Scholes-style, binomial trees, bespoke models).
Set a fair option premium and strike price.

6.2 First Look deals

Advertiser gets priority access to impressions:
- They see the impression before the open auction.
- They can buy at a pre-agreed floor/fixed price.
- If they decline, the impression goes to RTB.

Dynamic pricing here means priority prices/terms depend on:

Forecast distribution of RTB prices
Advertiser’s risk tolerance
Publisher’s revenue objectives

7. Why Dynamic Pricing Matters in RTB

7.1 Revenue maximisation (yield management)

Dynamic pricing boosts revenue per impression, smooths revenue over time, and enables price discrimination.

7.2 Risk management

RTB prices are volatile. Guaranteed deals and options let publishers lock in revenue and advertisers cap future costs.

7.3 Better market segmentation and control

Different channels for different advertiser types:
- Risk-averse, high-value buyers → guaranteed & options (higher prices)
- Price-sensitive buyers → RTB spot market with more uncertainty
Dynamic pricing aligns each buyer type with the contract that fits them best.