To Broad-Match or Not to Broad-Match : An Auctioneers Dilemma ?

We initiate the study of an interesting aspect of sponsored search advertising, namely the consequences of broad match-a feature where an ad of an advertiser can be mapped to a broader range of relevant queries, and not necessarily to the particular keyword(s) that ad is associated with. Starting with a very natural setting for strategies available to the advertisers, and via a careful look through the algorithmic lens, we first propose solution concepts for the game originating from the strategic behavior of advertisers as they try to optimize their budget allocation across various keywords. Next, we consider two broad match scenarios based on factors such as information asymmetry between advertisers and the auctioneer, and the extent of auctioneer’s control on the budget splitting. In the first scenario, the advertisers have the full information about broad match and relevant parameters, and can reapportion their own budgets to utilize the extra information; in particular, the auctioneer has no direct control over budget splitting. We show that, the same broad match may lead to different equilibria, one leading to a revenue improvement, whereas another to a revenue loss. This leaves the auctioneer in a dilemma - whether to broad-match or not. This motivates us to consider another broad match scenario, where the advertisers have information only about the current scenario, and the allocation of the budgets unspent in the current scenario is in the control of the auctioneer. We observe that the auctioneer can always improve his revenue by judiciously using broad match. Thus, information seems to be a double-edged sword for the auctioneer.

💡 Research Summary

The paper initiates a systematic study of “broad match” – a feature in sponsored search advertising that allows an advertiser’s ad to be shown for a larger set of related queries beyond the exact keyword it is associated with. While the importance of broad match is widely acknowledged in industry, no formal analytical framework existed to assess its impact on the strategic behavior of advertisers and on the revenue of the search engine (the auctioneer).

The authors first formalize the underlying auction environment. There are N advertisers, M keywords, and K ad slots. Each advertiser i has a daily budget B_i and a true valuation v_{i,j} for each keyword j, together with a relevance factor e_{i,j}. Their product s_{i,j}=v_{i,j}·e_{i,j} serves as a combined quality score that determines ranking under the standard Rank‑by‑Revenue (RBR) with Generalized Second Price (GSP) mechanism. Advertisers first decide how to split their daily budget across the M keywords (the “budget‑splitting stage”), and then, whenever a query for keyword j arrives, a static one‑shot GSP auction is run among the advertisers who still have remaining budget for that keyword.

Because computing an exact best response for the budget‑splitting game is NP‑hard, the paper introduces a weaker but algorithmically tractable equilibrium notion. A “Broad‑Match Equilibrium” (BME) is defined as a local Nash equilibrium where, given a budget split, no advertiser can improve its marginal payoff (bang‑per‑buck) on any keyword by moving a small amount of budget to another keyword. The authors prove that a locally optimal best response can be computed in strongly polynomial time, and that an ε‑approximate Nash equilibrium (ε‑NE) can also be found efficiently. Thus the analysis proceeds under either BME or ε‑NE, with all main results holding for both concepts.

Two distinct informational scenarios are examined.

Scenario 1 – Full Information: Advertisers know the exact broad‑match mapping (i.e., which new keyword‑advertiser edges are created and their quality scores) and they control their own budget reallocation. The paper constructs concrete examples showing that the same broad‑match configuration can admit multiple BME outcomes: in one equilibrium the auctioneer’s revenue exceeds the revenue without broad match, while in another equilibrium it falls below. Hence the auctioneer faces a genuine dilemma – broad‑match may be beneficial or harmful depending on which equilibrium is realized, and predicting the outcome is computationally non‑trivial. The same phenomenon persists under ε‑NE.

Scenario 2 – Limited Information & Auctioneer Control: Advertisers are unaware of the broad‑match extension; they only see the original keyword set. After the broad match is applied, any budget that remains unspent in the original scenario is re‑allocated by the auctioneer arbitrarily. The authors prove that, provided the quality of the broad match is “good” (i.e., the added edges have sufficiently high s_{i,j}), the auctioneer can always re‑allocate the leftover budget in a way that strictly increases his revenue. In this setting, information asymmetry actually benefits the auctioneer, turning the “double‑edged sword” observation from Scenario 1 on its head.

The paper also discusses social welfare (the sum of advertiser utilities plus user surplus). In the full‑information case, different BME can either raise or lower total welfare, whereas in the limited‑information case the auctioneer’s revenue‑maximizing reallocation typically does not dramatically harm welfare.

Algorithmically, the authors present strongly polynomial‑time procedures for computing a locally optimal best response, for finding a BME, and for constructing an ε‑NE. These algorithms do not depend on the volume of queries or the absolute size of daily budgets, only on the numbers of advertisers, keywords, and slots.

In conclusion, the work shows that the effect of broad match on auction revenue is highly sensitive to the information structure and to who controls budget splitting. It provides a novel game‑theoretic framework for analyzing multi‑keyword budget allocation, introduces tractable equilibrium concepts, and offers practical insights for search‑engine operators: the decision to enable broad match must consider not only the quality of the match but also the degree of information disclosed to advertisers and the extent of the platform’s control over budget redistribution. Future research directions include dynamic (time‑varying) broad‑match quality, richer slot models, empirical validation with real search data, and mechanisms that can steer the system toward revenue‑enhancing equilibria.