Auctions Do Not Always Beat Fixed Pricing. On a Thin Creator Market, They Lose.

Based on the research ofWang, Cui and Jiang, "Are Auction-Based Promotions More Profitable for Short Video Platforms than Fixed Pricing?," Information Systems Research, 2026

Why an auction is not automatically the better mechanism

Zhongbin Wang, Shiliang Cui, and Zhong-Zhong Jiang built a classical priority-auction model of the promotion tools short video platforms sell to creators: pay-to-boost systems that push a video higher in the feed. They compared it against a straightforward fixed-price policy, the same slot sold at the same price to whoever wants it. The result runs against the standard assumption that a live, competitive bidding process must be closer to economically efficient. Auctions generate higher platform revenue than fixed pricing only when the creator-to-viewer ratio is high enough that creators are genuinely competing for a scarce pool of attention. Below that threshold, fixed pricing can produce more revenue, and it does so while keeping engagement on promoted content stronger, because auctions pull in a wider set of creators, including ones whose videos have little chance of holding a viewer, which dilutes the average quality of what gets promoted. The paper's own numbers sharpen the point further: aggregate creator welfare, summed across all the creators on the platform, improves under auctions only when the creator-to-viewer ratio is comparatively small, the near-opposite condition from the one that makes auctions the better revenue engine for the platform. There is no single ratio at which auctions are simply "the correct" mechanism; there is a threshold, and which side of it a platform sits on changes what the right pricing model is.

The flat-fee side: TikTok Promote and Spotify's Discovery Mode

Two real products illustrate what a fixed, non-competitive promotion mechanism looks like in practice. TikTok's in-app Promote tool sells "promotion packs": a preset budget paired with an estimated result, built so a creator can tap Promote on a post, choose a package, and pay, without setting a bid or watching a live auction unfold. TikTok's own support documentation is explicit that Promote is a distinct, simpler layer sitting alongside TikTok Ads Manager, which is the platform's full real-time bidding system for advertisers, the two coexist because they serve different jobs: one is a menu, the other is a marketplace. Spotify's Discovery Mode is a cleaner fixed-rate case. An artist opts specific songs into Discovery Mode, which raises the odds those songs are surfaced in Radio and Autoplay; in exchange, Spotify applies a flat 30 percent commission to the recording royalties earned specifically from streams that occur in Discovery Mode contexts, with no upfront budget and no bidding against other artists for the slot. Every other stream of that song stays commission-free. Eligibility itself is gated by scale, an artist team needs at least 25,000 monthly listeners and at least three qualifying songs, which is its own signal that Spotify built the tool for artists past a certain floor of demand, not for the deepest, most saturated end of the market. Both tools read as products designed for a promotion market where competitive bidding would be overkill: a flat number is simpler to understand, cheaper to administer, and, per the paper's model, can leave more revenue and engagement on the table for the platform precisely because it does not need to solve a scarce-attention allocation problem yet.

The auction side: what platforms build once attention gets scarce

Contrast that with the tools platforms build once genuine scarcity sets in. Meta's ad auction, which runs the placements across Facebook and Instagram, picks a winner by "total value", a combination of the advertiser's bid, an estimate of how likely that particular person is to engage, and a measured ad-quality score, and Meta states plainly that this design will never charge an advertiser more than their own bid, only enough to win against the next-best competing ad. Amazon runs Sponsored Products the same way: an auction-based inventory model where advertisers set a maximum cost-per-click or cost-per-thousand-viewable-impressions bid, and Amazon recalculates a suggested bid and bid range daily from the past seven days of winning bids in similar auctions, because the market it is pricing shifts that fast. Google's search ad auction, the model much of the modern ad-tech industry inherited its logic from, works on the same principle: bid combined with ad quality produces an "Ad Rank," and an advertiser typically pays less than their maximum bid, with price rising only as competing bids close in. These are all markets with enormous numbers of advertisers or sellers chasing a comparatively fixed amount of viewer or shopper attention, precisely the "creator-to-viewer ratio is high" condition the paper identifies as the point where an auction stops being theoretically elegant and starts actually outearning a price list. It is not a coincidence that the biggest ad auctions in the industry sit on top of the biggest, most contested attention pools in the industry.

The squeezed middle: who auctions actually help

Once a platform is in genuine auction territory, the paper's second finding is the one operators are least likely to expect. Auctions do not lift all creators evenly. Content with very high appeal wins auctions on its own merit and gets promoted cheaply relative to the traffic it earns; content with very low appeal has so little to lose that bidding aggressively for the small chance of a payoff still makes sense. It is the broad middle, creators whose content is solidly good but not exceptional, who come out worse off under an auction than they would under a flat price, because they are competing against both the strongest content in the queue and the sheer volume of weak content the auction's low barrier to entry lets in. That crowding effect also explains the paper's engagement finding: more creators promoting means more low-appeal videos entering the promoted feed, and average engagement on promoted content falls even as the platform's promotion revenue rises. A mechanism sold as "letting the market decide" ends up doing the opposite of what that phrase implies for most participants, it decides in favor of the extremes and against the median.

Match the mechanism to the ecosystem's age

The paper's own framing is a life-cycle argument, not a static preference for one mechanism. Early in a platform's life, when the pool of creators wanting promotion is still small relative to viewers, fixed pricing is not a compromise, it is the revenue-maximizing and engagement-preserving choice. As the creator base grows and competition for the same viewer attention intensifies, the crossover point gets crossed, and an auction starts to outearn the price list, with the aggregate creator-welfare case for auctions actually running the other direction, strongest earlier rather than later. The paper adds one further complication worth carrying into any pricing review: stronger network effects, the more a platform's value depends on creators drawing in other creators and viewers drawing in other viewers, reduce the platform's own incentive to adopt auctions at all, because the dilution auctions cause to engagement quality costs more in a network-effects-heavy business than the extra promotion revenue is worth. Before treating "move to auction pricing" as a natural maturity milestone, a platform's own operators should be able to answer a narrower question: what is our current creator-to-viewer ratio, and is it actually past the threshold where an auction wins, or are we about to trade real engagement for a mechanism that, on paper, looks more sophisticated than the one it's replacing.