PredictBack

Thesis. Prediction markets are pricing and hedging infrastructure for promotional liabilities, not just forecasting venues.

PredictBack lets brands run "if it happens, you get cashback" campaigns, then price and hedge the payout on prediction markets. An uncertain promotional liability becomes a fixed cost.

The problem

Conditional promotions ("national team wins → 20% cashback") are creative campaigns, but financially they are event-contingent promotional liabilities with an unknown cost. In early 2026, Jordan's Furniture promised full refunds if both UConn's men's and women's teams reached their NCAA title games: about $50M of exposure across ~20,000 orders, which it covered with insurance. Most brands instead over-reserve, cap rewards conservatively, or eat the tail risk.

The idea

A prediction-market price p_i is the market cost of $1 paid out if outcome i occurs. So the headline reward a brand can advertise on a fixed hedged budget is:

R_i = b / (k · p_i)

where b is the promotional budget rate per unit of transaction value, k is the economic cost coefficient of the reward, and p_i is the prediction-market price of the contract that pays out under outcome i.

Lower-priced outcomes support larger headline rewards at the same hedged cost. Buy the matching contracts and the payout is funded whatever happens: the uncertain promotional liability becomes a deterministic hedged cost.

Terms

This project introduces three working terms:

• PredictBack — a conditional reward in which a brand promises customers a payout if a public event occurs, sets the reward rate from the event-market price, and hedges the resulting liability at that price.
• Promotional finance — pricing and hedging the financial liabilities that promotions create, instead of treating conditional rewards only as creative campaigns.
• Event-contingent promotional liability — a promotional obligation whose payout depends on the resolution of an external, uncertain event, so its cost is unknown at the time the campaign is offered.

Research

PredictBack's model comes from the working paper "PredictBack: Pricing and Hedging Promotional Liabilities with Prediction Markets" (Chan Park, 2026). The paper is the canonical reference; this repository is the hub that points to it.

• Source: paper.qmd
• Rendered PDF: output/paper.pdf
• Preprint / DOI: SSRN (to be added)

How to cite

@techreport{park2026predictback,
  title       = {PredictBack: Pricing and Hedging Promotional Liabilities
                 with Prediction Markets},
  author      = {Park, Chan},
  year        = {2026},
  type        = {Working paper},
  institution = {Independent Researcher},
  note        = {https://github.com/predictback/research}
}

Plain: Park, C. (2026). PredictBack: Pricing and Hedging Promotional Liabilities with Prediction Markets. Working paper.

Replace the note URL and add the preprint link above once the paper has an SSRN DOI. A dated DOI is the strongest priority anchor.

Status

Working paper, first posted May 2026, developed in public. The model in the paper is the contribution.

PredictBack is the effort to build on it; product work will live in a separate repo (predictback/predictback, coming).

Contact

Chan Park · chanpark0202@gmail.com

---

© 2026 Chan Park. All rights reserved. See LICENSE.