Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Uncertainty Quantification of Click and Conversion Estimates for the Autobidding

Ivan Zhigalskii, Andrey Pudovikov, Aleksandr Katrutsa, Egor Samosvat

TL;DR

The DenoiseBid method is proposed, which corrects the generated CTRs and CVRs to make the resulting bids more efficient in auctions and replaces noisy CTR or CVR estimates with those from recovered distributions.

Abstract

Modern e-commerce platforms employ various auction mechanisms to allocate paid slots for a given item. To scale this approach to the millions of auctions, the platforms suggest promotion tools based on the autobidding algorithms. These algorithms typically depend on the Click-Through-Rate (CTR) and Conversion-Rate (CVR) estimates provided by a pre-trained machine learning model. However, the predictions of such models are uncertain and can significantly affect the performance of the autobidding algorithm. To address this issue, we propose the DenoiseBid method, which corrects the generated CTRs and CVRs to make the resulting bids more efficient in auctions. The underlying idea of our method is to employ a Bayesian approach and replace noisy CTR or CVR estimates with those from recovered distributions. To demonstrate the performance of the proposed approach, we perform extensive experiments on the synthetic, iPinYou, and BAT datasets. To evaluate the robustness of our approach to the noise scale, we use synthetic noise and noise estimated from the predictions of the pre-trained machine learning model.

Uncertainty Quantification of Click and Conversion Estimates for the Autobidding

TL;DR

The DenoiseBid method is proposed, which corrects the generated CTRs and CVRs to make the resulting bids more efficient in auctions and replaces noisy CTR or CVR estimates with those from recovered distributions.

Abstract

Modern e-commerce platforms employ various auction mechanisms to allocate paid slots for a given item. To scale this approach to the millions of auctions, the platforms suggest promotion tools based on the autobidding algorithms. These algorithms typically depend on the Click-Through-Rate (CTR) and Conversion-Rate (CVR) estimates provided by a pre-trained machine learning model. However, the predictions of such models are uncertain and can significantly affect the performance of the autobidding algorithm. To address this issue, we propose the DenoiseBid method, which corrects the generated CTRs and CVRs to make the resulting bids more efficient in auctions. The underlying idea of our method is to employ a Bayesian approach and replace noisy CTR or CVR estimates with those from recovered distributions. To demonstrate the performance of the proposed approach, we perform extensive experiments on the synthetic, iPinYou, and BAT datasets. To evaluate the robustness of our approach to the noise scale, we use synthetic noise and noise estimated from the predictions of the pre-trained machine learning model.
Paper Structure (52 sections, 59 equations, 5 figures, 2 tables)

This paper contains 52 sections, 59 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related works.
  3. Autobidding formulations.
  4. Uncertainty quantification for CTR/CVR prediction.
  5. Propagating uncertainty into bidding.
  6. Problem statement
  7. DenoiseBid method
  8. Bidding under CTR-only Uncertainty
  9. Noise model.
  10. Gaussian mixture prior.
  11. Empirically reconstructed prior.
  12. Bidding under joint CTR-CVR uncertainty
  13. Noise model.
  14. Gaussian mixture prior.
  15. Empirically reconstructed prior.
  16. ...and 37 more sections

Figures (5)

  • Figure 1: Performance comparison on the synthetic, iPinYou and BAT datasets under CTR uncertainty.
  • Figure 2: Performance comparison on the BAT dataset under joint CTR-CVR uncertainty.
  • Figure 3: Relative uplifts in conversions and CPC deviation on the Criteo Attribution dataset. Markers * and ** denote statistical significance at the 10% and 5% levels, respectively.
  • Figure 4: Performance comparison on the synthetic dataset under joint CTR-CVR uncertainty.
  • Figure 5: Performance comparison on the iPinYou dataset under joint CTR-CVR uncertainty.