A Time To Event Framework For Multi-touch Attribution
Volume 22, Issue 1 (2024), pp. 56–76
Pub. online: 26 May 2023
Type: Statistical Data Science
Open Access
Open Access
†
Author was at Google when work was done.
Received
3 October 2022
3 October 2022
Accepted
22 April 2023
22 April 2023
Published
26 May 2023
26 May 2023
Abstract
Multi-touch attribution (MTA) estimates the relative contributions of the multiple ads a user may see prior to any observed conversions. Increasingly, advertisers also want to base budget and bidding decisions on these attributions, spending more on ads that drive more conversions. We describe two requirements for an MTA system to be suitable for this application: First, it must be able to handle continuously updated and incomplete data. Second, it must be sufficiently flexible to capture that an ad’s effect will change over time. We describe an MTA system, consisting of a model for user conversion behavior and a credit assignment algorithm, that satisfies these requirements. Our model for user conversion behavior treats conversions as occurrences in an inhomogeneous Poisson process, while our attribution algorithm is based on iteratively removing the last ad in the path.
Supplementary material
Supplementary MaterialThe supplementary material contains three files. The first is a PDF with additional technical material: In Appendix A we prove a result about
E
[
N
o
r
m
a
l
i
z
e
d
C
r
e
d
i
t
(
j
)
] mentioned in section 3.2.1 of the main text. In Appendix B we give a detailed discussion of the similarities and differences between Backwards Elimination (our proposed attribution method) and Shapley Values, another commonly used method. In Appendix C, we present additional simulation scenarios and their results. The other two files are a data file, data.csv, with sample data used in our simulations, and a README file with a detailed description of the data. We do not include code to replicate the simulation results. While our method can be applied using standard Poisson regression methods, our simulation framework is tightly integrated with our production environment and our proprietary data format. This makes sharing the code impractical.
References
De Haan E, Wiesel T, Pauwels K (2016). The effectiveness of different forms of online advertising for purchase conversion in a multiple-channel attribution framework. International Journal of Research in Marketing, 33(3): 491–507. https://doi.org/10.1016/j.ijresmar.2015.12.001
Dugan L (2011). The series hazard model: An alternative to time series for event data. Journal of Quantitative Criminology, 27: 379–402. https://doi.org/10.1007/s10940-010-9127-1
Kireyev P, Pauwels K, Gupta S (2016). Do display ads influence search? attribution and dynamics in online advertising. International Journal of Research in Marketing, 33(3): 475–490. https://doi.org/10.1016/j.ijresmar.2015.09.007
Lewis RA, Wong J (2018). Incrementality bidding & attribution. Available at SSRN: https://ssrn.com/abstract=3129350 or http://dx.doi.org/10.2139/ssrn.3129350.
Li H, Kannan P (2014). Attributing conversions in a multichannel online marketing environment: An empirical model and a field experiment. Journal of Marketing Research, 51(1): 40–56. https://doi.org/10.1509/jmr.13.0050
Xu L, Duan JA, Whinston A (2014). Path to purchase: A mutually exciting point process model for online advertising and conversion. Management Science, 60(6): 1392–1412. https://doi.org/10.1287/mnsc.2014.1952
Yao D, Gong C, Zhang L, Chen S, Bi J (2021). Causalmta: Eliminating the user confounding bias for causal multi-touch attribution. arXiv preprint: https://arxiv.org/abs/2201.00689.
