Propensity Score Matching for Attribution: Propensity score matching (PSM) claims to fix attribution bias. But does it actually work? We dissect PSM, expose its flaws, and offer a real solution: causal inference.
Read the full article below for detailed insights and actionable strategies.
Propensity score matching (PSM) sounds promising: a statistical technique to correct attribution bias by creating comparable groups. The reality? PSM is a band-aid on a bullet wound. It doesn't address the fundamental problem: correlation is not causation. If you're relying on PSM, you're still making flawed decisions based on incomplete data. We’ll show you why PSM falls short and how causal inference provides a real solution.
What is Propensity Score Matching?
Propensity score matching is a statistical method that attempts to estimate the effect of a treatment or intervention by accounting for the covariates that predict receiving the treatment. In the context of marketing attribution, the "treatment" is exposure to a marketing touchpoint (e.g., seeing an ad), and the goal is to estimate the causal effect of that touchpoint on a desired outcome (e.g., a purchase).
The core idea is to create two groups: one that received the treatment (exposed group) and one that did not (control group), but which are otherwise as similar as possible. This is achieved by calculating a "propensity score" for each individual, representing their probability of receiving the treatment based on their observed characteristics. Individuals with similar propensity scores are then matched, and the difference in outcomes between the matched groups is used to estimate the treatment effect.
Think of it like this: you want to know if showing ads to Gen Z on TikTok increases sales. PSM tries to create a control group of non-TikTok users who are just like the TikTok users in every other way (age, income, interests, etc.). Then, you compare the sales of the TikTok users to the sales of this matched control group. The difference is supposed to be the impact of the TikTok ads.
Why Propensity Score Matching Fails for Marketing Attribution
PSM suffers from several critical flaws that make it unsuitable for accurate marketing attribution. These flaws stem from its reliance on observational data, its inability to account for unobserved confounders, and its inherent limitations in establishing causality.
1. Garbage In, Garbage Out: Observational Data Problems
PSM relies on observational data, which is inherently biased. People are not randomly assigned to marketing touchpoints; they self-select based on their existing preferences and behaviors. This self-selection introduces confounding variables that PSM cannot fully address. PSM can only control for observed confounders. If there are unobserved factors influencing both the marketing touchpoint exposure and the outcome, PSM will fail to eliminate bias. Imagine trying to match customers based on demographics alone, while ignoring their brand loyalty or previous purchase history. The matched groups will still be fundamentally different.
2. Can't Control the Uncontrollable: Omitted Variable Bias
Omitted variable bias (OVB) is the bane of any statistical method that relies on observational data. OVB occurs when a relevant variable is excluded from the analysis, leading to biased estimates of the effects of included variables. In marketing attribution, countless unobserved factors influence customer behavior, such as word-of-mouth, competitor actions, and personal circumstances. PSM cannot account for these unobserved factors, leading to biased attribution estimates. You might attribute a sale to a recent ad click, when in reality, it was driven by a friend's recommendation that you didn't even track.
3. Correlation Isn't Causation: The Fundamental Flaw
At its core, PSM is a sophisticated form of correlation analysis. It identifies associations between marketing touchpoints and outcomes but cannot establish causation. Just because two groups are similar on observed characteristics does not mean that the marketing touchpoint caused the difference in outcomes. There may be other unobserved factors at play, or the relationship may be spurious. You see a correlation between website visits and sales, but that doesn't mean the website visits caused the sales. Maybe both were driven by a successful email campaign. PSM doesn't tell you.
4. LLMs Can't Fix It: Spider2-SQL Benchmark
Even advanced machine learning techniques, including Large Language Models (LLMs), struggle with the complexity of real-world attribution problems. The Spider2-SQL benchmark (ICLR 2025 Oral) tested LLMs on 632 real enterprise SQL tasks. GPT-4o solved only 10.1%, o1-preview only 17.1%. Marketing attribution databases have exactly this level of complexity. If LLMs can't even query the data correctly, they certainly can't perform causal inference. Trying to use LLMs to "fix" PSM is like putting lipstick on a pig.
A Real Solution: Causal Inference
Instead of relying on flawed methods like propensity score matching, embrace causal inference. Causal inference methods are specifically designed to identify causal relationships from observational data. Techniques like do-calculus, instrumental variables, and difference-in-differences can disentangle cause and effect, even in the presence of confounding variables. Causality Engine uses these methods to deliver accurate, actionable insights.
- Accuracy: Causal inference achieves 95% accuracy, compared to the 30-60% of traditional attribution models.
- ROI: Our clients see a 340% increase in ROI.
- Real Results: One customer increased ROAS from 3.9x to 5.2x, adding +78K EUR/month in incremental sales.
Questioning Common Beliefs About Marketing Attribution
Is Propensity Score Matching a Valid Approach for Attribution Bias Correction?
No, propensity score matching is not a valid approach for attribution bias correction. While it attempts to create comparable groups, it cannot account for unobserved confounders or establish causality. It's a band-aid, not a cure.
Can PSM Accurately Measure the Impact of Marketing Touchpoints?
No, PSM cannot accurately measure the impact of marketing touchpoints. It relies on observational data and correlation, not causation. It's prone to omitted variable bias and cannot disentangle cause and effect.
Are There Better Alternatives to Propensity Score Matching for Marketing Attribution?
Yes, there are far better alternatives. Causal inference methods, such as do-calculus and instrumental variables, are specifically designed to identify causal relationships from observational data. Causality Engine uses these methods to deliver accurate and actionable insights.
Stop wasting time and resources on flawed attribution methods. Discover how Causality Engine can unlock the true power of your data and drive incremental sales.
Sources and Further Reading
Get attribution insights in your inbox
One email per week. No spam. Unsubscribe anytime.
Key Terms in This Article
Attribution Model
An Attribution Model defines how credit for conversions is assigned to marketing touchpoints. It dictates how marketing channels receive credit for sales.
Causal Inference
Causal Inference determines the independent, actual effect of a phenomenon within a system, identifying true cause-and-effect relationships.
Confounding Variable
Confounding Variable is an unmeasured factor that influences both the marketing input and the desired outcome, distorting the true impact of a campaign.
Instrumental Variable
Instrumental Variable is a causal analysis method that estimates a variable's true effect when controlled experiments are not possible, using a third variable that influences the outcome only through the explanatory variable.
Machine Learning
Machine Learning involves computer algorithms that improve automatically through experience and data. It applies to tasks like customer segmentation and churn prediction.
Marketing Attribution
Marketing attribution assigns credit to marketing touchpoints that contribute to a conversion or sale. Causal inference enhances attribution models by identifying true cause-effect relationships.
Propensity Score
A propensity score is the probability a unit receives a specific treatment given observed characteristics. It reduces selection bias in observational studies, enabling causal inference when randomized experiments are not possible.
Propensity Score Matching
Propensity Score Matching is a statistical method that estimates the causal effect of a treatment from observational data. It matches individuals with similar likelihoods of receiving treatment to isolate its impact.
Ready to see your real numbers?
Upload your GA4 data. See which channels drive incremental sales. 95% accuracy. Results in minutes.
Book a DemoFull refund if you don't see it.
Stay ahead of the attribution curve
Weekly insights on marketing attribution, incrementality testing, and data-driven growth. Written for marketers who care about real numbers, not vanity metrics.
No spam. Unsubscribe anytime. We respect your data.
Frequently Asked Questions
What are the limitations of propensity score matching?
Propensity score matching (PSM) is limited by its reliance on observational data, inability to account for unobserved confounders, and failure to establish causation. It's a glorified correlation analysis, not a causal inference technique.
How does causal inference improve upon propensity score matching?
Causal inference methods, unlike PSM, are designed to identify causal relationships even with confounding variables. Techniques like do-calculus and instrumental variables disentangle cause and effect, providing accurate attribution.
Can propensity score matching be used in conjunction with other attribution methods?
While PSM can be combined with other methods, its fundamental flaws remain. It cannot transform correlation into causation. Using PSM alongside another method is like adding a broken wheel to a car. It won't improve performance.
