Regression Discontinuity Design
Regression Discontinuity Design is a quasi-experimental method that measures an intervention's impact by examining subjects near a specific cutoff point. It determines campaign effectiveness by analyzing customers just above or below a targeting threshold.
By Joris van Huët, Founder & CEOUpdated 4 min read
What is Regression Discontinuity Design?
Regression Discontinuity Design (RDD) is a quasi-experimental statistical technique used to estimate causal effects by exploiting a predetermined cutoff or threshold in the assignment of treatment or exposure. Originating in the 1960s within economics and social sciences, RDD has gained traction in marketing analytics for its ability to infer causality with higher credibility than traditional observational studies. The core concept involves comparing outcomes just above and below the cutoff point, under the assumption that units near the threshold are similar except for the treatment assignment. This design uses the discontinuity in the treatment variable to isolate the causal impact without randomization, making it especially valuable when randomized control trials (RCTs) are impractical or unethical.
In the context of marketing attribution and causal analysis, RDD enables ecommerce and fashion/beauty brands on platforms like Shopify to measure the direct effect of marketing interventions, such as promotional discounts, loyalty programs, or targeted ads, where eligibility is determined by a continuous score or variable. For example, a brand can offer a special discount to customers with a loyalty score above a certain threshold, and RDD would allow marketers to rigorously assess the impact of that discount on purchase behavior by comparing customers just below and just above the cutoff. This method helps overcome confounding variables and selection bias, providing findings into campaign effectiveness and customer response.
By applying RDD, brands can refine targeting strategies, improve budget allocation, and confidently justify marketing spend by understanding which interventions truly drive conversions.
Why Regression Discontinuity Design matters for ecommerce
For ecommerce marketers, especially in the competitive fashion and beauty sectors, understanding the true impact of marketing actions is vital to improving return on investment (ROI). Regression Discontinuity Design offers a reliable framework to uncover causal relationships in scenarios where randomized experiments are not feasible, such as when eligibility for a promotion or loyalty tier is determined by a score or demographic cutoff. By isolating the effect of specific marketing treatments, RDD helps marketers identify which campaigns genuinely influence customer behavior rather than relying on superficial correlations. This precision translates into smarter campaign design and resource allocation, allowing brands to invest in strategies that demonstrably increase sales, customer retention, or lifetime value. For Shopify merchants juggling multiple marketing channels, RDD enhances attribution accuracy by controlling for confounding factors inherent in observational data. Ultimately, this leads to improved business outcomes, reduced wasted spend, and a competitive edge in rapidly evolving marketplaces.
Formula
τ = lim_{x → c^+} E[Y | X = x] − lim_{x → c^-} E[Y | X = x]
Where:
τ is the treatment effect at cutoff c,
Y is the outcome variable,
X is the running variable,
lim_{x → c^+} and lim_{x → c^-} denote the limits from above and below the cutoff respectively.Common mistakes
- Ignoring the validity of the cutoff assignment mechanism, leading to biased estimates if the threshold is manipulable.
- Selecting an inappropriate bandwidth that is too wide or too narrow, which can distort causal effect estimates.
- Failing to check for continuity in covariates or potential confounders at the cutoff, undermining the assumption that units are comparable.
Frequently asked questions
What types of marketing campaigns are best suited for Regression Discontinuity Design?
RDD is ideal for marketing campaigns where treatment assignment is determined by a clear cutoff or threshold, such as loyalty program tiers, discount eligibility based on spend, or customer scoring systems. It works best when there is a continuous running variable and sufficient data points close to the cutoff.How does Regression Discontinuity Design differ from A/B testing?
While A/B testing randomly assigns subjects to treatment or control groups, RDD exploits a predetermined cutoff for assignment without randomization. RDD can estimate causal effects when randomization isn't possible, but requires strict assumptions about the cutoff's integrity and the continuity of other factors.Can RDD be applied to online advertising campaigns on Shopify stores?
Yes, RDD can be applied to Shopify-based online stores, especially when campaigns involve eligibility thresholds, such as special offers for customers above a certain engagement score or purchase history. This helps isolate the causal impact of those ads or promotions on sales.What are the key assumptions for valid Regression Discontinuity Design analysis?
Key assumptions include that the cutoff is strictly enforced and not manipulable, units near the cutoff are comparable, and other covariates evolve smoothly at the threshold. Violations of these assumptions can bias the estimated treatment effect.