Regression Discontinuity — Does Free Shipping Threshold Impact AOV?

The Scenario

BasketBloom is a U.S.-based online home décor and gifting store that introduced a free shipping offer two years ago: any order totaling $50 or more qualifies for free standard shipping. The policy was designed to encourage customers to add one more item to their cart rather than abandon at checkout due to shipping costs. Over time, the pricing team has observed an interesting pattern in the order data — there seem to be noticeably fewer orders just below the $50 mark, and orders just above it appear to be slightly larger than you would expect based on the general spending trend. The business wants to know whether this pattern reflects a genuine causal effect of the free shipping threshold or simply a coincidence in how customers naturally shop.

The Statistical Challenge

A sample of 500 orders placed within a $25–$75 order value range has been collected — the zone closest to the $50 threshold where the incentive effect is most likely to be felt. For each order, you have the final order value. The challenge is not simply to compare high spenders to low spenders, because customers who spend more are fundamentally different from customers who spend less in many ways. Instead, you need to isolate the effect of the threshold itself. Regression Discontinuity Design (RDD) does exactly this: by comparing orders just below $50 to orders just above $50 — people who are otherwise very similar except for which side of the line they landed on — you can estimate how much the free shipping offer causally nudged spending upward.

What's at Stake

If the free shipping threshold is genuinely causing customers to inflate their cart values, BasketBloom has strong evidence to keep — or even raise — the threshold. If the effect is negligible, the policy may be giving away free shipping to customers who would have spent that amount anyway, eroding margin without generating incremental revenue. A data-driven answer will directly shape BasketBloom's pricing and promotion strategy for the upcoming holiday season, when shipping costs and order volumes are at their highest. The analysis will also introduce you to one of the most important ideas in causal inference: using a policy boundary as a natural experiment.