How to Test a Free Shipping Threshold Without Chasing AOV

Average order value cannot tell you which free-shipping threshold to keep. A larger basket can contain lower-margin products or require a more expensive parcel. Before changing the number in a banner, build a candidate sheet that shows where the proposed threshold is affordable, where it fails, and what evidence would justify rollout.

This guide focuses on that decision sheet and its sensitivity checks. Our free-shipping dropshipping guide covers the broader shipping setup and mixed-supplier policy. Here, the objective is to turn a plausible threshold into a testable business decision without calling a spreadsheet result “optimal.”

Use the formula to screen candidates, not choose a winner

For a simplified illustrative model, assume merchandise contribution before shipping and handling is 40% of net merchandise revenue. Shipping and handling cost $8, and the desired remaining contribution is $12. The qualifying basket that meets that target is ($8 + $12) / 0.40 = $50. At $50, the assumed merchandise contribution is $20; subtracting $8 leaves $12.

That result depends on the assumptions. It excludes acquisition, after-sales losses, fixed overhead and tax collected for remittance. Payment costs must either be included in the 40% rate or deducted separately, never both. Write the cost definition beside the formula so another person can reproduce it.

Now vary one assumption at a time. At a 30% contribution rate and the same $8 fulfillment cost, the basket needed for the $12 target becomes $66.67 when rounded up to cents. At 40% but $13 fulfillment cost, it becomes $62.50. Neither sensitivity implies customers will buy those basket amounts. They show how quickly the original $50 screen stops protecting the target.

Replace the average basket with three physical baskets

Create rows for actual combinations customers can assemble. Record item-level net prices, variable merchandise costs, packed weight, package size, fulfillment source and destination. Then calculate the remaining contribution using the complete basket. Ask the packing or fulfillment owner to check the row before sending it to marketing.

Consider these fictional baskets at a proposed $50 threshold. All figures are USD, with no customer shipping revenue because each basket qualifies. Merchandise contribution is already net of the assumed merchandise-related variable costs; shipping and handling are deducted afterward. Acquisition, returns and overhead remain outside this example.

• Compact basket: $50 revenue, $20 merchandise contribution and $8 shipping and handling leave $12.
• Bulky basket: $60 revenue, $24 merchandise contribution and $15 shipping and handling leave $9.
• Lower-margin basket: $70 revenue, $21 merchandise contribution and $10 shipping and handling leave $11.

The biggest basket is not the strongest contributor, and all three satisfy the same revenue threshold. If the target is $12, only the compact example passes this preliminary screen. Raising the threshold alone might not repair the problem: shoppers could reach it with more low-margin or bulky products.

Record a decision beside each failing row. Options to investigate include changing the assortment eligible for the offer, improving packaging costs or choosing a different delivery proposition. Do not advertise exclusions until the checkout can enforce them. A cost concern in a spreadsheet is not an implemented shipping rule.

Estimate the cost of orders that would already qualify

A lower threshold can subsidize customers who would have purchased under the existing policy. Add a historical replay column before forecasting any new sales. For each past order, ask what the proposed offer would have changed while holding that order's basket and destination fixed.

For example, suppose 120 historical orders would have received a newly waived $5 delivery charge. With otherwise unchanged receipts and costs, the replay reduces contribution by 120 × $5 = $600. This is a fictional accounting sensitivity, not an estimate of the experiment's causal effect. Customers could change their baskets or purchase decisions under the new offer.

Keep that replay separate from the live test. Do not subtract the $600 again from experimental results that already include the lower shipping receipts. Its purpose is to reveal the subsidy you are committing before assuming that new demand will cover it.

Make the treatment specification reproducible

Write down the current threshold, candidate threshold, eligible destinations, delivery service, product rules and dates. State whether the implementation is a shipping rate or a promotional discount. For Shopify free-shipping discounts, the documented minimum uses product prices after applicable combinable discounts and excludes taxes. Do not assume a banner's displayed basket total is the same qualifying amount.

Attach controlled checkout evidence for values immediately below, exactly at and above each threshold. Include the bulky and lower-margin baskets from the candidate sheet, a discounted basket and an excluded destination. Record the actual available service and amount due, not just whether a “free shipping” message appeared.

If you already use Eggflow Instant Upsell, its listed clickable coupon banners can introduce another incentive into the journey. Keep their coupon rules and placement stable while testing the threshold, or explicitly define the combined change as the treatment. The app's banner and redemption reporting capabilities do not establish randomized assignment, actual carrier costs or every shipping eligibility rule.

Define the decision before opening the results

Use a testing setup that can keep eligible visitors in a consistent assigned group and connect outcomes to that assignment. Define eligibility before exposure; do not analyze only shoppers who reach the threshold. Keep non-buyers in the assigned denominator. If stable assignment cannot be implemented, describe a sequential comparison as observational and account for concurrent holiday promotions, traffic changes and stockouts.

Record three separate stopping rules. An operational stop covers incorrect charges or broken eligibility. A spending stop limits the experiment's affordable subsidy exposure. The planned analysis rule sets the sample and observation period needed to assess a commercially useful difference with appropriate uncertainty. A universal number of days cannot replace a design based on your traffic and outcome variation.

The primary decision can use contribution per assigned eligible visitor, with a consistent cost definition and acquisition-spend treatment across groups. Add completed-order rate, shipping-related support issues and the distribution of contribution across the predefined basket types. Do not choose a winner from tiny favorable subgroups after seeing the results.

A Shopify Community threshold discussion illustrates merchants' concerns about higher baskets and lost orders. Its experiences and suggested test durations are anecdotes, not proof of a universal threshold or sufficient sample size. Use it to identify questions, then answer them with your own experiment.

Reconcile costs before rollout

At the planned readout, label unreceived carrier invoices and immature returns explicitly. Apply the same provisional-cost method to each group, then schedule a second review when those costs mature. Preserve the early snapshot so a later adjustment is visible rather than silently rewriting the initial result.

Choose one of three documented outcomes: roll out within the tested scope when the evidence and operating checks support it; retain the current policy when the candidate fails; or gather better evidence when uncertainty remains material. Record which basket types or destinations were not tested. A successful compact-parcel result does not authorize the same promise for bulky deliveries.

Sources checked November 11, 2026. All basket amounts, cost rates, targets and replay counts are illustrative, not merchant results or promised profit. The reused feature image illustrates shopping and delivery costs.