The Free Shipping Policy Review Framework

FREE SHIPPING POLICY REVIEW — original Growthcraft worksheet
Decision / owner / intended launch date:
Currency / eligible zone / delivery service / period:
Current threshold / proposed threshold / below-threshold fee:
Eligibility basis: merchandise after discounts, before tax and shipping:
Exclusions: bulky goods, special profiles, subscriptions, remote zones:
Threshold inclusive? Refunds and discount stacking policy:

1. DATA CONTRACT — analyst and finance
Order export / source version / count reconciliation:
Basket values / product and payment costs / carrier expense:
Cost coverage gaps / actual versus estimated costs:
All eligible historical orders retained, including negative contribution?
Hold if source totals, currency or eligibility cannot reconcile.

2. FIXED-ORDER REPLAY — analyst
Orders / current and proposed free-shipping counts:
Newly qualifying / losing eligibility:
Current contribution / proposed contribution / change:
Check change equals fee times (lost eligibility minus newly qualifying).
No assumed conversion or basket response in this step.

3. RESPONSE HURDLE — finance and growth
Static loss to recover:
Net contribution per genuinely additional order / cost evidence:
Whole additional orders required / share of historical orders:
Top-ups, cannibalization, split orders and acquisition costs handled separately:
Unknown input is unknown; a plausible assumption is not a forecast.

4. CUSTOMER AND TEST CONTRACT — UX, operations, growth
Product/cart/checkout wording and eligibility agree?
Discount change, threshold equality, mobile layout and excluded zone tests:
Stable assignment unit / eligible visitor denominator / exposure logging:
Primary contribution metric / conversion, returns and delivery guardrails:
Decision rule / observation window / rollback trigger / accountable reviewer:
Status: hold / propose controlled test / approved by named human
Evidence references and unresolved questions:

Treat free shipping as a policy with a cost

This original Growthcraft framework is for D2C founders, ecommerce growth leads, finance partners and checkout teams considering a different shipping threshold. It separates the cost on orders that already exist from the behavioral response you hope to create. Use it to prepare a controlled test, not to declare an optimal threshold from average order value.

It is unsuitable for stores whose shipping charge depends on several profiles, parcel weights or destination rules unless those populations are separated first. It does not determine taxes, carrier tariffs, legal disclosures or an experiment sample size. Keep those decisions with their respective owners.

Start with an eligibility contract

Define one currency, zone, shipping service and merchandise subtotal. Record whether discounts reduce the qualifying subtotal and whether equality qualifies. The companion implementation uses post-discount merchandise value excluding tax and shipping, with equality qualifying. Check the store's actual configuration rather than assuming this convention matches it.

The analyst supplies anonymous order bands or individual aggregate rows, finance defines variable cost coverage, operations checks carrier expense, and UX checks the promise displayed to shoppers. The growth owner names the decision and test population. One person can hold multiple roles, but an unowned assumption remains a gap.

Shopify's September 26, 2026 shipping guide revisits thresholds and cost coverage. That is a timely editorial signal, not proof that a threshold change will improve this store. Read the Shopify guide.

Four gates before a launch

1. Reconcile: retain every eligible historical order, including low-margin and negative-contribution orders. Match counts to the scoped export. Hold for missing currency, costs or shipping eligibility.
2. Replay: apply current and proposed thresholds to the same unchanged baskets. Sum merchandise less variable costs and carrier expense plus customer shipping revenue. Validate the change against the newly qualifying and losing-eligibility counts.
3. Set a hurdle: divide the static loss by defensible net contribution per genuinely additional order, rounding up. Document costs and separate top-ups by existing buyers from new orders. A zero-contribution incremental order cannot recover a positive loss.
4. Design the test: align product, cart and checkout promises, then define assignment, eligible visitors, observation window and contribution guardrails. Do not select the winner from conversion or average order value alone.

No generic pass score is assigned. Reconciliation and comparable scope are logical requirements. Commercial decision limits come from the business's risk and economics, not an invented industry benchmark.

Worked example

Synthetic EUR example. A threshold falls from 75 to 50 with a fee of 5 below it. There are 100 orders at basket/cost/carrier values 40/24/6, 80 at 60/36/6 and 40 at 90/54/7. Existing contribution falls from 4,500 to 4,100. Eighty orders now get free shipping; each loses 5 of shipping revenue. Carrier expense is already present under both policies.

If a genuinely additional order contributes 12 after all marginal costs under the proposed policy, 34 such orders are needed to cover the 400 loss. That is 15.45% of the historical 220 orders. It is not a forecast or a 15.45 percentage-point conversion lift. If the 12 input has no supporting cost evidence, hold the economic recommendation even though the arithmetic passes.

Make the promise understandable

Baymard's research warns against communicating free shipping only through a site-wide banner. Read the product-page research. In this workflow, check eligibility beside relevant product/cart decisions, explain exclusions and ensure the checkout agrees. This is a design checklist, not a claim that adding a progress bar causes sales lift.

Test an order exactly at the threshold, a discount that drops it below, an excluded zone, an empty basket and mobile text wrapping. A top-up suggestion should explain the real additional spend and delivery benefit, not imply that the extra item is free.

Where the model stops

A frozen replay contains only historical purchasers. It cannot see abandoned baskets, changed traffic or new customers attracted by the offer. Raising a threshold can improve replayed contribution while reducing demand. Lowering it can increase conversion yet lose contribution. Treat those responses as hypotheses for a separate test and keep shipping-profile complexity visible.