How to Test Your Ecommerce Welcome Offer: Discount, Gift or Free Shipping?
By Trayan Hristov · Thrivelia · Published 2 November 2026

A welcome offer can produce more subscribers and still leave the business worse off. The extra orders may carry a larger discount, higher shipping costs or a gift that costs more than expected. Some customers may have bought without the incentive.
I would test an ecommerce welcome offer against the decision it needs to improve: acquiring customers whose purchases justify the cost. The signup rate helps explain the result. The commercial outcome decides whether the offer deserves to stay.
This guide compares percentage discounts, gifts, free shipping and an offer without a purchase incentive. It focuses on the offer itself. Our Klaviyo A/B testing guide covers the broader testing process.
Download the editable lifecycle workbook (XLSX) and open the Welcome offer tab. Replace the example inputs with your own evidence and record the decisions your team needs to make.
Choose two offers with a clear commercial reason
Start with the current offer and one credible alternative. Write why the alternative might improve customer value. A useful gift may make the first product easier to enjoy. Free shipping may address a checkout concern. A smaller discount may preserve contribution while still helping the buyer decide.
Test a no-incentive route when the brand can give someone a useful reason to subscribe without a purchase reward. Examples might include product-selection guidance or access to relevant launches. Describe the exchange honestly.
Offer | Question to test | Cost to make visible |
Percentage or fixed discount | Does the increase in retained orders justify the reduction in selling price? | Discount across all qualifying orders, including buyers who might have paid full price |
Gift with purchase | Does the gift add enough useful value to change the buying decision? | Landed gift cost, picking, packing, extra shipping and return treatment |
Free shipping | Does removing the shipping charge improve the total economics? | Shipping subsidy by destination, weight and basket |
No purchase incentive | Can the content or access promise attract valuable customers on its own? | Content or service delivery cost, where it changes between treatments |
A real test question: a discount or a useful accessory?
In our published Manuel Dreesmann case, the popup work compared a 10% discount with a free ring wallet. The record confirms the two options and the launch work. It does not establish an offer-test winner.
The useful lesson is the question. One offer reduces the price; the other adds a product. Those have different economics and may appeal to different buyers. A brand needs to follow both paths through signup, welcome, checkout and the eventual retained order.
A gift's retail price is not its cost to the business. Nor does a low unit cost make it a good offer automatically. It still needs to be desirable, available and practical to fulfil.
Define the audience before building the variants
State who can enter the test. That might be new visitors in specified markets who are eligible for a first-order offer. Decide how existing customers, returning subscribers and visitors already exposed to another incentive are handled.
Use concurrent random assignment with a consistent allocation. Keep a visitor's treatment stable where the testing setup permits, and document the limitations across devices, browsers and consent choices. A person seeing both offers can blur the comparison.
Keep the other parts of the experience comparable: targeting, display timing, device rules, number of steps, consent choices and the surrounding campaign treatment. Change the copy needed to explain the offer accurately. If several other elements change, interpret the result as a comparison of complete experiences.
For a commercial readout, use contribution per assigned eligible visitor when your measurement can reliably connect assignment with orders. Include people who never subscribe. Comparing only subscribers conditions the analysis on a behaviour the offer itself can change.
Make the welcome deliver the promise
Record which offer the subscriber saw and route them to the matching message. Klaviyo's offer-testing instructions describe saving an offer value as a hidden form property and using it to split the welcome journey. Check the actual saved value and exclude overlapping welcomes.
Use an experiment-specific property or event record where appropriate. A mutable profile field is useful for current routing, but it is not automatically a durable history of original assignment. Klaviyo's custom-property documentation explains that later submissions can replace a property's previous value.
Place test orders through each complete route. Check the code, qualifying products, minimum spend, expiry, gift availability, shipping destination and offer combinations. Confirm what happens when someone buys before the next welcome message.
Have support and fulfilment approve the operational promise. A free gift that requires a manual workaround can create a hidden cost in both the result and the customer experience.
Read the form report and the business result separately
Klaviyo provides revenue metrics for form variations. They associate orders with identified form submissions or completions within a configured lookback window. They do not automatically measure every assigned visitor, long-term contribution or additional revenue caused by the offer.
Save the reporting settings before the test. Changing the form lookback configuration can alter historical associated-revenue totals. Privacy and tracking restrictions can also affect the events available for analysis.
Use the native report to diagnose the journey: exposures, submissions, orders and associated revenue. Use reconciled store and cost data for the commercial comparison. If visitor-to-order assignment is incomplete, state the measured population and investigate whether coverage differs between variants.
Our attribution and incrementality guide explains the broader distinction between credited revenue and causal evidence.
Calculate contribution without subtracting costs twice
Choose one consistent definition. A practical order-level starting point is sales after discounts and refunds, less relevant product, payment, fulfilment, shipping, gift and return-related costs. Exclude taxes collected for authorities. Document any shipping income and how recoverable returned inventory is treated.
If discounts are already removed from net sales, do not subtract them again. Count the gift at its actual economic cost, and include the operational costs it adds. Aggregate the same cost categories for both test groups over the same observation window.
The table below is fictional. It demonstrates a readout, not a result achieved by Thrivelia.
Measure | Discount route | Gift route |
Assigned eligible visitors | 10,000 | 10,000 |
Subscribers | 1,000 | 800 |
Retained first orders | 300 | 280 |
Total contribution after defined costs | $6,000 | $6,400 |
Contribution per assigned visitor | $0.60 | $0.64 |
The gift route has fewer subscribers and orders but higher observed contribution per visitor. That does not establish a reliable winner. The uncertainty around the difference, assignment quality and completeness of the cost data still matter.
Decide the measurement window and stopping rule in advance
Plan the test around the expected purchase delay, order volume and commercially meaningful difference. Allow for refunds and gift fulfilment. Avoid ending the test as soon as a favourable number appears.
Klaviyo's native form significance assessment concerns form submit rate. An automatic winner on that measure does not establish a contribution winner. Review the ending settings so the platform does not silently make a different decision from the one your test is meant to support.
If volume is limited, test fewer variants and retain uncertainty. An inconclusive result can still reveal broken redemption or a costly fulfilment issue. It does not justify presenting a small observed difference as a proven uplift.
Check who comes back after the first order
For a later view, compare contribution over equal observation windows and examine genuine repeat purchases. Keep the offer assignment and first product attached to the cohort. Use the same follow-up period for both groups.
Repeat rate among buyers is a useful diagnostic, but the groups of buyers were partly created by the offers. The randomised comparison per originally assigned visitor remains the stronger basis for judging the overall policy when tracking supports it.
Our first-to-second purchase guide covers equal-age measurement. Do not call a recently acquired customer a failed repeater before they have had the agreed opportunity to return.
The test brief to complete before launch
Write the commercial decision and the two offers.
Define eligibility, assignment and the treatment of repeat visitors.
Verify routing, redemption and fulfilment for both paths.
Set the primary contribution measure and diagnostic metrics.
Record the observation window, cost definition and stopping rule.
Name the owner who will turn the result into an operating decision.
The strongest welcome offer is the one your evidence supports for that audience and business. Start with a comparison you can interpret, then make the whole journey deliver what the form promised.

.png)



Comments