Your link should go to Shopify if you're promoting a website-only subscription or web-only bundle. But what if your product is sold on both Shopify and Amazon?
Do you want to make sure the offer can be delivered? How much will you earn from this email campaign, after accounting for the costs associated with these orders?
If the product is available on both platforms, Amazon should be tested, because some of your subscribers may prefer to purchase there, but don't default to it just for convenience.
This is different from sending your entire list to Amazon just because it's convenient. Shopify has value in its customer relationship too, so neither of these reasons definitively guides you to where this email should go.
Consider the offer in your email
Write down your audience and what you are asking them to purchase, then click on it for yourself.
If you're sending an email asking subscribers to build a mixed-flavor bundle, you better have a page that allows them to do so. If you're sending a replenishment email, it should take subscribers to the right product and size.
If you're advertising a three-pack offer, it should not take subscribers to a single unit or to a product page that they have to search for.
Use Shopify if your email requires an offer or experience that is only available on your website.
If subscribers have indicated a preference to purchase on Amazon (and you know that the offer is available there), you should test this on a small segment of your list. Make sure Amazon will make the delivery promises that you advertise in your target market (as these may be different).
If you do not have a preference indicated, test on a small scale instead of sending your entire list to Amazon.
Verify on Amazon before sending
Have the account owner verify the specific variation, price, seller offer, delivery promise and stock available for the test. The product page's images and information should appropriately reflect what you are promoting in your email.
Establish who will have the ability to pause your campaign if the offer changes prior to send.
The account owner should communicate with your email team about whether the offer is ready to send or whether it needs to be fixed first, or if Shopify should be used.
"Ready to send" means that the correct offer is in place, that there is adequate stock and that cost assumptions are up to date.
A viable product with a page issue requires fixing prior to send. If Amazon won't make the delivery promises that you advertise in your email, use Shopify if it delivers the promised offer.
Consider this alongside your discount code and mobile preview checks. Attribution tags help you track visitors to a product page that they may not convert on, but they cannot fix your offer.
What those extra orders earn
Hypothetical test, not a real client: same offer, two groups, two channels.
The same offer goes to two similar groups of 1,000 delivered recipients. Shopify generates 20 orders, an estimate of $18 contribution per order. Amazon generates 30 orders, an estimate of $13 contribution per order.
Shopify earns $360 in contribution. Amazon earns $390. That's 36 cents versus 39 cents per delivered email.
Amazon has 10 more orders and a $30 lead. If its contribution estimate falls by $30 while Shopify's stays unchanged, the groups tie. That's a flimsy reason to shift all future campaigns.
In this scenario, contribution begins with revenue minus discounts, minus product cost, minus channel fees, minus fulfillment, minus shipping, minus expected returns, minus campaign specific expenses.
Use the same buckets on both sides. If you're excluding fixed overhead, exclude it from both sides. Don't credit either channel with future customer value you haven't validated.
Returns require care. If you're already estimating with an allowance, replace the allowance with better data as it comes in. Don't double-count the returns.
I'd keep the full list sends going to the same place while exploring this outcome, and only shift the destination once the contribution difference holds up across a larger send. Another limited test might be worthwhile.
Ten more orders, for 30 extra dollars, doesn't tell us how statistically reliable this difference is, or whether shifting is worthwhile.
Agree what would change your mind
Before launch, agree what additional contribution per delivered email would change your mind about shifting, and write that number down so you have a clear line to act on. This is a business requirement, not a statistical confidence level.
In this scenario, we have an apparent three-cent improvement per delivered email. Does that still beat your requirements, across a credible range of unsettled costs? If not, hold the destination change.
If you don't have enough data to make a call, hold it for that reason.
Once costs settle and you have enough of a credible comparison to act on, you can decide whether you want to run another limited test, or go broad.
Where possible, randomly split a similar, permissioned audience between the two destinations. Keep the messaging similar, and note any differences you can't avoid.
If you're comparing different discounts or pack sizes, you're comparing more than checkout destinations. A handful of orders isn't enough to change out the next full list send.
Give the campaign owner a quick snapshot of who received it, what the offer was with links, what the cost estimates were, what the return allowance was, when they said they'd review, and what return and other reporting gaps you have to fill.
Before anything else, name the decision maker and what it would take to continue, fix, or stop. This doesn't have to be a second project. Someone needs to be responsible for making the decision.
Get reporting squared away before launch
Amazon Attribution lets eligible advertisers measure Amazon shopping activity associated with off-Amazon marketing like email in supported markets. Amazon Attribution has a 14 day attribution window.
Plan separate tags for each campaign variation and test the whole link path before send.
For setup and reconciliation work, ALFI has an Amazon Attribution guide that covers the Amazon side of things. Share the tags and destination plan with whoever owns the email campaign.
Document how Shopify and your email platform credit sales as well. Adding their credited revenue to Amazon's risks double-counting orders.
An attributed sale isn't a settled contribution figure. Review attribution and returns separately. At the end of 14 days, you don't necessarily know all your return costs.
An approved AI tool is handy for sorting supplied numbers and hunting down missing cost fields. Keep inputs inside your data permissions and have someone manually review the calculations.
Don't let the machine invent a return rate or patch a reporting gap with an assumption nobody approved.
An Amazon purchase alone doesn't give you permission to add the buyer to your email list.
There's one reconciliation limit even after you're done. Comparing two email destinations doesn't mean the email drove additional demand. Some recipients might have bought anyway. A separate no-send holdout, if feasible and measurable, addresses that question.
