If you have already decided automated cart recovery is worth setting up, the next question is how to turn a generic reminder into a flow that can be measured and improved. This guide gives you a practical sequence for email and SMS, shows where Klaviyo, Omnisend and Recart fit, and separates useful AI assistance from features that sound more automatic than they really are.
In short: Allow about two to four hours for the initial build, then two to four weeks to collect enough data for sensible optimisation. Difficulty is moderate. Start with two emails, add one SMS only where you have valid consent, exclude anyone who completes a purchase, and change one variable at a time.
What you will have at the end
An online store owner running a five-person team may currently rely on one generic cart email with the same wording and delay for every shopper. After this setup, the store will have a staged flow that restores the cart, changes the message by customer or cart context, uses AI to refine variants, and reports whether it produces additional completed orders.
A platform can attribute revenue to a message even when the shopper might have returned without it. Use an A/B test or holdout group rather than trusting attributed revenue alone.
Data and privacy: Klaviyo, Omnisend and Recart receive customer identifiers, browsing or checkout events, cart contents and order data from your store. Review the vendor's data-processing terms, make sure email and SMS consent status syncs correctly, and do not add SMS to the flow until your records can prove the customer opted in under the rules that apply in their location.
What you need before you start
You need administrator access to the ecommerce store and marketing platform, a working product catalogue, a tested checkout, and at least one completed test order. You also need separate email and SMS consent fields, access to your existing campaign results, and a person responsible for checking the flow after it goes live.
Turn off any abandoned-cart message already being sent by Shopify, BigCommerce or another store platform before activating a replacement. Klaviyo's setup guidance warns that leaving both systems active can create duplicate reminders.
Which platform fits this workflow
| Klaviyo | Omnisend | Recart | |
|---|---|---|---|
| Best fit | Stores needing detailed segmentation, branching and reporting | Small and mid-size stores wanting fast ecommerce presets | Shopify brands prioritising managed SMS and list growth |
| Recovery channels | Email and SMS flows | Email, SMS and push automations | SMS-first flows with managed support |
| AI role | Email drafting, copy revision and advanced campaign send-time features | Content tools, product recommendations and campaign send-time optimisation | Personalised text campaigns, segmentation support and managed testing |
| Important limitation | Personalized Send Time is not currently available inside flows | AI send-time optimisation applies to email campaigns, not automations | Higher starting cost and a narrower Shopify and SMS focus |
Use the platform already connected to your store unless it cannot support the segmentation or channel you need. Klaviyo offers a free tier for up to 250 active profiles and 500 monthly emails. Omnisend Standard starts at $16 USD per month and Pro at $59 USD per month, with new accounts needing Pro for SMS. Recart's relevant managed List Growth + SMS plan is shown from $500 USD per month.
Pricing was checked in July 2026 and scales with contacts, sends, messages or service level. Compare the monthly bill with the gross profit from genuinely incremental recovered orders.
Step 1: Define the trigger and exit conditions
Decide whether the flow starts when someone adds a product to cart or when they begin checkout. These are different signals. Omnisend recommends separate abandoned-cart and abandoned-checkout automations so a shopper who reaches checkout leaves the earlier cart flow and receives only the more relevant checkout sequence.
Your primary exit condition is a completed order. Add filters that remove recent purchasers before every message, not only at the start of the flow. Also exclude cancelled test orders, staff accounts and anyone without permission for the selected channel.
Step 2: Verify event and cart data
Run a test checkout using a real email address and confirm that the platform records the cart or checkout event, product image, title, quantity, price and return-to-cart link. Then complete the order and confirm the profile exits before the next reminder.
Integration delays matter. Klaviyo notes that some ecommerce integrations do not sync checkout events immediately, which is one reason very short delays can produce incorrect sends. If the event data is late or incomplete, fix the integration before writing clever copy.
Step 3: Build the first two messages
Start with an email two to four hours after checkout begins, then a second email roughly 20 to 48 hours later. This matches Klaviyo's current broad guidance, but it is a starting point rather than a universal rule. Higher-consideration products often need longer delays than low-cost impulse purchases.
The first message should restore the exact cart, answer likely friction and make the return button obvious. The second can add reassurance such as shipping information, returns, stock availability or a clear support route. Do not lead with a discount unless your data shows it improves total profit rather than training customers to abandon deliberately.
Step 4: Use AI for variants, not unchecked decisions
Use the platform's AI writing features to produce three or four variations of the subject line, opening sentence and call to action. Give the tool real constraints, including brand tone, product category, delivery promise and words it must avoid. Klaviyo's Email AI can draft and revise content, but its own documentation says generated sections still need manual configuration for product feeds, images, links and buttons.
Human review remains essential. Check prices, availability, delivery claims, discount terms and the cart link before publishing each variant. AI should reduce blank-page work, not decide what your business promises a customer.
Step 5: Add segmentation and SMS carefully
Create simple branches that can plausibly change the message: new versus returning customer, high versus low cart value, first purchase versus repeat purchase, and domestic versus international delivery. Keep the first version small. Four useful branches are better than twenty conditions nobody can maintain.
Add one SMS reminder only for customers with recorded SMS consent. Keep it short, restore the cart directly and respect country-specific quiet hours, message limits and opt-out requirements. Recart's current guidance is a useful reminder that provider and carrier rules can limit cart-reminder timing and frequency. Recart is the strongest fit of the three when SMS is the main recovery channel and the business values a managed service, while Klaviyo and Omnisend make more sense when email remains central.
Step 6: Measure the baseline before optimising
Run the basic flow long enough to collect a stable baseline, usually at least two to four weeks unless the store has very high order volume. Track delivered messages, clicks, completed orders, revenue per recipient, unsubscribe rate, complaint rate and the share of purchases that happen before the first message.
Where possible, keep a small holdout group that receives no recovery message. The difference in completed orders between the messaged and holdout groups is more useful than the platform's attributed revenue figure on its own.
Important AI limitation: Klaviyo's Personalized Send Time and Omnisend's AI send-time optimisation are currently campaign features, not timing controls inside abandoned-cart automations. For flows, improve timing through controlled delay tests and conversion data rather than assuming the platform chooses the best hour for every shopper.
Step 7: Improve one variable at a time
Test one change against the baseline: first-message delay, subject line, message length, reassurance block, incentive, email versus SMS, or a segment-specific branch. Leave the rest unchanged so you know what caused the result.
Use order rate or profit per recipient as the deciding metric. Open rates can help diagnose a weak subject line, but they do not prove the flow recovered more revenue. Stop tests that increase complaints or unsubscribes even when attributed revenue rises.
Common problems and fixes
- Customers receive reminders after buying: check order-event sync and repeat the purchaser exclusion before every send.
- The cart button opens an empty cart: verify the checkout URL on mobile, desktop and a second device.
- Products are missing or out of stock: repair catalogue sync and suppress unavailable products.
- SMS complaints rise: reduce frequency, confirm consent and check quiet-hour settings.
- AI copy sounds generic: provide product, customer and brand context, then edit it manually.
- Revenue looks high but orders do not change: add a holdout group and compare incremental conversion.
Launch checklist
- Choose one clear cart or checkout trigger.
- Exclude completed purchasers before every message.
- Disable default store reminders.
- Test product details and return-to-cart links.
- Launch two emails before adding extra branches.
- Limit SMS to consented subscribers and allowed hours.
- Have a person check AI-generated copy.
- Document the baseline, test plan and holdout method.
- Assign one owner for weekly monitoring.
Frequently asked questions
Methodology (Real-World, Verified)
We score AI tools against real SMB workflows using named vendor documentation, pricing pages, and independent sources, not enterprise demos. Pricing is verified at the vendor's published rates, with local-currency conversions noted where relevant. Compliance notes reference the legislation and regulatory guidance relevant to each article's region. Every tool is judged on one question: could a business with no dedicated IT department actually pick this up and use it on Monday morning.
Related reading: our AI governance by region.
How many abandoned-cart messages should an ecommerce store send?
Start with two emails. Add one SMS only for consented subscribers when the likely value justifies the extra interruption. Klaviyo's current guidance says two to three cart messages often perform well, but your holdout and complaint data should decide the final number.
Should the first abandoned-cart email include a discount?
Usually not. The first message should restore the cart and remove practical friction. Test a discount later for price-sensitive or high-value segments, then compare profit per recipient rather than conversion rate alone.
Can AI choose the best send time for every abandoned cart?
Not automatically in the main tools covered here. Klaviyo and Omnisend offer AI-based send-time features for campaigns, but their current documentation does not extend those features to abandoned-cart flow delays. Test flow timing directly against completed orders.
Is Klaviyo, Omnisend or Recart best for a small store?
Omnisend is usually the fastest starting point for a smaller store that wants ecommerce presets and simpler pricing. Klaviyo is stronger when detailed customer data, segmentation and branching justify more setup. Recart is better suited to a Shopify brand that wants SMS-first recovery and managed help.
Still choosing the platform? Compare Klaviyo and Omnisend on setup, automation depth, pricing and the type of ecommerce team each one suits.
Compare Klaviyo and Omnisend