Quick answer
An abandoned-cart SMS should give an eligible shopper a useful route back to the purchase. That may be a working checkout link, an answer to a product question or an approved offer. Start by identifying the event and the obstacle. A shopper who viewed a product is not in the same state as one who entered checkout and stopped.
Source context: Postscript: recovery event and control definitions; Attentive: coordinating checkout and existing cart journeys.
Recover the decision that stopped the checkout
Suppose your store already sends a cart reminder and you add a checkout flow with a similar delay. A shopper can qualify for both. Each message may look reasonable in its own builder, while the combined experience looks careless. The first job is to define ownership of the next message and the conditions that cancel it.
This guide builds one inspectable flow. It gives you the event map, a reply example, a discount hurdle and a launch test. Timing and permitted frequency must follow your provider’s current rules and the programme’s approved consent and compliance design. A generic three-text sequence is not a safe universal starting point.
Write the event definition in plain language
| State | What happened | What to verify before messaging |
|---|---|---|
| Product viewed | A known subscriber viewed a product | Identity, eligibility and whether they moved further |
| Added to cart | A product entered the cart | Whether checkout or purchase has since occurred |
| Checkout started | The shopper began checkout | Whether purchase completed and which return link works |
Postscript’s current recovery guide separates product-view and add-to-cart activity from a checkout-started abandonment. Its documentation also describes opted-in eligibility, cancellation and frequency controls. Use the precise event supplied by the integration, then verify it in an actual test checkout.
Keep an event receipt with a test identifier and time. Confirm when it reaches the messaging platform, which fields arrive and what happens after a purchase. If the store cannot reliably cancel a scheduled reminder when an order completes, copy polish will not fix the underlying experience.
Give the next message one owner
Define precedence across browse, cart and checkout activity. When the shopper moves forward, the earlier path should be cancelled or suppressed as designed. A completed purchase should remove the person from recovery messaging for that purchase. Opt-out and other applicable suppressions must remain effective across the programme.
Attentive’s Shopify checkout-started guide presents different setups depending on whether an existing cart-abandonment message is already running. It includes purchase checks and spacing to prevent duplicate contact. Treat those as examples of orchestration, not as evidence that one delay is best for every merchant.
The launch sheet includes constructed cases for a purchase just before send, a second cart, an email already sent and an unmatched reply. Give each expected outcome an owner. A test is complete when the observed outcome matches the intended behaviour, not when a green switch says the flow is active.
Write the message for the next useful action
Make the store identifiable and the route clear. Use the actual return link supplied for that event and integration. Test it on a phone and after the cart changes. A link that lands on an empty cart can turn an otherwise useful reminder into another task for the shopper.
Suppose a shopper replies, “Does this include the refill, or do I buy that separately?” The useful next turn is the approved product answer. A coupon does not resolve the uncertainty. If the answer depends on a variant or bundle, ask which item they mean before making a claim.
| Observed obstacle | Useful next move | Avoid |
|---|---|---|
| Unclear bundle contents | Use the current item or variant description | Guessing what the shopper selected |
| Delivery uncertainty | Check destination and approved delivery information | A guaranteed arrival date without support |
| Price objection | Consider an approved offer within the margin rule | Automatic discount on every reply |
| Order already placed | Stop recovery for that purchase | Another “complete your order” reminder |
Offer a reply route only if someone or something owns the response. If questions go to a staffed queue, make the hours and escalation path work in practice. If an automated service answers them, test missing information, conflicting policies and the limit of its authority.
Make a discount earn the margin it gives away
Take a $100 order with $60 in variable costs. It contributes $40 before the recovery programme cost. A $10 discount leaves $30, assuming the other costs stay unchanged. At 100 orders the original contribution is $4,000. The discounted treatment needs 134 orders to exceed that amount, because 133 orders contribute only $3,990.
The required increase is 34 orders in this example. It is not a forecast of what an offer will achieve. If a higher basket changes fulfilment costs, if refunds differ or if the discounted group uses more messages, rebuild the calculation with those costs. An attributed revenue increase can still leave contribution lower.
Use a consistent eligible population and outcome window when comparing the offer. Assign the shopper consistently if they can create several carts. Keep the no-offer baseline, the actual discounts used and the extra programme costs. The calculator can help compare bills; the ROI article covers the separate question of incremental contribution.
Run the launch cases before expanding the flow
The launch sheet opens with a completed checkout-flow example and expected outcomes. It then gives you an editable test log. Keep the trigger, eligibility, cancellation, message owner, link behaviour and reply owner together. Those details are hard to recover from a collection of screenshots after a launch goes wrong.
Run a purchase before the delay expires, an opt-out, a repeat checkout and a reply that needs a person. Record the actual send or suppression beside the expectation. Investigate mismatches before widening the audience. An empty observed-result field means the case has not been tested, even if the intended design is clear.
Download the cart-flow launch sheet. It includes a completed example and an editable version. Get the short printable brief.
After launch, review sends that should have been suppressed, unresolved replies, failed return links and contribution over the defined window. Keep the first observation period narrow enough that the team can read the exceptions. Add complexity only when you can still explain the route a specific shopper took.
The next test may be the reply, not the incentive
If checkout follow-up is already live, read the replies before adding more reminders. Identify which shoppers are asking a product or purchase question and whether the current flow gives them an answer. That is a concrete place to evaluate conversational recovery.
Zobi’s EQ models are designed to read a shopper’s question in context and choose a relevant reply from merchant-approved information, within the programme parameters the merchant approves.
Use a checkout question, its approved product source and the handoff rule as the starting brief. Then evaluate the proposed Zobi flow on answer relevance, source support and contribution under your actual programme costs. The discount example is a decision tool, not a claim about Zobi merchant results.
Practical questions about abandoned-cart texts
How many messages should the flow send?
Follow the provider rules and the approved programme design. Postscript’s reviewed guide specifies one message for its cart-recovery automations within 48 hours. That vendor rule is not permission to copy a multi-message sequence elsewhere.
Should the first reminder contain a discount?
There is no universal answer. Compare the obstacle, existing offer exposure and contribution hurdle. A useful answer or working return link may be the relevant change. Test the actual offer against a suitable baseline.
What counts as a successful launch?
The correct eligible event enters, purchases and suppressions cancel as intended, the link works, replies have an owner and the measurement window is defined. A flow being active is only one part of that result.
Evidence and measurement limits
Primary event and orchestration documentation was checked on 28 October 2026. Provider instructions can change, so recheck the applicable setup before implementation. They are not a complete statement of legal requirements. The design, launch cases and contribution hurdle are original teaching material under explicit assumptions.
Publisher disclosure
Zobi publishes this guide and offers conversational commerce services. Product descriptions, worked examples and source documentation have distinct evidential roles. This is an unranked editorial guide, not independent product testing.
Rene writes about ecommerce retention and conversational commerce at Zobi. This guide combines primary documentation with practical decision tools.
Tell us about your store.
Start growing revenue