Welcome
Confirm the value of joining, set expectations, and help a new subscriber find the right product or category. Do not make the first relationship entirely about a coupon.
Home / Learn / Shopify Marketing
Ecommerce growth · lifecycle
The ecommerce lifecycle marketing plan to build first
Start with customer moments the store can recognize and serve. Add personalization only when the signal is clear, useful, consented, and easy to explain.
Direct answer
Build four automations first: welcome, abandonment, post-purchase, and winback.
- Tie each flow to a real customer state, suppress completed goals, and cap overlapping messages.
- Personalize only with declared preferences or recent behavior that clearly improves the experience.
- Measure repeat purchase, time to second order, retained gross profit, unsubscribes, and complaints, not email-attributed revenue alone.
01 · Build the core sequence
Four flows cover the first lifecycle jobs
Abandonment
Separate browse, cart, and checkout intent. Remind the shopper of the exact context, suppress purchasers, and stop when the reason to message disappears.
Post-purchase
Confirm use, care, delivery, setup, and the likely next need. Cross-sell only when it helps the purchased product succeed.
Winback
Use the normal reorder or repeat-purchase window, not a generic number of days. Diagnose why the relationship cooled before reaching for a discount.
Klaviyo’s own flow guidance highlights welcome, abandoned cart, post-purchase, and winback as common starting flows. The implementation must still reflect the store’s products, purchase cycle, consent rules, and message volume.
02 · Personalize with a permission test
Useful personalization is visible, proportional, and explainable
A shopper usually accepts personalization when the input is obvious and helpful. Examples include size, stated interests, a recently viewed product, a recent purchase, or a replenishment interval. It becomes unsettling when the data source is hidden, the inference is sensitive, or the message reveals more surveillance than service.
Declared preference
The customer told the store the category, size, channel, or frequency they prefer.
Recent behavior
The context is clear, the inference is modest, and the message stops when it is no longer relevant.
Sensitive inference
The store cannot explain why it knows, the attribute is sensitive, or the targeting would surprise a reasonable customer.
Consent and frequency are part of the design. Segment on subscription status and channel permission, keep suppression logic readable, and prevent multiple flows from colliding with campaigns in the same short window.
03 · Measure customer value
Measure customer value, not claimed email revenue
Use cohorts so a newer customer group is not compared carelessly with an older one. Read retention through three connected views.
Repeat purchase, time to second order, purchase frequency, and cohort retention.
Average order value, contribution margin, refunds, and discount dependency.
Unsubscribes, complaints, collisions, and the share of orders bought with a code.
Shoppers can browse more and buy less when traffic quality falls, price resistance rises, product information is incomplete, mobile performance weakens, or the path creates uncertainty. Browse abandonment is a symptom. Before adding reminders, compare engagement, cart progression, checkout errors, source, device, customer type, inventory, and shipping constraints.
Use discounts for a named job: first-order risk reduction, inventory clearance, bundle economics, loyalty recognition, or a bounded winback test. Protect full-price demand with expiration, eligibility, exclusions, and holdout measurement. If every campaign has a code, waiting becomes rational customer behavior.
Keep a lifecycle control sheet
For each flow, write the customer state, trigger, consent requirement, exclusions, exit event, message count, minimum interval, owner, and commercial measure. Add a collision check for campaigns or other flows that can reach the same person.
Review the system as a customer would experience it. Test a new subscriber, browser, cart abandoner, first-time buyer, repeat buyer, refunded buyer, and unsubscribed customer. Confirm what each profile receives and what it does not receive.
Questions this owner resolves
Short answers, with the conditions that change them
Q086
Which email automations should an ecommerce store build first?
Build automations in customer-state order: welcome for new subscribers, browse or product-interest follow-up when consent and event quality support it, cart and checkout abandonment, post-purchase education, replenishment when timing is predictable, and winback after a defensible lapse window. Start with the flows closest to revenue and customer service, suppress people who completed the goal, and cap overlapping messages before adding more branches.
Use this rule: Launch a flow only when its trigger is reliable, the customer state is meaningful, the message changes the next best action, and exit and suppression rules prevent stale or contradictory sends.
Example: If one shopper qualifies for welcome, browse, and cart messages in 24 hours, give cart the highest priority, pause lower-priority sends, and exit all abandonment messages immediately after purchase.
Field note: Map flows as a state machine before writing emails; most lifecycle damage comes from collisions and late exits, not from missing another subject-line test.
Sources1
Q087
How do you personalize ecommerce marketing without being creepy?
Personalize with data the customer knowingly provided or behavior that is recent, relevant, and necessary to improve the next choice. Explain material data use, offer controls, minimize retention, and avoid sensitive or surprising inferences. Product preferences, replenishment timing, language, or declared size can be useful; inferred health, financial stress, pregnancy, or identity traits can create harm and distrust even when technically available.
Use this rule: Use a personalization signal only if a customer could reasonably expect it, it improves a specific decision, consent and policy allow it, and the experience still works when the signal is absent or wrong.
Example: Use a declared men's size 10 to filter available shoes; do not infer a medical condition from browsing and target it. Set and document a behavior window, such as 30 days, then test whether older signals still help.
Field note: Add a 'why am I seeing this?' review to every personalization brief; if the truthful explanation sounds invasive, the signal is not ready for production.
Q088
Why are shoppers browsing more but buying less?
More browsing with fewer purchases can come from lower-intent traffic, weaker product-market fit, price or delivery friction, stock and variant gaps, trust problems, or checkout failure. Separate acquisition mix from on-site behavior: compare product-view-to-cart, cart-to-checkout, and checkout-to-purchase rates by channel, device, new versus returning customer, and product. Diagnose the first falling stage before adding discounts or retargeting.
Use this rule: If engagement falls first, fix traffic or merchandising; if carts fall, fix product value and availability; if checkout completion falls, fix total cost, delivery, trust, account, payment, and form friction.
Example: Sessions rise 40% and product-view-to-cart remains 8%, but checkout completion drops from 50% to 35%; the evidence points to checkout economics or friction, not a need for more product-page traffic.
Field note: Segment 'just browsing' from blocked buying by asking a one-question exit survey and matching the answer to observed funnel behavior, not by treating all abandonment as recoverable intent.
Q089
Which retention metrics matter most for an ecommerce brand?
Track retention by acquisition cohort: repeat purchase rate, time to second order, orders per customer, purchase frequency, average order value, gross margin, cumulative contribution-margin lifetime value, lapse or churn, and unsubscribe and complaint rates. Report them by product, channel, and first-order offer. Email-attributed revenue alone is not retention because it can credit demand that already existed and ignore discount, return, and service costs.
Use this rule: Use a retention metric only if it identifies a customer cohort, a time window, and an economic outcome; prioritize contribution margin and repeat behavior over platform-attributed revenue.
Example: Of 1,000 first-time buyers, 280 place another order within 90 days, so 90-day repeat purchase rate is 28%. Compare their cumulative gross profit, returns, discounts, and acquisition source before calling the cohort valuable.
Field note: Time to second order is often the earliest operational signal; it can reveal the right education, replenishment, or service intervention before annual LTV stabilizes.
Q090
How do you use discounts without training customers to wait?
Give every discount a defined job, eligible audience, duration, margin floor, and holdout: overcome a first-purchase hurdle, move specific inventory, increase bundle size, reward a valuable behavior, or reactivate a truly lapsed customer. Avoid predictable sitewide cycles and perpetual countdowns. Measure incremental contribution profit and repeat full-price behavior, not code redemptions or discounted revenue alone, and display truthful terms and reference prices.
Use this rule: Run a discount only when the incremental orders and downstream value are expected to exceed lost margin, leakage to customers who would buy anyway, and added returns or service cost.
Example: A $100 item with $40 cost yields $60 gross profit. At 20% off it yields $40, so the promotion must create enough incremental profitable orders to replace one-third of gross profit per discounted order.
Field note: Measure coupon leakage by comparing eligible recipients, exposed non-recipients, and actual redeemers; a successful-looking code can mostly subsidize existing demand.
Primary and observed sources
What this guide is grounded in
Official documentation supports platform and search requirements. Third-party pricing and practice pages are cited as observed market examples, not universal facts or proof of ranking causation.
Continue the decision