Abandoned shopping carts in an online store a retargeting strategy

Over the past few years I have observed the same pattern: on average 7 out of 10 carts remain unpaid. International e‑commerce studies confirm this, and on mobile this figure is often higher. Each such cart: your already-earned user interest and already-paid traffic, which can easily be turned into revenue if you build your online store’s retargeting strategy correctly.

The problem of abandoned carts hits unit economics: CAC rises, CLV falls short, and the payback period for advertising stretches. An audit of dozens of projects has shown me that root causes are always multilayered: from UX and delivery methods to trust in payment and site speed. Are you ready to challenge conventional approaches and build a system where the recovery rate grows and the cost of a recovered order decreases?

I suggest reading the material to the end. I will break down abandoned cart recovery into concrete steps: from event architecture and email sequences to dynamic remarketing, attribution and legal requirements. I rely on the practice of BUSINESS SITE: solutions that we designed and scaled for online stores across different price points and purchase frequencies in Ukraine.

Abandoned cart rate

Abandoned carts are sessions where a user added items, started checkout or reached the checkout step but did not complete payment. Key metrics are abandonment rate (the share of uncompleted carts) and recovery rate (the share of recovered and paid carts). I believe they should be interpreted together through the lens of revenue and incrementality: not just “how many were recovered”, but “how many were recovered in addition to natural purchases”.

Reasons for abandonment are usually distributed like this: UX/checkout (extra fields, lack of one‑click recovery, poor mobile optimization), cost and delivery times (“Nova Poshta”, courier, pickup), site speed and errors on the payment step, trust in payments (absence of recognizable payment logos, 3‑D Secure from “PrivatBank”/Monobank). To investigate, I connect session replay and heatmaps: behavioral analytics shows points of frustration, address data discrepancies, bottlenecks in choosing a “Nova Poshta” branch, delays in card validation.

Prioritization is influenced by cart value: carts with a high cart value deliver a greater incremental lift with the same contact budget. In BUSINESS SITE projects I always set rules: high‑value gets earlier and multichannel touches, while low‑value receives more restrained frequency capping and shorter time‑decay windows.

Retargeting for abandoned carts

The goal of the strategy is to manage recovery rate, incremental revenue, and reduce CAC while growing CLV. I set KPIs: cost per recovered order, ROAS for retargeting, recovery rate by segments, share of incremental orders (incremental lift). It’s important to synchronize this with lifecycle marketing: immediately after leaving we work on immediate recovery, then on winback and loyalty, without undermining margins with discounts.

I base the approach to timing on time‑decay recovery windows. First contact within 30–60 minutes, then: after 20–24 hours, a third touch at 48–72 hours with a stronger incentive. For repeat buyers the window is wider, for first‑time visitors it’s shorter. The channel combination looks like this: email → SMS → push → dynamic remarketing in Google/Meta, where value‑based bidding helps push real cart values into bids.

Prioritization is dictated by segments: cart total (segmentation by cart value), customer status (new vs returning), intent signals (begin_checkout, payment_info_added), and uplift modeling: a forecast of the touch potential. In one of our cases the BUSINESS SITE team reduced touches for users with high natural conversion, increased intensity for borderline segments, and achieved a greater incremental lift with fewer impressions.

Cart recovery automation

The architecture is built on event tracking: add_to_cart, view_cart, begin_checkout, add_shipping_info, add_payment_info, purchase. It’s important to capture webhooks and real‑time events without delays: the lower the latency in the trigger system, the higher the chance of recovery. For this I use server‑side queues and retries so notifications go out even during traffic spikes.

Scenario templates have been tested dozens of times: an immediate soft reminder, a reminder with social proof/delivery details, then an offer of a limited incentive (a bonus, free shipping via «Nova Poshta» for orders above a threshold). For high-margin assortments we add a series of emails with dynamic content: recommendations, bundle offers, and alternatives when an item is temporarily unavailable.

Examples of email sequences and triggers

I use a three-level logic: soft reminder → urgency → incentive. First email: after 30–60 minutes: «You left items in your cart», with one‑click recovery and deep‑linking directly to the checkout step. Second: after 24 hours: we emphasize benefits, delivery times, availability of «Nova Poshta» branches, and promise fast processing. Third: after 48–72 hours: we add a limited incentive or a more favorable bundle.

I pick send times by testing: I try different trigger delays and run A/B tests on subject and CTA. Metrics: open rate, click‑to‑recover, revenue per recipient, recovery rate. I always personalize the email based on the catalog: dynamic content shows specific items, their prices and availability, using product feed management and feed optimization.

For mobile-first segments, deep‑linking into the app with a saved cart works especially well. On the web I add one‑click recovery: token-based authorization in the link and instant cart restoration without searching for items.

Email retargeting and deliverability

The email structure is shaped by practice: product photo, price, benefits, delivery options (“Nova Poshta”, courier), a clear CTA – “Complete order”. A dynamic recommendations block adds cross-sell/upsell, and a noticeable offer deadline stimulates action. I keep the design concise so the email loads quickly and doesn’t get caught in additional spam-filter checks.

Testing is a mandatory routine. I combine A/B and multivariate: subject, preheader, creative, the “delivery today/tomorrow” block, a payment button via “PrivatBank”/Monobank. I measure uplift vs baseline and set criteria for statistical significance. As ESPs, Klaviyo, Mailchimp and Iterable work well: they support triggers, segmentation and hashed email matching.

Dynamic ads and remarketing

Dynamic remarketing shows the user exactly the products they viewed or added. This requires a high-quality product feed and synchronization with advertising platforms. I link the feed to Google Merchant Center, perform feed optimization (titles, attributes, availability), and synchronize prices and availability.

In Google Ads and Meta I launch dynamic product ads with event parameters (view_item, add_to_cart, initiate_checkout). Smart bidding and value-based bidding help the algorithms account for cart value and margin. For creatives I use DCO: personalization of images, prices and callouts by segment to increase ROAS for retargeting.

Retargeting: SMS, push and programmatic

I select the channel by segment. Email is the baseline, especially for returning audiences and average order values. SMS is for high-intent and high-ticket customers, and for urgent reminders before an offer expires. Push notifications are a fast, low-cost contact for mobile audiences and web push subscribers. For each channel I calculate cost per recovered order and choose the mix with the best ROAS.

SMS campaigns require proper integration with providers and strict handling of consent. I recommend storing consents in the CMP, setting frequency limits and sending messages with personalized deep-linking. For push: I set up the SDK in the app, web push integration in the browser, frequency capping and “quiet” reminder scenarios.

Message and catalog personalization

Personalization increases in levels: from basic (name, product) to behavioral (browsing history, discount-sensitivity) and catalog- and context-oriented (availability at the required branch of «Nova Poshta», seasonal triggers, local holidays). This approach significantly increases the recovery rate without excessive discounts.

Recommendations are especially noticeable: cross-sell (accessories), up-sell (a higher model), similar products and bundle offers with free shipping when cart value thresholds are reached. I apply value-based segmentation: segments with a high probability of conversion receive more benefit-focused content, segments with low margins – cautious incentives that protect margin.

Cross-device identification and tracking

Client-side tracking often loses signals due to blockers. I consider a hybrid optimal: server-side tagging for resilience in a cookieless environment and client events for interface responsiveness. This provides a more stable data layer for retargeting and correct event import.

For identifiers I use hashed identifiers: email/phone hashing with a salt, storage in encrypted form (encryption at rest and in transit) and strict access control. PCI compliance and careful handling of PII are mandatory requirements. For cross-device identification, cross-device matching helps: hashed matching, identity resolution in the CDP, and data clean rooms for secure joins.

I design the architecture to be scalable: real-time event pipelines, queues, retries, latency monitoring, alerts for trigger delays. This setup withstands sale spikes and preserves reaction speed: emails and ads go out while intent is hot.

Attribution and measurement of retargeting ROI

Attribution strongly affects conclusions. Last click often underestimates email and overstates paid retargeting, data‑driven and multi‑touch distribute credit more fairly. I use incremental lift measurement and incrementality testing: I create holdout‑groups to separate actual recovery from natural purchases and channel cannibalization.

KPIs include cost per recovered order, ROAS, incremental revenue, LTV and CAC for cart recovery. I track UTM and source tracking for all touches, use cohort analysis: I evaluate cohort behavior over time since the cart was abandoned. Benchmark recovery rates by vertical are useful: low‑ticket – 10–20%, mid‑ticket – 12–25%, high‑ticket: 5–12% with quality automation.

I make business decisions based on incremental value: if retargeting delivers high ROAS but low incremental lift, I reallocate budget to segments with greater uplift, and shift some touches to winback and loyalty.

Testing and optimization of creatives

Creative testing is an ongoing process. I start with A/B, then bring in multivariate for large audiences. The metrics lineup: open rate → CTR → click‑to‑recover → recovery rate → revenue per contact. The key is not only the winner by clicks, but also by recovered orders and profitability.

Ad frequency requires care. Frequency capping by channel and by user protects against ad fatigue and preserves ROI. I use dynamic creative rotation, DCO and creative personalization to prevent messages from getting worn out. At the same time I test different “offer end” windows and track how response changes when the deadline shifts.

Privacy and consent management

Regulatory frameworks, GDPR, CCPA, ePrivacy. I implement a consent management platform (CMP), maintain consent storage and an audit trail: who, when and what they consented to. For retargeting I apply privacy-first approaches: hashed identifiers, minimal data sets, cookieless techniques and server-side processing.

Martech tools and services

I build the stack from these categories:

• ESP: Klaviyo, Mailchimp, Iterable, triggers, segmentation, deliverability.
• CDP/DMP: segmentation, identity resolution, channel orchestration.
• Ad platforms: Google Ads remarketing, Meta Ads (Facebook/Instagram), programmatic DSP.
• Feed managers: product feed management and merchant center synchronization.
• SMS‑providers: integration, two‑way messaging, reports.
• Push: web push integration, mobile SDKs.
• analytics: attribution, cohort analysis, session replay, heatmaps.

Selection criteria: integration with ecommerce platforms (Shopify, Magento, BigCommerce and local CMSs), scalability, support for server‑side tracking and APIs. I proactively design API catalog integration, webhooks, and synchronization with Merchant Center. If the business also sells through Prom.ua or «Rozetka», consistency of prices and availability between catalogs is important.

Improving recovery rate in e-commerce

Case 1 (home goods, average ticket 1500–2000 UAH). Initial data: abandonment rate ~78%, recovery rate 6%, absence of server‑side events, fragmented email sequences. Hypotheses: speed up first touches, personalize the feed and move to CDP‑driven orchestration. The BUSINESS SITE specialists implemented server‑side tagging, triggers with delays of 45 min / 24 h / 60 h, DCO creatives in Google/Meta, SMS for high‑value carts (5000+ UAH). Results after 8 weeks: recovery rate 16.4%, incremental revenue increase by 21%, cost per recovered order −28%, retargeting ROAS +34%.

Case 2 (e‑commerce pharma segment, frequent repeat purchases). Objective: minimize discounts and rely on “Nova Poshta” delivery + pickup. The solution we developed at BUSINESS SITE: triggers focused on speed and availability, recommendations of alternatives when out of stock, integration with CRM to account for patient status and seasonality. Outcome: recovery rate growth from 9% to 15%, share of recovered orders without a discount: 73%.

Checklist for implementing a similar solution

• Set up events and server‑side tagging.
• Segment by cart value and customer status.
• Launch triggers 45 min / 24 h / 60 h with one‑click recovery.
• Enable dynamic product ads with value‑based bidding.
• Introduce SMS for high‑value carts.
• Set up incrementality testing with holdout groups.

Step-by-step cart recovery checklist

Phase 1. Funnel audit (1–2 weeks).

• Tasks: assessment of abandonment rate, site speed, checkout flow, event map.
• Owners: product/marketing + development.
• Metrics: baseline abandonment, recovery, page response time.

Phase 2. Data collection and segmentation (1 week).

• Tasks: setup of add_to_cart/begin_checkout/purchase, UTM and source tracking, cohort analysis.
• Owners: analyst + development.
• Metrics: event completeness, catalog tagging, cart value thresholds.

Phase 3. Channel and tool selection (3–5 days).

• Tasks: ESP/CRM/CDP, SMS provider, ad platforms, feed management.
• Owners: marketing + IT.
• Metrics: integration readiness, CMP, hashed identifiers.

Phase 4. Automation of scenarios (1–2 weeks).

• Tasks: email‑sequences, SMS/push, suppression rules, frequency capping.
• Owners: CRM marketer + ESP specialist.
• Metrics: trigger latency, deliverability, opt‑out rate.

Phase 5. Testing and launch (2–3 weeks).

• Tasks: A/B and multivariate, creative and timing tests, holdout groups.
• Owners: performance marketer + analyst.
• Metrics: uplift vs baseline, ROAS, cost per recovered order.

Phase 6. Measurement and scaling (ongoing).

• Tasks: incrementality testing, value‑based bidding, DCO, expansion into programmatic.
• Owners: head of e‑commerce + marketing.
• Metrics: recovery rate by segment, incremental revenue, LTV/CAC, retention cohorts.

• One‑click recovery + deep‑linking into checkout.
• Early soft reminder + dynamic content.
• Segmentation by cart value with priority for high‑value.

• Temporary drop in open rate during domain warm-up.
• Increased costs for launching DPA until feed calibration.
• Misestimation of ROAS without proper attribution: holdout helps.

Frequently Asked Questions

Conclusion and call to action

I am convinced that a combination of multichannel automation, catalog-based personalization, and rigorous analytics yields a steady increase in recovery rate and reduces CAC. My colleagues have repeatedly confirmed this on BUSINESS SITE projects: early triggers, DCO, and value-based bidding generate incremental revenue without “overheating” with discounts.