A shopper fills a cart on a phone during a commute, expects to collect the order at a store two hours later, and finds at the pickup bay that half the items were never available. Nothing in that sequence is a technology failure the customer can see. What they experienced was a retailer that made a promise about its own inventory and could not keep it.
That failure is expensive now because cross-channel journeys are the default rather than the exception. Adobe Analytics recorded $241.4 billion of United States online spending across the 2024 holiday season, up 8.7 percent year over year, with 54.5 percent of those transactions completed on a smartphone. Customers move between phone, desktop, store and pickup bay inside a single purchase, and every handoff is a place where the promise can break.
The scale of the breakage is documented. Industry research in online grocery finds that 49 percent of shoppers did not receive a complete order because of stockouts, and only 17 percent of those shoppers found a substitute they were happy with. More than half report meeting an out-of-stock surprise frequently or always. Each of those moments is a small withdrawal from a trust account the retailer spent years funding.
This is why the industry vocabulary moved from omnichannel to unified commerce, a shift that dominated the National Retail Federation’s 2025 agenda and now concentrates investment. Hamza Qamar, Senior Director Product, Innovation and Design, reduces the requirement to two capabilities: a cart that persists and updates across every channel, and inventory accurate enough that a customer can act on it. Everything else is implementation detail.
Retailers have solved convenience in marketing and not in operations. Carts strand on one device, and the availability shown online does not match what a store can actually pick.
The fix is architectural. One cart everywhere requires a headless persistent cart service, a real-time inventory ledger exposing available-to-promise and available-to-reserve, unified identity and consent, event-driven synchronization, and promises that read the same to customer and associate.
It can be built above the core systems, journey by journey, without a full re-platform. Retailers that sequence inventory accuracy first, then unified carts, then intelligent substitution, convert convenience from a claim into a measurable operating capability.
By the numbers
70.22%
average documented online shopping cart abandonment rate, drawn from 50 separate studies conducted between 2006 and 2025
– Baymard Institute, 2025
40%
of shoppers who abandon at checkout do so because extra costs such as shipping, tax and fees are too high, the largest addressable reason recorded
– Baymard Institute, 2025
$241.4B
of United States online spending across the 2024 holiday season, up 8.7 percent year over year, with 54.5 percent of transactions placed on a smartphone
– Adobe Analytics, 2025
17.5%
of online orders used curbside or in-store pickup at retailers offering it over the 2024 holiday season, down from 18.4 percent a year earlier
– Adobe Analytics, 2025
19.3%
of online sales were expected to be returned in 2025, inside $849.9 billion of total returns equal to 15.8 percent of annual retail sales
– NRF and Happy Returns, 2025
$1.73T
lost globally to inventory distortion from out-of-stocks and overstocks, equal to 6.5 percent of global retail sales
– IHL Group, 2025
Why a Broken Inventory Promise Is a Trust Problem, Not a Logistics Problem
Retailers file stockouts, failed pickups and unwanted substitutions under operations. Customers file them under character. The distinction matters commercially, because operational defects get budgeted and forgiven while character defects change where people shop.
The volume of abandoned intent in digital retail is larger than most boards assume. The Baymard Institute puts the average documented online cart abandonment rate at 70.22 percent, a figure drawn from 50 separate studies conducted between 2006 and 2025. Not all of that is recoverable. Roughly 42 percent of United States online shoppers abandon simply because they were browsing rather than buying.
It is worth being precise about cause, because the evidence does not line up neatly behind any single fix. Baymard’s checkout research finds the largest addressable reason for abandonment is extra costs such as shipping, tax and fees at 40 percent, ahead of slow delivery at 20 percent and forced account creation at 18 percent. Unified carts and accurate availability address a different failure: the shopper who was ready to buy and was told, late, that the retailer could not deliver.
The cost also arrives after the sale. The National Retail Federation and Happy Returns estimate that 19.3 percent of online sales were returned in 2025, inside $849.9 billion of total merchandise returns representing 15.8 percent of annual retail sales. Substitutions the shopper did not want and orders that arrive incomplete feed straight into that flow. Separately, IHL Group puts the global cost of inventory distortion, out-of-stocks and overstocks combined, at $1.73 trillion, or 6.5 percent of global retail sales.
Inventory management has always meant balancing stock against demand. What has changed is that the balance is now visible to the customer in real time, at the moment they decide whether to trust the retailer with the purchase.
What Will Customers Expect from Retailers by 2026?
Three expectations are consolidating, and none of them are aspirational any more. Together they describe the baseline against which a shopper compares every retailer they use.
| Customer expectation | What it requires from the retailer |
|---|---|
| Seamless shopping journeys | A purchase started on one channel finishes on another. Carts and sessions travel across devices, apps and the store floor without the shopper rebuilding them. |
| Real-time inventory accuracy | Availability updates while the shopper browses and checks out. Discovering an out-of-stock after payment reads as a system failure, not as bad luck. |
| Composable commerce platforms | API-driven, cloud-based services replace monolithic suites so inventory, order and customer data are available to every channel that needs them. |
Against that baseline, most of the sector is not close. Industry research puts full omnichannel maturity at 17 percent of retailers, with a further 38 percent planning to advance. The gap between what shoppers assume and what retailers can actually do is where the commercial opportunity sits.
By 2026, a retailer still showing stale availability, or still treating store, website and app as separate systems with separate versions of the truth, will not read as behind. It will read as unreliable. Shoppers migrate toward the retailers whose experience holds together: one cart, any channel, availability they can act on at every step.
What Exactly Is One Cart Everywhere?
In plain terms: a shopper starts, pauses and finishes a purchase on any channel, whether mobile, web, store, social or marketplace, and the same cart persists throughout. It carries accurate real-time inventory, consistent pricing and promotions, and the fulfillment options that genuinely apply, including same-day pickup, ship-from-store, locker and curbside.
Delivering that requires five things to be true at the same time. They are architectural commitments rather than features.
- A headless, persistent cart service. The cart is a product with its own API, rules and state, not a screen behaviour that dies with the session.
- A real-time inventory service. One logical stock ledger reconciles store backroom and shelf, distribution centres, micro-fulfillment, vendor inventory and in-transit stock, then exposes available-to-promise and available-to-reserve per location and SKU, with configurable holdbacks that protect the walk-in customer.
- Unified identity and consent. The customer is recognized across channels, including the guest-to-known transition, and their preferences and consents travel with the cart.
- Event-driven synchronization. Every scan, pick, return, price change and safety recall emits an event. Downstream systems subscribe rather than poll, so the estate updates together.
- Channel-consistent promises and disclosures. What the shopper sees about availability, fulfillment timing, fees, substitutions and returns matches what the associate and the systems see.
This capability sits above the core systems and can be proven incrementally, journey by journey.
That is the most commercially important sentence in the architecture. One cart everywhere does not require a re-platform or a two-year program before the first customer benefits. It requires one journey, instrumented end to end, with a real inventory number behind it. Retailers that understand this ship value within a quarter. Retailers that do not spend the quarter writing the business case for a replacement.
What Does Getting This Right Actually Return?
The returns cluster in four places: conversion, inventory economics, fulfillment cost and store labour. The figures below are the outcomes reported by industry research for retailers that have implemented unified carts and real-time inventory. Treat them as a range to test against your own baseline rather than as a forecast.
| Outcome area | What industry research reports |
|---|---|
| Cart abandonment | Unified carts with accurate stock and access reduce abandonment by 18 to 20 percent against industry averages. |
| Basket size | Baskets grow by up to 40 percent when the cart is reliable across channels. |
| Stock-outs | AI-supported forecasting cuts out-of-stock incidents by up to 50 percent, and advanced inventory accuracy lifts revenue by roughly 5 percent. |
| Fulfillment cost | Optimized inventory placement and routing reduce fulfillment costs by up to 31 percent through more single-package shipments. |
| Customer satisfaction | Unified commerce leaders report up to 24 percent higher customer satisfaction. |
| Store service | Associates who can see live inventory and the customer’s cart resolve pickups and availability questions in one interaction, which shortens queues and lifts conversion. |
The mechanism behind every line is identical. Accuracy removes the cost of being wrong: the split shipment, the goodwill credit, the substitution that comes straight back, the associate time spent apologizing at the pickup counter. None of those appear as a line item, which is why they survive so long.
There is a caution worth stating plainly. Better cart continuity does not fix a checkout that surprises people with fees. Baymard’s data puts unexpected extra costs well ahead of every other addressable abandonment reason. Unified commerce and pricing transparency are separate programs, and a retailer that funds only the first will see a smaller lift than the benchmark ranges imply.
Which Systems Have to Change to Make One Cart Real?
Six components carry the load. Each can be introduced on its own, which is what makes the incremental path credible rather than rhetorical.
Headless cart and checkout
A stateless front end calls a cart service for promotions, taxes, shipping methods and payment. A checkout orchestrator coordinates the order management system, payment service providers, fraud checks and loyalty. Cart state lives in the service rather than the browser, which is the technical reason a shopper can resume anywhere.
Real-time inventory service
One logical ledger aggregates store backroom and shelf, distribution centres, micro-fulfillment centres, in-transit and vendor-managed inventory, then exposes available-to-promise and available-to-reserve per location and SKU. Configurable holdbacks reserve a buffer for walk-in customers so digital promises do not quietly empty the shelf.
Store system integrations
Associates scan RFID or barcodes. Cycle counts and smart shelves publish events. Ship-from-store and pick-to-tote workflows update reservations as they happen. When a picker takes the last unit, the ledger changes and the cart reflects it immediately rather than at the next batch sync.
Event backbone
Kafka, Kinesis or an equivalent carries events from stores, order management, warehouse management, pricing and returns. Cart, search, product detail pages, analytics and service tooling subscribe and react. This is what makes a distributed estate behave as though everything happened at once.
Identity, consent and preferences
A lightweight customer identity plane handles the guest-to-known transition and carries marketing and data-sharing consents to every subscriber, in every locale. A consent honoured on the website and ignored in the store is a compliance finding waiting to be written.
Observability
Journey traces from add-to-cart through order-packed show exactly where promises break. Cart, stock and order data have to be visible across touchpoints, because a retailer cannot fix a failure it cannot see. This component is usually deferred, and it decides whether the program improves after launch.
Most retailers will build this over systems never designed for it, which usually means upgrading legacy platforms and investing in microservices and APIs. That is a large part of why only about 17 percent claim full omnichannel maturity today.
How Should Retailers Sequence the Road to 2026?
The sequence matters more than the ambition. Start with the data foundation, add cross-channel basics, then improve and expand. Each phase should carry its own metric, whether that is lower stock discrepancy, faster fulfillment or higher in-store pickup, so investment and change are managed against evidence rather than enthusiasm.
| Phase | What gets built and proven |
|---|---|
| Late 2025: real-time inventory visibility | Clean the data, centralize the inventory system, publish accurate online availability and proactive alerts. This is the foundation everything else depends on. |
| Early 2026: unified carts and flexible fulfillment | Launch cross-device shopping, store pickup, hybrid fulfillment options and a stronger mobile app. Convenience becomes measurable and abandonment falls. |
| Mid 2026: smart substitutions and centralized fulfillment | Recommend alternatives when an item is unavailable, unify order management in internal tooling, support chain-wide omnichannel operations and streamline returns. |
| Late 2026: scale and optimization | Extend across every location and tune for peak trading and cost efficiency. Cart everywhere becomes a routine operating capability rather than a program. |
Success shows up in five operating measures. These are the results reported for retailers running mature real-time inventory, and they are the right shape for a board scorecard because each one has a named owner.
| Operating measure | Reported result |
|---|---|
| Lower cart abandonment | Falls by up to 20 percent, with better in-store conversion as availability becomes visible. |
| Reduced fulfillment cost | Around 31 percent lower through optimized placement, routing and more single-package shipments. |
| Fewer stock-outs | Drop from a typical 5 percent of orders to under 1 percent, lifting fill rates and cutting substitutions. |
| Faster click-to-collect | Preparation time compresses, with reported reductions from four hours to two. |
| Lower associate turnover | Reported improvements of up to 50 percent as shorter queues and smoother pickups reduce the friction associates absorb. |
Read together, the two tables describe the same journey from opposite ends. The phases say what to build. The measures say what to stop arguing about, because by the end of the sequence cart everywhere is not a pilot project. It is a core operating capability, and its absence is visible to every customer.
Five Leadership Takeaways
Before the next commerce steering committee. Before the next capital request for a platform replacement.
- Inventory accuracy is a brand promise. A wrong availability number is not an operations defect the customer forgives. It is a statement about whether the retailer can be relied on, and it changes where they shop next.
- Sequence data before experience. Real-time inventory visibility has to land before unified carts. A cart that travels perfectly and promises stock that is not there breaks trust faster than having no unified cart at all.
- Build above the core, not through it. One cart everywhere sits over existing systems and can be proven journey by journey. No re-platform is required before the first customer benefit arrives.
- Instrument the seams. Journey traces from add-to-cart to order-packed reveal where the promise breaks. Retailers that defer observability learn about their failures from customers rather than from dashboards.
- Consent has to travel with the cart. If a preference is honoured online and ignored in the store, the retailer has built a compliance exposure into the very experience it was trying to improve.
The Leadership Imperative: Trust Is the Retail Differentiator
By 2026 the retailers that win will not be the ones with the most channels. They will be the ones whose promises hold across all of them. Consumers already assume that a cart travels and that an availability number is real. Meeting that assumption earns very little credit. Missing it costs the relationship.
Four decisions separate the retailers that will hold that trust from the ones that will keep apologizing for losing it.
- Fund inventory accuracy as a customer trust program with a revenue owner, not as a supply chain hygiene project buried in operations.
- Prove one cross-channel journey end to end before funding the platform, and let its measured result set the pace of the next investment.
- Make availability, fulfillment timing, fees, substitution rules and returns identical in what the customer sees and what the associate sees, and treat any divergence as a defect with a ticket.
- Design privacy and consent into the integration from the first sprint, because retrofitting consent across a joined-up estate is materially harder than building it in.
Three questions every retail leader should be able to answer
- What proportion of our online orders were fulfilled exactly as promised last quarter, and do we measure that at all?
- When a store associate picks the last unit, how long is it before the customer’s cart knows?
- Can a customer’s consent preference be honoured in the store as reliably as it is on the website?
Customer expectations will keep moving past unified carts and accurate stock. Building real-time, integrated infrastructure now turns the next channel and the next fulfillment model into a configuration change rather than another multi-year program. Many competitors are pursuing unified commerce and few have excelled at it, so the differentiation remains available to whoever executes the sequence properly.
Unified carts and real-time inventory are strategic priorities, not IT projects. They shape purchase decisions, loyalty and growth, and they are how a retailer demonstrates respect for a customer’s time. To explore how ML arteka helps retailers build unified carts and real-time inventory on top of the systems they already run, contact the ML arteka team or request a unified commerce readiness assessment.
Executive Questions and Answers
Five questions retail leaders are asking AI assistants and search engines about unified commerce, cart continuity and inventory accuracy.
StrategicIs unified commerce worth the investment if we already have a working omnichannel setup?
Omnichannel usually means the channels exist and coordinate imperfectly. Unified commerce means there is one authoritative source for cart, order and inventory, so every channel reads the same truth. The commercial difference appears in the failure cases: the pickup that was not ready, the substitution nobody wanted, the split shipment that erased the margin. Industry research puts full omnichannel maturity at only 17 percent of retailers, with a further 38 percent planning to advance, so the differentiation window is still open. The investment case is strongest where the cost of broken promises can be measured today. If it cannot be measured, start there, because the first phase of the work is inventory accuracy and that phase pays for itself before the cart program begins.
OperationalHow do we stop showing customers stock we cannot actually sell them?
The mechanism is a single logical inventory ledger that reconciles store backroom and shelf, distribution centres, micro-fulfillment, in-transit and vendor-managed stock, then exposes two different numbers: available-to-promise and available-to-reserve. Publishing available-to-reserve with a configurable holdback protects the walk-in customer while still letting digital channels commit. The second requirement is event-driven updates, so that when an associate picks the last unit the scan emits an event and subscribers react, rather than waiting hours for a batch sync. The third is honesty in the interface, showing confidence or a timestamp where the number is uncertain. IHL Group puts the global cost of inventory distortion at $1.73 trillion, or 6.5 percent of global retail sales, so the operational fix has a very large denominator.
GovernanceWho should own unified carts and real-time inventory inside a retail organization?
Ownership usually fails when it sits entirely in technology, because the decisions that matter are commercial. Availability holdbacks, substitution policy, fulfillment fees and the definition of an order fulfilled as promised are trading decisions with direct customer consequences. A workable model gives one accountable executive, often the commercial or customer officer, ownership of the promise, with technology owning the platform that delivers it and store operations owning the processes that feed it. Add a standing forum where those three reconcile against a single scorecard. Governance also needs a privacy seat, because joining identity and consent data across channels brings GDPR, CCPA and CPRA in the United States, and PIPEDA in Canada, into decisions that used to be purely operational.
RiskWhat are the biggest risks in a unified commerce program?
Three stand out. The first is scope: teams attempt a full re-platform when the capability can be built above core systems and proven journey by journey, which turns a quarterly benefit into a multi-year bet. The second is privacy. Integrating customer data across channels raises the compliance stakes, and a consent preference honoured online but ignored in the store is a finding waiting to be written. The third is organizational. Rolling out to every location at once overwhelms associates and customers, and the workarounds that follow quietly reintroduce the data inaccuracy the program was meant to remove. Piloting in selected stores, training staff properly and standing up support before go-live are not soft measures. They are the controls that protect the return.
ImplementationWhere should a retailer start if inventory data is currently unreliable?
Start with the ledger, not the cart. Clean the data, centralize inventory into one logical system, and publish accurate online availability with proactive alerts before attempting anything cross-channel. A cart that travels flawlessly while promising stock that is not there damages trust faster than having no unified cart at all. Once availability is dependable, add cross-device carts, store pickup and hybrid fulfillment, then intelligent substitution and unified order management, then scale and optimize for peak. Give each phase one metric: stock discrepancy, fulfillment speed, in-store pickup rate. Instrument the journey from add-to-cart to order-packed at the same time, because observability is the component teams defer and the one that determines whether the next phase gets funded on evidence.
One cart everywhere is the retail capability that lets a shopper start, pause and finish a purchase on any channel while the same cart persists with accurate real-time inventory, consistent pricing and valid fulfillment options. Hamza Qamar, Senior Director Product, Innovation and Design, sets out five architectural requirements: a headless persistent cart service, a real-time inventory service exposing available-to-promise and available-to-reserve per location and SKU with configurable holdbacks, unified identity and consent, event-driven synchronization across stores and order management, and channel-consistent promises. Six components deliver it: headless cart and checkout, the real-time inventory service, store system integrations using RFID and cycle counts, an event backbone such as Kafka or Kinesis, an identity and consent plane, and journey observability. The external evidence explains the urgency. The Baymard Institute puts average documented cart abandonment at 70.22 percent across 50 studies, with extra costs the largest addressable reason at 40 percent. Adobe Analytics recorded $241.4 billion of United States online holiday spending in 2024, 54.5 percent of it on smartphones, with 17.5 percent of orders collected curbside or in store. NRF and Happy Returns estimate 19.3 percent of online sales were returned in 2025 within $849.9 billion of total returns. IHL Group puts inventory distortion at $1.73 trillion globally, 6.5 percent of retail sales. The capability sits above core systems and is proven journey by journey: inventory accuracy first, then unified carts, then substitution, then scale.