Loctek relies on owned warehouses as Amazon click costs climb

Loctek faces an e-commerce marketplace where Amazon cost per click has climbed continuously across all categories for multiple years. According to 36kr, Loctek marketing manager Zhang Yifei tracks that steady advertising inflation alongside higher commission rates from marketplaces like Shopee and Lazada. Margins are vanishing. To remain profitable, participants at a 36kr event reported that sellers must keep CPC under 30% of sales.

Traffic subsidies no longer disguise unprofitable operations. Participants at the 36kr event noted that customer acquisition costs for brands like Temu are now six times higher than SHEIN's early benchmarks.

When paid acquisition ceases to carry balance sheets, the operational role of brand fundamentally shifts. As woshipm author Yuan Zhen argues, true brand value resides neither in promotional reach nor sheer sales volume, but in the choice probability and repurchase certainty stored in customer minds. Surviving without paid platform placement requires tangible operational proof that endures when artificial intelligence shopping assistants strip away sponsored visibility.

Measuring brand resilience when paid traffic stops paying for itself

According to 36kr, Loctek cross-border marketing manager Zhang Yifei reported that cost per click on Amazon has risen continuously across all categories for multiple years. Event participants at 36kr noted that sellers on the platform generally need to keep cost per click under 30% of sales to maintain profitability. Commission rates have climbed as well across Shopee, TikTok Shop, and Lazada. Alongside rising advertising and logistics costs, these platform fees leave most cross-border sellers facing stagnating or declining profits.

Renting platform demand has become punitive. Participants at the 36kr event stated that customer acquisition costs for platforms like Temu are now six times higher than early acquisition costs were for SHEIN.

For direct-to-consumer brands, higher acquisition expenses erode first-order net margins until initial checkouts produce losses. Repeat purchase rates therefore decide whether customer lifetime value ever covers that initial outlay. Marketing theory cited by woshipm holds that retaining an existing customer costs one-tenth of acquiring a new one. Furthermore, woshipm contributor Yuan Zhen argues that choice probability and repurchase certainty stored in the brains of consumers represent a brand's true asset, outranking promotional reach or volume.

Pop Mart demonstrated the financial mechanics of that retention in its 2025 interim financial report. In the first half of 2025, registered members generated 91.2% of the company's total sales. The brand's member repeat purchase rate reached 50.8% during that same period. When paid search auctions erase margins, customer retention ceases to be an abstract communications goal and becomes a baseline survival metric.

Passing the physical shelf test without online explanatory crutches

Digital storefronts routinely cushion weak packaging with interactive explanation. Zhang Yifei argues in 36kr that while e-commerce platforms secure initial orders quickly, independent direct-to-consumer websites drive user retention and repeat purchases. Emerging brands frequently misread those early platform metrics. Shuze Brand Notes observes in woshipm that companies mistake livestream conversion spikes and viral clips for lasting brand loyalty. Once an item leaves algorithmic feeds, those promotional crutches vanish.

Initial retail shipments merely fill distributor pipelines and store setups without reflecting actual end-consumer sell-through. Physical retailers often provide trial placements without guaranteeing replenishment orders. Without algorithmic nudges, the product must sell itself.

Physical presentation faces brutal comparisons. In offline audits reported by woshipm, Shuze Brand Notes monitored a brand with high online popularity that failed to attract shoppers despite occupying an entire dedicated display rack. A consulted functional beverage brand performed well online but required perpetual discounts and dedicated floor staff to move offline inventory. Its package size and shelf price proved fundamentally uncompetitive beside adjacent competitors. Retail channels evaluate shelf allocations on whether an item expands total category shelf value rather than siphoning volume from neighboring stock-keeping units.

Relying on traditional intuition rather than direct capital investment drives the trial-and-error costs of expansion, argues Ocean Stone managing partner Yang Fei in 36kr. When teams define specifications around livestream formats, offline economics break. Shuze Brand Notes argues in woshipm that companies must benchmark planned retail prices and unit specifications against adjacent offline competitors during initial project conception rather than after finalizing the product.

Structuring technical feeds and infrastructure for AI shopping agents

According to 36kr, mainstream platforms such as Amazon and TikTok Shop are embedding conversational question-answering systems and AI shopping guides into their search engines to replace traditional keyword advertising systems. At the same time, third-party AI shopping agents bypass marketplace storefronts to connect consumers directly with brands. When software conducts product discovery, automated evaluation replaces visual browsing. As 36kr notes, machine recommendation logic weighs structured product data, user reviews, brand content richness, and domain credibility to select products.

According to 36kr, Alibaba Cloud Senior Product Expert Wang Xiao argues that domain names and digital certificates will serve as the foundation for AI agent identity verification. Secure DNS records confirm brand authenticity before an autonomous agent transacts.

To establish verifiable merchant identities, businesses must secure baseline technical credentials. Alibaba Cloud and Alibaba.com offer registration across nearly 400 domain suffixes to anchor external trust. Through tools like the Wanxiaozhi AI Website Builder (万小智AI建站) and integrations with OKKI, companies link domain registration, site building, CRM, and marketing. These verified technical registries give automated agents direct access to clean product feeds.

This structural shift matches the expanding scale of automated decision-making. According to geekpark, Wu Yongming projected that machine thinking will supply more than 1,000 times the volume of human thinking. Alibaba built an Agentic Cloud architecture using AgentCore and CPFS to support that computational demand, while introducing Agent Studio on Qianwen AI to unite models, tools, memory, and governance. In geekpark, author Zhang Peng argues Alibaba is positioning itself as economic infrastructure for machine intelligence rather than an operating system. Structured feeds determine which inventory gets chosen.

Cross-Channel Resilience: Traffic Acquisition vs. Generative Content vs. Trust Infrastructure

Controlling generative output before automated content erodes trust

Deploying automated infrastructure lets teams generate external assets at industrial volume. Alibaba Cloud launched the Wanxiaozhi AI Website Builder (万小智AI建站) alongside OKKI to package domain registration, site building, CRM, and marketing into automated pipelines, according to 36kr. Yet woshipm's author Yuan Zhen argues that artificial intelligence provides mechanized capacity for mediocre output rather than an immediate lift in quality. Costs do not disappear during this transition. Instead, expenses shift directly from the initial writing desk to downstream evaluation and revision.

That downstream drag is substantial. A survey of 1,150 full-time U.S. knowledge workers cited by woshipm revealed that 40% received hollow, surface-complete AI work in the prior month, spending nearly two hours to remediate each task.

Left unmonitored, synthetic generators insert embarrassing factual blunders into customer-facing channels. Commercial systems have rendered peanuts growing on trees, chili peppers taller than adult humans, distorted product structures, and extra human fingers. Reputational exposure runs deeper than surreal imagery. Tone-deaf campaigns-such as a personal care brand promoting laundry detergent with "Mom, you use it first," or an outdoor apparel label detonating fireworks in a fragile Himalayan ecosystem-highlight the hazards of unanchored messaging. AI systems lack the context to foresee public blowback.

Scale alone cannot fix this judgment gap. According to geekpark, Alibaba's Qwen team is exploring recursive self-improvement while targeting models spanning 5 trillion to 10 trillion parameters. Even so, geekpark noted Wu Yongming comparing current artificial intelligence tools to the electric light bulb in 1882. Models mimic cadence, but they cannot accept responsibility when business interests clash. As Yuan Zhen observed in woshipm, falling production costs make human accountability the rarest asset in enterprise management. Editorial review is an operational firewall against brand erosion.

Decision Framework for AI-Era Commerce Infrastructure and Brand Asset Allocation

• A cross-border brand faces climbing advertising costs and rising commission fees across marketplaces like Amazon, Shopee, TikTok Shop, or Lazada. Cap marketplace ad spend strictly-such as keeping Amazon cost per click (CPC) under 30% of sales to preserve profitability-while routing retention efforts to an independent direct-to-consumer website. Marketplaces can secure initial orders, but independent sites are necessary to capture user retention and repeat purchases against customer acquisition costs that can run six times higher than early industry benchmarks.
• A marketing team seeks to deploy generative AI tools to rapidly produce commercial product imagery and customer-facing collateral. Do not treat generative models as an autonomous end-to-end production pipeline. Build mandatory human evaluation checkpoints into the workflow to catch structural errors-such as chili peppers taller than adult humans, peanuts on trees, or extra human fingers-and allocate staff time for review, recognizing that resolving a piece of substance-lacking AI output takes knowledge workers an average of nearly two hours.
• An emerging direct-to-consumer brand with viral livestream sales plans an initial expansion into brick-and-mortar retail channels. Benchmark the stock-keeping unit's planned package size and retail price directly against adjacent physical shelf competitors during initial project conception. Initial shipments only reflect channel pipeline stocking, and offline retailers evaluate placements on whether a product expands total category shelf value rather than relying on permanent promotional discounts or floor staff to clear inventory.
• A brand wants its products recommended by third-party conversational AI shopping agents and generative search engines. Transition discovery investments from keyword bidding toward structured digital infrastructure. AI shopping agents bypass traditional search rankings and evaluate brands using structured product data, verified reviews, domain credibility, DNS, and digital certificates as the foundational verification layer.

Mapping capital allocation across category risk and repeat cycles

Capital allocation across product categories depends on whether an industry rewards emotional differentiation or distribution mechanics. Writing for woshipm, Chen Hao defines commercial brand power across a four-tier system: cognitive power, trust power, premium power, and repeat purchase power. In high-margin lifestyle categories, emotional value commands extreme markups. Basic domestic yoga pants in China sell for around 100 yuan, yet individual Lululemon items sell for 1,000 yuan. As Chen points out, that premium reflects lifestyle framing around health and discipline rather than raw textile costs.

Traffic inflation breaks brands that cannot command that pricing leeway. At a 36kr event, attendees noted that acquisition expenses for storefronts like Temu ran six times higher than SHEIN's early customer acquisition costs. Burning capital on ephemeral buyers is ruinous.

Where categories cannot sustain luxury premiums, capital must target operational retention mechanisms instead. Gillette sustains recurring revenue by selling inexpensive razor handles and monetizing consumable replacement blades. In high-consideration durable goods, memory erosion is fatal. As woshipm contributor Yuan Zhen observes, one frequent failure occurs when consumers recognize a brand but overlook it during the actual purchase decision. To bridge that gap, Loctek, the first A-share listed company in the ergonomic furniture and health equipment industry, signed basketball player Paul George for endorsements.

Channel selection anchors this split between transactional volume and durable asset value. According to 36kr, Zhang Yifei argues that while e-commerce platforms secure initial orders rapidly, independent direct-to-consumer websites deliver actual customer retention and repeat purchases. Without repeat loyalty or pricing power, high acquisition expenses eventually consume the business.

Auditing a product on a physical shelf exposes structural weaknesses that digital promotions conceal. According to woshipm, research by Shuze Brand Notes examined a functional beverage brand that succeeded online but stalled offline because its package size and shelf price could not compete with adjacent products. Moving inventory required continuous promotional markdowns and dedicated sales staffing. Early channel shipments easily disguise this friction because initial orders only reflect pipeline stocking and store setup.

Offline retail channels evaluate whether an additional stock-keeping unit lifts total category shelf value rather than cannibalizing existing items. With chains developing proprietary private labels, uncompetitive packaging loses floor space.

Shuze Brand Notes argues that consumer brands must benchmark planned retail prices and specifications against adjacent physical competitors during initial conception. Designing solely around livestream or e-commerce formats leaves inventory vulnerable once customer acquisition shifts to unassisted environments.

For readers outside China

• Availability: Alibaba Cloud and Alibaba.com offer domain name registration services across nearly 400 domain suffixes alongside the Wanxiaozhi AI Website Builder and Agent Studio on the Qianwen platform. While Alibaba Cloud targets operating more than 20 gigawatts (GW) of data center capacity worldwide by 2032, the availability, international localization, and overseas commercial onboarding for Wanxiaozhi and OKKI toolsets are not disclosed in sources. Loctek's public overseas warehouses are physically operational in the United States, spanning the East Coast, West Coast, and central regions to support 1-to-2-day order delivery.
• Pricing: Specific subscription tiers, API pricing, and licensing costs for Alibaba Cloud's Wanxiaozhi AI Website Builder, OKKI CRM, Agent Studio, and domain registry packages are not disclosed in sources. Related economic benchmarks mentioned in the source material include domestic Chinese yoga pants retailing around 100 yuan compared to 1,000 yuan for Lululemon items, Amazon advertising viability requiring CPC to stay under 30% of sales, customer acquisition costs for platforms like Temu running six times higher than SHEIN's early costs, and customer retention costs estimated at one-tenth of new customer acquisition costs.
• Closest Western equivalents: Shopify; HubSpot; Amazon Bedrock Agents; ShipBob
• Data residency: The source material highlights that domain names, DNS, and digital certificates will serve as the foundational identity verification layer for AI agents, but it does not disclose specific cross-border data residency protocols, server regions for overseas customers, or regulatory compliance mechanisms outside mainland China.

Sources

• 36kr 从「流量依赖」到「信任复利」，跨境电商的下一个增长公式是什么？｜超级会氪厅 https://36kr.com/p/3982902160063239
• woshipm 新消费的品牌资产，正在被货架重新计价 https://woshipm.com/share/6450876.html
• 36kr 流量成本攀升，品牌信任为什么是跨境电商的终极壁垒？丨超级会氪厅活动回顾 https://36kr.com/p/3995657950433158
• woshipm 当所有品牌都有一支"无限内容团队"，品牌还剩下什么？ https://woshipm.com/share/6445530.html
• woshipm 究竟什么是真正的品牌力？ https://woshipm.com/share/6457094.html
• geekpark 拆解下阿里的 AI 经济学，与它的下注 https://geekpark.net/news/370865