The Sales Funnel Was Invented in 1898. TikTok Shop Killed It in 30 Seconds.

In 1898, an advertising executive named Elias St. Elmo Lewis wrote a column for a trade magazine about how to capture a reader's attention. He proposed a sequence: first get their attention, then hold their interest, create desire, and finally prompt action. Advertisers called it AIDA. Marketers called it the purchase funnel. For 128 years, it was the operating model for selling things to people.

Last Black Friday, TikTok Shop generated over $100 million in a single day. Most of those purchases followed a path that Lewis would not recognize: a user scrolling through videos saw a creator demonstrate a product, tapped a button overlaid on the video, and bought it. No search query. No comparison shopping. No reading reviews. Attention, interest, desire, and action compressed into one algorithmic moment.

Three months later, at the National Retail Federation's Big Show in January 2026, Shopify announced what it called "the agentic commerce platform": infrastructure that connects merchants to AI agents acting on behalf of consumers. In this model, the consumer doesn't scroll, doesn't search, doesn't tap. The AI agent identifies a need, evaluates options, and purchases. The human never enters the funnel because there is no funnel to enter.

These are not incremental improvements to the same model. They are two fundamentally different visions of what buying something looks like, and both assume Lewis was wrong.

What the Funnel Actually Assumed

AIDA worked because it described a real constraint: information was scarce and attention was directional. You learned about products through ads, asked a salesperson questions, compared maybe two or three options, and decided. Each stage had friction. Each stage took time. Marketers could measure how many people entered each stage and how many dropped out. The metaphor held because the physics held.

By 2009, McKinsey's David Court and colleagues had already challenged this. Their "Consumer Decision Journey" paper argued that consumers don't move linearly through a funnel. They loop between evaluation and active consideration, and brand loyalty can short-circuit the process entirely. Google's internal research team took it further in 2020 with the "Messy Middle" concept, showing that consumers oscillate between exploration and evaluation in patterns that look nothing like a funnel.

Neither of these frameworks predicted what happened next: platforms that don't just disrupt the funnel but eliminate stages from it entirely.

The Compression Ratio

Here is a way to measure what each platform actually does to the purchase journey. Call it the funnel compression ratio: the number of traditional funnel stages that a platform's dominant commerce model traverses between initial exposure and completed transaction.

Traditional retail traverses all four stages (awareness, consideration, conversion, and retention) over days or weeks. A customer sees an ad, visits a store, compares options, buys, and maybe comes back. Four stages, high friction, high intent at each transition.

Amazon search compresses this to roughly three stages. By the time someone types "noise cancelling headphones" into Amazon, awareness has already happened elsewhere. Amazon's job starts at consideration: showing relevant options, surfacing reviews, comparing prices. The purchase happens on the same page. Retention is handled by Subscribe & Save and Prime lock-in. Amazon eliminated the top of the funnel and built a machine for everything below it.

Now Amazon is compressing further. Rufus, its generative AI shopping assistant launched in 2024, replaces keyword search with conversation. Instead of typing product names, customers ask questions: "what do I need for cold weather golf?" Rufus suggests product categories, compares options, and answers questions about specific items using review data and community Q&As. Consideration becomes a conversation, and the conversation leads directly to purchase. Three stages compressed toward two.

TikTok Shop compresses to roughly 1.5 stages. A user scrolling their For You Page sees a creator applying a skin care product. The creator isn't running an ad. They're earning commission through TikTok's Creator Affiliate Program. The product is tagged. One tap leads to checkout. The user didn't search for the product. They didn't compare alternatives. They weren't shopping at all. TikTok's own research found that 48% of its subscribers have expressed interest in purchasing directly within the app. The platform calls this "shoppertainment." The accurate term is impulse capture at algorithmic scale.

Shopify's agentic commerce model takes this to its logical endpoint: zero stages. An AI agent identifies that you're running low on coffee beans based on your purchase history and consumption rate, evaluates three options across price, roast preference, and shipping speed, and orders. You receive a notification that coffee is arriving Tuesday. You never entered a funnel because the agent operated outside of one.

China Already Proved This

Before TikTok Shop launched in the US in September 2023, its Chinese counterpart Douyin had already built a $281 billion livestream commerce industry. According to Influencer Intelligence, 60.9% of Chinese social commerce sales in 2023 came through livestreaming alone. Statista reported that 84% of Chinese consumers had shopped on social media by 2022, compared to about 36% of US internet users.

China's social commerce market was valued at $363 billion in 2022. ResearchAndMarkets projects it will reach $2 trillion by 2028, growing at 33.7% annually. WeChat Mini Programs alone had 450 million daily active users. Pinduoduo's group-buying model turned customers into salespeople through viral discount sharing, growing the platform to 263 million monthly visits and $4.99 billion in quarterly revenue.

The US market is behind, but the gap is closing faster than most predictions assumed. US social commerce was roughly 5% of online retail in 2025, totaling an estimated $107 billion, up from $39.5 billion in 2021. The global social commerce market is projected to exceed $1 trillion by 2028.

The Race Beneath the Race

Each platform is building toward a different version of post-funnel commerce, and the differences reveal what each company actually believes about consumer behavior.

Amazon believes search is dead. Not web search, but product search. Rufus converts the typed query into a conversation, and the conversation into a curated shortlist. Amazon has also deployed AI-generated review highlights (summarizing thousands of reviews into a paragraph), AI-powered size guidance, and AI-generated product listings. The store is becoming a system that understands what you need and narrows choices for you. Amazon's advertising business, now exceeding $50 billion annually, funds this transition. It doesn't need you to browse. It needs you to convert.

TikTok believes intent is manufactured, not found. In TikTok's model, you didn't want the product before you saw the video. The video created the want. This inverts the traditional funnel completely: instead of awareness leading to interest leading to desire, the algorithm generates desire first and the purchase is an afterthought. Awareness and desire are the same event.

Temu believes price eliminates deliberation. When a phone case costs $1.47, nobody comparison shops. Temu reached 700 million monthly website visits by December 2024, and Sensor Tower data showed users spent 23 minutes per week on the app versus 18 on Amazon. PDD Holdings, Temu's parent, runs a 34% operating margin with 17,000 employees, compared to Alibaba's 15% margin and JD.com's 3%. The margin comes from cutting everything between the factory and the consumer. No brand. No marketing. No consideration stage. Just a price so low that the friction of thinking about it exceeds the cost of buying it.

Shopify believes the consumer is leaving the loop. Its January 2026 NRF announcement positioned Shopify not as a store builder but as a protocol: infrastructure that connects any merchant's catalog to any AI agent. When an AI assistant decides you need new running shoes, it doesn't open a browser. It queries an API. Shopify wants to be that API. This is the most radical bet because it eliminates the consumer from the transaction entirely.

Where the Funnel Still Lives

The strongest case against the funnel-is-dead narrative is that it only describes one end of the purchase spectrum. A $12 phone case on TikTok doesn't need consideration. A $45,000 car still does.

B2B enterprise software purchases involve 6 to 10 decision makers evaluating options over 6 to 12 months. Home buying involves inspections, financing, legal review, and emotional attachment. Medical devices require clinical evidence, regulatory approval, and institutional committee review. For these purchases, the funnel isn't just alive. It's deeper than it was in 1898.

What's actually happening is a bifurcation. Below roughly $100, the funnel is collapsing toward zero stages. Impulse, algorithmic surfacing, and price floors make deliberation irrational. Above roughly $10,000, the funnel is expanding: more stakeholders, more data sources, more evaluation criteria. The contested territory is between $100 and $10,000: the laptop, the winter coat, the furniture set. In this range, some consumers still research extensively while others buy on a creator's recommendation. The funnel exists for some people and not for others buying the identical product.

This bifurcation has consequences for where marketing dollars flow. If the funnel still works for high-consideration purchases, then awareness and consideration spending is rational for cars, real estate, and enterprise software. But for consumer goods under $100, that spending is increasingly wasted on stages that consumers skip. The $107 billion flowing into US social commerce by 2025 is money that used to flow into awareness campaigns, brand building, and search ads. It's being redirected to the single point where impulse and conversion overlap.

The Part Nobody Wants to Talk About

AI-generated reviews are already a documented problem on TikTok Shop and Amazon. If the consideration stage was the consumer's chance to do due diligence, and that stage is being eliminated or contaminated by AI-generated content, then the quality signal disappears. Consumers making purchases in seconds based on a creator's recommendation have no time to check if the product is counterfeit, poorly made, or dangerous.

Temu's business model illustrates the extreme case. Items that don't sell 30 units in 14 days are removed. Sellers are pressured to price below AliExpress. Users spend 23 minutes per week on the app, more than Amazon's 18 minutes, not because they're researching purchases but because the app is gamified. Spin-the-wheel discounts, referral bonuses, countdown timers. The experience is designed to sustain engagement, not inform decisions. By December 2024, Temu's global app users had surpassed Amazon's. In 2024, it was the most downloaded iPhone app in over 20 countries.

The de minimis loophole, which allowed packages under $800 to enter the US duty-free, enabled Temu's direct-from-China model. When the Trump administration closed it in May 2025, Temu announced it would stop shipping Chinese goods directly to US consumers. The pivot to local sellers happened fast, but the underlying model hasn't changed: price so low that deliberation is irrational, engagement so gamified that shopping becomes entertainment.

Agentic Commerce: The Real Cliff

Shopify's agentic commerce announcement is the most important signal in this landscape, and also the least proven. If AI agents begin mediating a meaningful share of consumer purchases, the entire marketing industry faces an existential problem: you can't build brand awareness with an algorithm that doesn't have preferences. You can't create desire in a system that optimizes for price-to-quality ratios.

Today, agentic commerce is mostly demo-ware. No meaningful volume of consumer purchases is mediated by AI agents. But the infrastructure is being built. Shopify's platform connects merchant catalogs to any AI conversation. Amazon's Rufus already recommends specific products within a chat interface. The distance between "AI that answers shopping questions" and "AI that completes purchases on your behalf" is a single permission toggle.

When that toggle flips, the funnel doesn't compress. It ceases to exist as a concept that involves humans.

Limitations

Most social commerce revenue data comes from platform self-reporting or estimates by research firms with varying methodologies. eMarketer, Statista, and Sensor Tower use different panel sizes and measurement approaches, which is why US social commerce market size estimates range from $80 billion to $130 billion depending on the source and definition.

China's social commerce dominance (84% penetration versus ~36% in the US) may reflect cultural differences in consumer behavior that don't transfer. WeChat's super-app model has no Western equivalent, and live commerce adoption in the US remains a fraction of Douyin's.

TikTok's US future is uncertain due to regulatory pressure. If TikTok Shop is forced to divest or exit, its share of US social commerce reverts to Meta, YouTube, and Amazon, platforms that haven't cracked the same impulse-purchase mechanic.

Shopify's "agentic commerce" positioning is strategic framing. No significant volume of consumer purchases is currently agent-mediated. The infrastructure exists but the behavior doesn't, and predicting adoption timelines for agent-based shopping is speculation.

Temu's pre-2025 growth data may not predict its post-de minimis trajectory. The shift to local sellers fundamentally changes its cost structure and value proposition.

The Bottom Line

Lewis's funnel survived for 128 years because it described the physics of information scarcity: people who didn't know about products couldn't buy them. Algorithms broke that physics. When a system can surface a product to a consumer who never searched for it, manufacture desire through entertainment, and complete the purchase before the consumer has a chance to compare alternatives, the funnel doesn't narrow. It short-circuits.

But the funnel isn't universally dead. It's bifurcating. For anything under $100, the funnel is collapsing and the platforms that win will be the ones that compress attention-to-purchase into the fewest possible seconds. For everything over $10,000, the funnel is expanding and the winners will be whoever builds the best AI-assisted evaluation tools. For everything in between, we're in a transition where the same product can be an impulse buy for one consumer and a researched purchase for another.

The marketing industry built $700 billion in annual global spending on the assumption that consumers move through stages. Four of the largest commerce platforms in the world are now operating on the assumption that they don't. Somebody is going to be right, and $700 billion is going to move.