Is Every Scroll a decision you didn't know you were making?

Imagine you're lying on your couch at midnight, mindlessly scrolling Instagram. You weren't looking to buy anything. Then a filtered video of a filtered shower head glides across your screen — same one you've seen three times this week. By morning, you've added it to your cart. Did you choose it? Or did something else decide for you?

That question — deceptively simple, analytically complex — sits at the heart of this investigation. In 2026, the machinery of digital commerce has become so seamlessly woven into how we browse, discover, and buy, that the line between personal preference and algorithmic suggestion has almost entirely dissolved.

The global social commerce market crossed $2 trillion this year, growing at over 31% annually. In the United States alone, social commerce sales hit $100.99 billion in 2026 — an 18% year-over-year increase. These numbers don't just reflect consumer enthusiasm. They reflect the extraordinary effectiveness of systems built to predict — and shape — what you want next.

"They don't just predict our behavior — they actively shape it."

Tim Cook, Apple CEO — on the attention economy

This piece is an attempt to follow the data. Across five datasets spanning 120,000 marketing sessions, 12,719 customer journeys, 1,000 trending products, and 354 survey respondents, we searched for evidence of algorithmic influence — not just in rhetoric, but in behavior, revenue, and revealed preference.

0

marketing sessions give this story its first anchor: not what platforms claim, but what users actually did after arriving through ads, influencers, search and social media.

The scope of this influence is staggering

Before diving into what our data shows, it helps to understand the terrain. Social media platforms have transformed from communication tools into full-stack commerce engines. 90% of online shoppers are on social media, and according to Hootsuite, 75% of e-commerce sales are now influenced by social platforms. The feed is no longer just entertainment — it's a shop window that knows exactly what you're likely to buy.

🔒 Locked — pick an answer to keep reading

Make a Guess

Out of 120,000 tracked marketing sessions, what percentage ended in a purchase?

The actual figure: 6.82% — across 120,000 sessions, 8,182 ended in a purchase. That might sound low, but it represents the real-world baseline of digital commerce conversion. Influencer campaigns and paid ads achieved almost identical rates (both ~6.8%), suggesting reach matters more than channel type — and that the algorithm's job is to maximize who gets shown what.

100-session isotype

Only a small fraction crosses the line from browsing to buying.

Each person represents approximately 1% of sessions. Seven highlighted people approximate the observed 6.82% conversion rate.

0.00%converted
ConvertedDid not convert8,182 purchases / 120,000 sessions
Marketing Channel Conversion Rates

Every channel is made of the same people — and converts almost identically.

Each row is built entirely from person icons — no colored bar behind them, just the count itself. Watch how close the rows are in length — that's the finding.

0.00pp spread, top to bottom
1 icon ≈ 0.2% conversion rateEmail 7.31% · Organic 6.81% · Social 6.81% · Paid Ads 6.72%
All channels cluster tightly between 6.7–7.3%, suggesting algorithms have equalized the playing field — the bottleneck is the product, not the channel.

What the data actually shows about purchasing behavior

If algorithms were secretly hijacking consumer choice, we'd expect to see dramatic differences in how people behave depending on where they arrived from. Social media visitors, manipulated by curated feeds, should convert at dramatically higher rates than someone who searched organically, right?

The data tells a more nuanced story.

The Customer Journey Funnel

The same crowd keeps falling — and keeps thinning out.

Small figures inside each stage stand in for the people there. Watch how few remain by the bottom; the highlighted stage is who actually bought.

0.0%made it all the way
Home 5,000 visits 100% of traffic Product Page 3,987 visits 79.7% of traffic Cart 1,599 visits 32% of traffic Checkout 1,123 visits 22.5% of traffic Confirmed 1,010 purchases 20.2% of traffic
Still in the funnelConfirmed purchase1,010 / 5,000
The sharpest drop is between cart and checkout — a 30% abandonment rate suggesting users are still actively deliberating, not passively completing algorithm-driven purchases.
0

of users who start checkout still abandon the purchase. That is the moment where algorithmic exposure meets human hesitation.

The funnel reveals a striking pattern: the biggest drop-off happens at checkout, not at discovery. Users reach product pages, add items to their carts — and then pause. They re-evaluate. A full 30% who start checkout don't complete it. This isn't passive consumption; it's active deliberation.

And when we break down conversions by referral source, the results are even more revealing:

Purchase Rate by Referral Source

The same people, sorted by how they arrived — four slightly different rates.

Each bar is textured with the people it represents. Social Media, the algorithmic feed, is the shortest of the four.

0.0%highest: Google
Same colors as the channel chart above — Email and Social carry throughGoogle 41.9% · Direct 39.3% · Email 39.1% · Social 38.4%
Social Media has the lowest conversion of all four sources — users driven in by algorithmic feeds are the most likely to browse without buying.

Social Media produces the lowest purchase rate of all four acquisition channels — 38.4% versus Google's 41.9%. Users arriving via algorithmic feeds are, paradoxically, the least likely to convert. They browse, compare, reconsider. The algorithm gets them to the page. The user decides whether to stay.

⚡ Key Finding

Algorithms are extraordinarily effective at generating exposure. They are far less effective at overriding conscious deliberation. The conversion gap between Social Media and direct/organic channels suggests users arriving via feeds apply more skepticism to their visit — not less.

Fast buyers, slow buyers, and the engagement threshold

One of the most compelling patterns in our data involves the temporal dimension of purchasing decisions. Using time-on-page as a proxy for engagement depth, we examined whether buyers and non-buyers behave differently during their sessions.

🔒 Locked — pick an answer to keep reading

Make a Guess

Do buyers spend significantly more time on product pages than non-buyers?

Buyers averaged 97 seconds on page. Non-buyers: 97.7 seconds. The difference is statistically negligible. This is a profound finding — time spent browsing does not reliably predict purchase intent. What matters is decision style, not total engagement time. Some users arrive knowing exactly what they want and check out fast. Others deliberate at length — and still don't buy.

The data from our customer journey analysis reveals a striking behavioral split. Rather than a single engagement pattern, we observe what the slides described as "polar effects": a cohort of fast, determined buyers who arrive, evaluate briefly, and convert — and a separate cohort of slow, hesitant non-buyers who browse extensively and ultimately leave.

Purchase Rate by Time Spent on Page
Conversion varies by session length — but not in the direction you'd expect
Customer Journey Dataset
Short-session users (<60s) convert at a higher rate than long-session browsers (>200s), supporting the "fast-deliberate split" thesis. High time + no purchase = decision fatigue, not deeper consideration.

This pattern — which our analysis calls the "decision fatigue zone" — suggests that extended browsing often signals confusion or uncertainty rather than deeper engagement. Users who spend the most time on pages are frequently those least likely to commit. The algorithm brought them to the door; their internal deliberation kept them from entering.

This is the crucial nuance that aggregate conversion statistics miss: users are not passive recipients of algorithmic nudges. They arrive influenced, but they leave by choice.

Exposure in 100 users

Ad exposure is common — but exposure is not the same as causation.

Sixty-four highlighted view symbols approximate the reported 63.8% who bought after seeing an ad. The visual shows scale while keeping the causal limitation explicit.

0.0%after ad exposure
Reported ad exposure before purchaseOtherAssociation ≠ proof of causation

When a product goes viral, the revenue gap is extraordinary

Our Shopify trending product dataset covers 1,000 products across 2025, tagged by their primary trend source: Google Trends, Amazon Best Sellers, Instagram Shop, Shopify Trending List, and TikTok Shop Viral. The last category is small — only 6 products — but its financial signature is unmistakable.

0.0

viral TikTok products generate almost three times the average revenue of products trending through other discovery channels in this dataset.

Head-to-head

TikTok Viral vs. everything else, averaged

Same dataset, one direct comparison: TikTok's 6 viral products against the average of all 994 others.

TikTok Viral$24.4M avg
vs
0.0×
All Other Sources$8.9M avg
Average Revenue by Trend Source
Where a product trends shapes how much it earns
Shopify Trending Products 2025 · n=1,000
TikTok viral products average $24.4M — nearly 3× the next-highest source. The platform's short-form algorithm produces the most commercially potent exposure of any channel.

TikTok viral products average $24.4 million in revenue — compared to $9.7M for Amazon Best Sellers, $9.1M for Instagram Shop, and $8.9M for Google Trends. That's a 2.7× multiplier just from going viral on one platform's recommendation algorithm.

And the correlation between trend score and revenue is sharp: products scoring 95 or above on trend score average $20.1M in revenue, compared to just $5.9M for those below 85. The algorithm doesn't just influence what people discover — it dramatically amplifies the commercial outcomes of the products it decides to surface.

Circle-packed revenue comparison

Category revenue is encoded in the size of each packed circle.

Larger circles represent higher average revenue. Exact values remain visible inside each circle, so the visual comparison does not replace the numbers.

Circle area follows average revenueRange: $8.0M–$10.9M average revenue
Pet and Eco-Friendly categories outperform Fashion and Beauty, suggesting trend algorithms favour novelty and utility over traditional retail categories.

Do users feel influenced? The answer is complicated

Our SMI Attitude Questionnaire collected 354 responses measuring attitudes toward social media influencers across five dimensions: Convenience, Interactivity, Attractiveness, Expertise, and Trustworthiness. The picture that emerges is not of naive consumers manipulated against their will — it's something more complex.

0.00

Trustworthiness is the lowest construct in the influencer attitude dashboard: users find social media convenient, but they do not fully trust the people selling through it.

Consumer Attitudes Toward Influencers
Average Likert scores (1–5) across five perception dimensions
SMI Attitude Questionnaire · n=354
Convenience scores highest — users find the platform useful. Trustworthiness scores lowest — they know it's commercial. This combination is the defining psychological tension of social commerce.

The survey shows respondents rate influencer Convenience highly — social media is genuinely useful for product discovery. But Trustworthiness scores are the lowest of all dimensions. Users are aware of the commercial nature of what they're seeing. They find it convenient to browse, but they don't blindly trust the source.

This tension — useful but not fully trusted — explains much of the behavioral data. High exposure, moderate conversion, active deliberation at checkout. The algorithm surfaces products effectively. The consumer filters skeptically.

Users recalculate. They are not fully passive.

Here is where the narrative resists easy conclusions. Across all of our datasets, a consistent pattern emerges: users are influenced to arrive, but not coerced into buying.

The funnel data shows 30% checkout abandonment — consumers who got deep into the purchase process and still walked away. The journey data shows Social Media arrivals converting at lower rates than organic or direct visitors. The survey shows awareness of algorithmic personalization sitting alongside purchasing behavior — consumers who know they're being influenced and adjust accordingly.

The truly passive scenario — where the algorithm fully determines the purchase — doesn't match the data. What we see instead is a two-stage process: algorithmic influence shapes the discovery landscape, and human deliberation determines what happens next. The algorithm defines the menu. The consumer still orders.

"More browsing ≠ buying. Some users are lost, not engaged. Algorithms get them to the door — the user decides whether to walk through."

From the analysis of customer journey data

Limits of our evidence

Analytical honesty requires naming what this data cannot establish. We do not have access to the actual recommendation signals — what the algorithm showed, and how often. We cannot observe the counterfactual: would the user have discovered this product anyway, without algorithmic surfacing? We cannot measure the long-run effect of repeated exposure on preference formation.

Our survey sample skews young (18–29) and student/low-income, which limits demographic generalizability. We know whether people bought, but not how much they spent or how often they repeat. And while we measure both ad and influencer influence, the two channels aren't always directly comparable across identical scales.

⚡ What The Data Can Say

Observable behavior patterns show influence at the exposure and discovery stage, but preserved human agency at the conversion stage. The mechanism of influence is real. Its power to override deliberation is not absolute.