Business
The Tiny Designs That Keep You Scrolling
Nine months working on Facebook search taught me how tiny designs keep people scrolling, and how that attention gets sold.
I joined Facebook earlier this year to work on search, starting my social media career. Since then I have spent 9 months rewatching things I once took for granted: the red dot I must clear, the Reels I cannot stop, the little query pill that guesses what I want to look up. Slowly they sorted themselves into patterns. Product details that compound. Four old weaknesses they all exploit. One business they all converge into. This post follows that order, from the obvious to the money.
Demand, Supply, and the Hooks
Here are the features I found can be categorized into three piles.
The first pile makes consuming cheaper. The Like button (2009) pressed feedback cost to one tap and produced the densest training signal on the internet, opening the engagement era with all its pathologies. The News Feed (2006, hated at launch, then universal) flipped consumption from pull to push, and infinite scroll removed the stopping cue. Short video UI pushed the same idea to the limit: no progress bar means never knowing where you are, vertical framing means no rotation plus one clean binary signal per item, swipe up means zero decision with an unknown next, and looped replay holds you three more seconds while annoying you into swiping away by hand.
The second pile makes posting cheaper, and it holds far less. Stories set posting shame to zero because everything vanishes, while the 24-hour clock pressures the other side to watch before it is gone. Duets and stitches turned responding into creating, since reacting is easy and originating is hard. The pile stays thin for a reason: lowering the cost of consuming is engineering, lowering the cost of producing is psychology, and the second is much harder.
The third pile pulls everyone back. The red dot, pull to refresh, and Snap streaks optimize sessions per day rather than minutes per session. Read receipts and online status rewrote chat etiquette with accountability. The unread ring on Stories is the red dot's sibling. People You May Know cured the empty room, because every edge added multiplies feed inventory.
One line to hold: milestone products never invent new human nature. They only build new buttons for it.
Four Old Weaknesses
That raised the real question: why do these buttons work on everyone? BJ Fogg's formula says behavior needs motivation, ability, and a trigger. Product design exists to guarantee the last two. Motivation is where the four weaknesses live.
First, variable rewards. Pull to refresh and the infinite swipe deal unknown hands to consumers; the Like count deals unknown payouts to creators. The platform runs the casino on both sides and takes a cut from each. Skinner proved a century ago that uncertain payouts beat certain ones, and every feed since is a Skinner box with better graphics.
Second, open loops. The Zeigarnik effect says unfinished business occupies memory until closed. The red dot is a visible todo list, pull to refresh is a promised unknown, and the curiosity gap industry (Upworthy headlines in 2013, withheld answer videos today) manufactures open loops on purpose. The unread ring on Stories works the same way: visible unfinished viewing plus a gap that only tapping can close. The urge to clear the badge is not OCD. It is the brain checking out, by design.
Third, loss aversion. Streaks turn consecutive days into owned assets that reset to zero if skipped. Stories run the same fear on a 24-hour clock: watch now or lose it forever. Research says losing hurts about three times more than equivalent winning pleases. FOMO is the same fear wearing information clothes: the dread of losing an edge you never had.
Fourth, social accountability. Read receipts created evidence for ignoring people, online status ended alibis, and unread counts became social debt. Nobody designed guilt into the spec. It arrived free with the feature.
Together they form the complete machine. The feed owns duration (stay once here), triggers own frequency (come back tomorrow), and every return deepens investment, which deepens the next return.
Fashion moves in cycles, so here is my question: TV shopping died years ago, unwatched. Why did the same idea come back to life as livestream? TV shopping failed on every axis livestream fixed: an anonymous corporation shouting (no trust), one-way broadcast (no interaction), three channels for an imagined average viewer (no matching), and call this number to pay on delivery (no infrastructure). The host one follows for months cannot fake a live demo, the room interrogates the product in real time, and what converts is the person, not the pitch. Variable rewards, open loops, loss aversion, and social proof, all four weaknesses running in one room.
Intent and Discovery
My own confusion started at work, and it cuts against everything above. My team helps build search inside Reels, including the query pill that guesses what you want to look up. A feed should weld every exit shut. So why do we open one on purpose? The resolution is pricing. The feed plants demand, search harvests it, and harvest rates run an order of magnitude above discovery rates. Curiosity gap videos open the loop, the pill routes it, and the session monetizes at intent prices while feeding the cleanest targeting data back into recommendations. The whole game is gating, because a query that returns junk burns the session and every ad slot behind it. An exit is only worth opening when it leads to a more expensive room.
Search commands its premium because questions are money. Nobody pays for explain quantum entanglement, while find a divorce lawyer auctions for tens of dollars a click. That is also why Google barely fears large language models: the commercial queries (buy this, hire them, go there) never migrated, because chatbots cannot transact yet, and defaults plus a buying machine tuned over twenty years defend the rest. The real threat sits further out, in assistants that compare and purchase on your behalf and route around the toll booth entirely.
Another one came from daily life: around the fifth time I see the same product, I start wanting it. Advertising calls this effective frequency, and psychology calls the mechanism the mere exposure effect: familiarity feels like liking. One exposure builds nothing, the curve rises then falls, early views plant awareness, middle views convert, late views annoy. Retargeting concentrates those middle views on people who already showed intent. Targeting second: the merchant's Pixel reports the visit, hashed email or phone joins the devices, and the product follows you to Instagram. Google never tells Meta anything. The store you walked into does.
Both prices flow through the same three-link chain: creative agencies make the ads, media agencies spend the budgets, platforms distribute and measure. Platforms keep eating upward. Self-serve storefronts took the long tail of small advertisers, algorithmic buying took media planning, and now AI-generated creative plus scrappy creator-shot phone videos take production itself, since polished brand films test worst in short video. What remains for the holding companies is brand narrative and cross-platform counsel. Put WPP's stock next to Meta's and you see the migration drawn as a chart.
Why Ads Never Die
Making money has exactly one shape: selling things. Selling runs in three steps, be known, be wanted, be bought, and advertising owns the first two. It is unavoidable because supply always exceeds attention, so whoever holds the chokepoint of being known taxes all commerce.
The alternatives fail at scale. Word of mouth converts best but cannot be bought, only earned. Shelves charge their own rent. Sales teams are slow and expensive. Only paid ads are scalable, controllable, and measurable at once, which makes them the only predictable growth machine a business can buy.
The exceptions prove the rule. Nobody advertises where demand exceeds supply: utilities, early Tesla, early Haidilao, Pangdonglai. Tesla started advertising in 2023, the moment demand fell back below supply and the tax came due.
Productivity never moves backward, so supply stays permanently ahead, the buyer's market never ends, and the tax only rises. That is why everything with attention becomes media, from TV operating systems to charging stations. The trend runs one way.