Blog
Notes on the platform economy
Shorter-form commentary between the papers: new findings, method notes, and takes on platform governance and performance in the wild.
ChatGPT Didn't Push Freelancers Off the Platform. It Trapped Its Best Ones in Place.
The standard fear was that AI would hit the most skilled freelancers hardest and chase the exposed ones off the platform. New evidence says the opposite: exposed freelancers stayed and bid their way upmarket, and it was the deepest experts in the threatened categories who got stuck.
The Best Onboarding Trick for AI Products Is Showing Users Less
A field experiment on a generative AI platform found that revealing only part of what users just created, not all of it and not none of it, drove more sign-ups than either extreme, upending the instant-gratification instinct behind most AI product design.
Cut the Commission, Feed the Competitor: The Take-Rate Trap
Pricing, matching, and bundling look like three separate dials on a platform's control panel. A single antitrust complaint against Amazon shows they are wired to the same circuit.
Your Neutral Algorithm Is Not Neutral. Your Follower Graph Decides That.
Swap a popularity feed for a collaborative filter that reaches the whole catalogue, and visibility should spread out, unless the follower graph feeding candidates into the ranker quietly hands it back to the same creators.
Uber One Doesn't Reward Riders. It Quietly Subsidizes Drivers.
A new Information Systems Research paper shows a platform's loyalty subscription helps whichever side didn't buy it, so copying Uber One without checking who is actually scarce in your market is a wealth transfer, not a growth strategy.
The Stance Tax: Why Launching an AI Tool Can Cost You Your Best Creators
A new field study of two Chinese art platforms finds that shipping an AI image generator drove creators away, while banning AI-made art drew them in, because creators were reading a signal, not testing a feature.
The Sleep Reminder That Trained the Algorithm to Keep You Scrolling
A campaign built to cut late-night scrolling on a short-video platform instead taught its recommender to serve more of it, usage stayed elevated for weeks after the campaign ended.
A Claimed Title Beats a Verified Badge. Every Single Time.
Users lock in an opinion after a handful of consistent posts, and a self-declared "Dr." wins more trust than a verified account with a huge following. The judgment forms before anyone checks a fact, and a correction issued afterward mostly fails to undo it.
Gig Workers Can Fight the Algorithm. Only the Ones Who Can Afford To.
A new theory of "algoactivism" built from Uber driver data finds that resisting algorithmic management is not a right workers share equally. It is a resource, and most workers cannot afford it.
Interoperability Doesn't Set Users Free. It Redraws Who Holds The Network.
A new formal model of platform interconnection shows that opening the wall helps a small rival's users, drains a giant's edge, and does either depending on the shape of the network, not on any general virtue of openness.
Nothing New Happened. The Market Moved Anyway.
Researchers added old, already public litigation news back onto the Wikipedia pages of public firms and watched retail trading volume jump 15.7% and bid-ask spreads narrow 6.9 basis points, on stories the market had already priced in years earlier.
Reviewer Honesty Follows a Smile Curve. The Middle Is Where It Breaks.
Newcomers rate candidly because they have not yet learned the room. Then they learn it, inflate for years to keep the peace with whoever they are rating, and only turn honest again once their loyalty shifts from that one person to the community itself.
Big Tech's M&A Didn't Shrink. It Went Underground.
New evidence from nearly 6,900 software acquisitions shows that tougher antitrust enforcement against platform giants didn't reduce their dealmaking. It rerouted it toward targets regulators don't flag, while the market overlap enforcers wanted gone never actually went away.
Practice What You Preach Is Not a Cliché. It Is a Governance Mechanism.
Physicians who spend unpaid hours answering strangers' questions on a Chinese health Q&A platform go on to prescribe less and keep hospital stays closer to clinical guidelines, not because anyone is watching, but because they already said the words in public under their own name.
The Cheapest Robotaxi Fleet Is One Nobody Buys
Waymo and Tesla are racing to own more driverless cars. A new model of AV crowdsourcing finds the cheaper fleet is the one platforms lease from private owners between their own trips, and it only pencils out in the exact pockets of the city where the two schedules refuse to collide.
The Waiting Game: How Gig Platforms Turn Your Uncertainty Into Underpay
New theoretical work shows a platform can hold its total wage bill almost flat no matter how large its workforce grows, simply by posting a low price and waiting for whoever hasn't yet learned her own costs. It also shows why a mass movement can't stop that trick, but the right handful of workers can.
Opting out of tracking was supposed to make you harder to price against. New research on platform competition finds it can do the opposite: platforms fold the untraceable into the same priced-up pool as the barely traceable, and everyone in that pool pays more.
The Pickup Premium: How BOPIS Quietly Raises Your Price
A new study of online marketplaces finds that letting you pick up your order locally isn't a favor to you. It's a way to reveal your address to sellers so nearby ones can stop fighting over your business.
Silence the Stars, and the Middle Class Disappears First
Platforms that demote their biggest creators expect attention to flow down and lift everyone. New research on a major creator platform finds it flows down unevenly, past the middle, straight to the bottom.
Taxi fleets that Uber spent a decade trying to destroy are now voluntarily plugging themselves into Uber's app. New research shows exactly when that trade pays off, and when it just outsources the next disruption.
The Ban That Built Amazon's Bait Shop
Amazon's rule against sellers swapping outside contact info was built to stop revenue leakage. Instead it taught sellers to keep the boring stuff on Amazon and quietly launch their best new products somewhere else.
The Discrimination Flywheel: How a 30% Gap Becomes a 1,007% One
A clean one-week audit of an online teaching platform found a modest booking gap against Black teachers. Twelve weeks of watching the same teachers found the platform's own reputation math had multiplied that gap elevenfold, with no new prejudice required.
One provider's decision to stop screening guests can lift prices, demand, and profit for itself, its rivals, and the platform at once. The moment the platform engineers that same decision into the product for everyone, the welfare gain flips into a platform-only profit grab.
Bad Reviews and Disagreeing Reviews Send New Entrants in Opposite Directions
A rival's low star average tells a new app to sharpen its core feature and differentiate hard. A rival's scattered, disagreeing star ratings tells the same new app to broaden out and copy what already works. Same review data, opposite move.
Sharing Your Market Data Doesn't Help Sellers Sell More. It Helps Them Build Better Products.
Give a retailer platform-level market intelligence and the natural guess is that they'll spend it on price cuts or ad bids. They spend it on invention instead: new SKUs, wider assortments, and sales gains that price and advertising never touch.
Royalty Revenue Doesn't Calm Console Wars. It Arms Them.
When a console expects a cut of every game a gamer buys later, the smart move is to slash the price of the box today and send the bill to the publisher. A new ecosystem model of consoles, publishers, and gamers shows why sharing more revenue makes platforms fight harder, not softer, and why chasing ever more multihoming eventually stops paying off.
Lower Fares Don't Move More Riders. A Pricier Shared Seat Does.
A ride-hailing platform chasing maximum volume, or social welfare, can end up charging more for both solo and shared trips than one chasing revenue, because a pricier shared seat is exactly what nudges riders into sharing, and sharing is what manufactures the driver capacity needed to move everyone else.
Reviews Regulate the Dining Room. Not the Kitchen.
Yelp reviews carry a real, causal hygiene signal, but only for the problems a diner can catch with their own senses. Mistake that partial signal for full oversight and you get a comforting illusion exactly where regulation still has to do the work.
Your "Global" Platform Is Actually Four or Five Regional Ones, and Nobody Designed It That Way
A single hour of time-zone offset cuts cross-region views by 13.6 percent and cross-region votes by 20.9 percent on Stack Overflow, a fully asynchronous, fully global platform. The culprit is not geography. It is the most ordinary ranking choice on the internet: reward whoever is awake right now.
Slow-Rolling a Quality Tier Is the Riskiest Way to Launch One
Sellers left outside a new quality-controlled submarket are not safe by default. They lose early to a newly credible rival tier and recover only after adoption crosses a tipping point, which means a cautious rollout maximizes the time everyone spends in the worst zone.
Auctions Do Not Always Beat Fixed Pricing. On a Thin Creator Market, They Lose.
Real-time bidding is the mechanism ad tech converted to decades ago because it was supposed to be the fair, efficient one. A new analytical model of short video promotion tools shows it only wins once creators are genuinely crowding each other out, and even then it rewards the extremes of content quality while squeezing everyone in the middle.
More Control Over Your Ad Load Is Making You Poorer
Let creators pick their own advertising and they invest less, publish worse content, and earn less money. A single ad-intensity rule the platform imposes on everyone can out-earn the one that lets each creator decide for themselves.
Why the Platform Giant Losing Money Forever Is Not Doing Anything Wrong
A dynamic-game model shows that permanent below-cost pricing by a dominant ecosystem player can be a rational, stable equilibrium rather than predation, which means the regulators and rivals watching its prices are looking in the wrong place.
Personalization Was Supposed to Help the Little Guy. It Did the Opposite.
A 1.7-million-observation study of algorithm upgrades on a food-sharing platform found that popularity-based ranking, the cruder of the two systems, narrowed the gap between top and bottom sellers, while personalization, the one marketed as fairer, widened it by funneling ever more revenue back to the same stars.
More Answers For Less Money Can Leave the Platform Poorer
A new model of paid Q&A platforms shows that letting users pay a small fee to peek at answers already given to others can shrink platform profit, push expert fees up, and make the personalized answers it was meant to complement more expensive.
Showing Women Fewer Profiles Got Them Better Matches
A matrimonial platform restricted which men could even see a woman's profile. Unwanted requests fell, matches nearly doubled, and women started reaching out first, without men doing any worse.
The Rating System That Cancels Its Own Bias
New minority drivers on a large French ridesharing platform earn 11.6 percent less than comparable peers the moment they start, not because they perform worse, but because passengers expect them to. Customer reviews erase almost the entire gap, and they erase it fastest exactly when reviews pile up fastest.
The Marketplace That Made Brands Stop Chasing Follower Counts
When RedNote gave brands a searchable, transparent way to compare creators, sponsorship dollars didn't pile up further on the platform's biggest names, across more than 152,000 posts from 1,161 influencers, they shifted decisively toward the middle tier.
Cloud Computing's Efficiency Miracle Is Also Its Throttling Machine
The statistical predictability that lets AWS, Google Cloud, and Azure pool millions of workloads into cheap, reliable capacity is the same predictability that lets them safely slow down or interrupt lower-paying customers, Spot Instances discounted up to 90% but reclaimable with a two-minute warning are the mechanism working in plain sight.
The Child-Privacy Law That Made Kids' YouTube Worse
After the FTC's $170 million COPPA settlement stripped ad personalization and engagement tools from made-for-kids channels, those creators cut output 18%, let manual captioning fall 28%, and watched views drop 20%, while content and viewership concentrated further among the platform's largest survivors.
Trapped Complementors Don't Quit. They Route Around.
Spotify and Epic Games could never credibly threaten to leave Apple's App Store, so instead they spent five years building coalitions and courting regulators, and walked away with a €1.84 billion fine against Apple, a mandated end to US anti-steering rules, and a 15 percent commission tier that didn't exist before they started fighting.
The Most Effective Sherlocking Never Happens
When Apple entered 22 App Store categories between 2016 and 2021, new competitor entry in those categories fell by roughly 22 percent, and that number, striking as it is, still only counts the founders who tried and stopped; it says nothing about the ones who never started.
The Boycott Was Never the Threat. The Playlist Was.
Public outrage barely moves a stream count. Quietly pulling an artist from the recommendation engine cut listening by a quarter, until one fanbase searched him out anyway and broke the mechanism entirely.
Stop Fencing In Your Creators. Their Second Platform Feeds Your First.
Platforms fight multihoming with watermarks, suppression, and exclusivity clauses. The evidence says the creator who also posts on a rival hands the home platform 23.5% more followers, not fewer.
Stop Defending the Questions ChatGPT Was Always Going to Take
The chatbot is not killing your Q&A platform. It is amputating the low-value half you should have been trying to shed anyway, and the strategic error is fighting to keep it.
Cutting Your Platform Fee Doesn't Buy You Better Apps. It Buys You More Bad Ones.
When Apple and Google halved their cut for small developers, the wave that followed wasn't hidden talent. A study of both fee cuts finds the fairness move drew a flood of lower-quality, less original apps and left the platform to pay the bill in catalog noise.
Certifying Your Best Listings Quietly Devalues All the Others
The moment a platform stamps a favored few "verified," the unlabeled majority stops looking ordinary and starts looking risky. The badge you gave to 3% can cost the other 97% real revenue.
A More Accurate Rating Can Provoke More Cheating
The backfire has a rule. When a score sets a price, half-cleaning it can raise gaming, because trust climbs faster than the loophole closes; when a score sets a rank, sharpening it is always safe.
Free Beats Paid: Why Giving It Away Grows the Tab
The instinct to wall off free features to protect revenue is backwards. In viewer gifting, gig bidding, delivery, and shopping, the free action is the on-ramp to the paid one.
Your Ratings Do Not Measure Quality. They Measure Who Holds Power.
When the rated side controls the rater's livelihood, your five stars are a fear reading, not a quality signal. Break the retaliation channel or govern on a flattering illusion.
Stop Fixing Your Dashboard. Give Your Data to Someone Who Cannot Lie.
The credibility of your analytics falls as your profit from bending them rises. The fix is to hand the raw numbers to a party with no stake in the answer.
Punish Your Best Users, Not Your Worst: The Enforcement Rule Most Platforms Get Backwards
The same suspension that makes an invested user better makes an uninvested one worse. Build losable status before you enforce, or your penalties will backfire.
When a Rival Platform Enters, One Big Sale Will Not Save You
The complementors most likely to defect are the ones your broad loyalty programs never reach. Here is how to spend your retention budget across food delivery, e-commerce, and home-sharing.
Fake Reviews Aren't a Content Problem. They're a Shape Problem.
Amazon, eBay, and Alibaba all catch fraud by mapping who transacts with whom, not by reading what anyone wrote.
Creator Platforms Don’t Govern Creators. Their Networks Do.
Multi-channel networks capture up to 70% of streamer revenue by telling talent a different story about the algorithm than the one they tell themselves. We call it interpretive arbitrage.
Raise Gig Workers’ Pay Floor. Their Earnings Stay Flat.
New York and Seattle data show delivery platforms neutralize a minimum-pay mandate through three offsetting levers: fee pass-through, tip compression, and task rationing. We call it the Wage Floor Offset.
The Friction Dividend: Why Ride-Hailing Apps Don’t Need to Be Cheaper
A one-week audit of 2,238 matched trips found a 14% price gap between Uber and Lyft, and riders check the other app just 16% of the time. The margin platforms collect when you don’t compare.