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.

Theme: Platform Competition · Industry: Mobile App Platforms

_Paper: Benjamin Leyden, "Sherlocking: The Effect of Platform-Owner Entry," Platform Strategy Symposium, 2026 | https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6309158_

Every retrospective on platform copying counts the same way: an app existed, a platform giant built something similar, the app's growth stalled or its company folded. Watson lost to Sherlock. f.lux lost attention to Night Shift. Snapchat's growth rate cratered after Instagram Stories. These are real, well-documented events, and they are also the least interesting part of the story, because they are the cases where someone built the thing in the first place. The harder question, the one a platform-owner's incumbent position lets it answer just by existing, is how many category-defining products never got built at all because a founder ran the calculation and decided the category was Apple-adjacent, or Google-adjacent, or Amazon-adjacent enough that building it would be donating six years of work to a company with better distribution.

The Handcar and the Steam Engine

The term "Sherlocking" has a specific, well-documented origin. In 2001, Karelia Software, run by developer Dan Wood, released an app called Watson that extended Apple's built-in search tool, Sherlock, with a set of internet-connected lookup features, movie showtimes, package tracking, dictionary definitions. It sold for $29.95 and did well. In 2002, Apple's Phil Schiller told Wood that Mac OS X's next version would ship Sherlock 3, which reproduced Watson's functionality as a free, built-in feature. Wood pushed back through Apple's developer relations channel, and Steve Jobs called him directly. According to Wood's own account of the call, Jobs offered a metaphor that Karelia later worked into its company logo: "You know those handcars, the little machines that people stand on and pump to move along on the train tracks? That's Karelia." Watson could not compete with a free, bundled equivalent, and Karelia shut the product down. The episode is why "sherlocked" entered developer vocabulary as the verb for what happens when a platform absorbs a complementor's product wholesale.

That 2002 event set the template that recurs across platform history. In January 2016, Apple's iOS 9.3 beta introduced Night Shift, a blue-light-reduction feature functionally overlapping with the popular third-party app f.lux, notably, just two months after Apple had blocked an unofficial iOS build of f.lux from being distributed. f.lux's developers responded publicly, calling Apple's move "a big commitment and an important first step," while asking Apple to go further and let f.lux itself ship an App Store version with equivalent system-level access. Apple never granted that. In August 2016, Instagram launched Stories, replicating the disappearing-photo-and-video format Snapchat had pioneered. The effect on Snapchat's own growth numbers is unusually well documented: Snapchat's daily-active-user growth rate went from roughly 17 percent quarter-over-quarter in Q2 2016 (its best quarter ever) to about 7 percent in Q3 and just over 3 percent by Q4, as Instagram Stories reached roughly 150 million daily users of its own within months of launch.

What the Paper Actually Measures

Benjamin Leyden's "Sherlocking" study, presented at the 2026 Platform Strategy Symposium, moves past the anecdote collection and asks what happens systematically after a platform owner enters a category its own developers created. Using Apple's entries into 22 App Store submarkets between 2016 and 2021, with same-market apps on Google Play (where Apple never entered) serving as a control group, Leyden finds that Apple's entry deters new third-party entry into that category by roughly 22 percent, with no detectable increase in exit among incumbents already there, and, notably, that surviving incumbents shift toward paid pricing even when Apple's own competing product is free. The pooled effect hides real variation: many submarkets show no measurable effect at all, the pricing response is largest when Apple's version is integrated directly into the OS rather than shipped as a separate app, and the deterrence effect scales with how close the incumbents' products sit to Apple's.

This is a direct empirical descendant of a much older idea. Thomas Eisenmann, Geoffrey Parker, and Marshall Van Alstyne's 2011 Strategic Management Journal paper "Platform Envelopment" described how a platform in one market can absorb a target market's core functionality by bundling it with its own user base and distribution, the mechanism is not a better product, it's a shared-relationship shortcut around the target's competitive advantage. Leyden's contribution is to show that envelopment's deterrent power operates prospectively as well as retrospectively: it doesn't just displace whoever's already in the market, it discourages whoever was thinking about entering it next.

The Damage You Can't See in Any Dataset

Here is where the paper's design gets interesting for anyone thinking about the mechanism rather than just the measurement. A 22 percent drop in entry into categories Apple has already visibly entered is a lower bound on the chilling effect, not an upper one, because it only counts categories where the counterfactual, Apple staying out, is observable in the Google Play control group. It cannot count the founder who looks at Apple's track record of entering 22 submarkets over five years, concludes that any sufficiently valuable single-feature product is a plausible acquisition target for the operating system itself, and picks a different, less "Apple-adjacent" problem to work on before writing a line of code. That decision produces no company, no funding round, no app-store listing, and therefore no row in anyone's dataset, including Leyden's own. It is deterrence with a perfect alibi: the product that never launches can't be counted as evidence of anything, which is exactly why the threat channel of Sherlocking is so much cheaper for a platform than the entry channel. Apple pays engineering costs only for the features it actually ships; the fear of Apple shipping a feature is free, self-enforcing, and permanently invisible.

There is real, if inevitably impressionistic, journalistic evidence that this fear shapes developer behavior directly. NPR's 2024 reporting on Apple's WWDC feature announcements quoted Philip Shoemaker, a former Apple App Store executive, describing routinely fielding complaints from small developers convinced Apple had copied their functionality, and noted that developers often grow wary when Apple's team reaches out to them at all, worried that the contact is really reconnaissance. On the investor side, venture writing on "platform risk", the standard warning against building a company whose core value proposition a larger platform could absorb, has become a fixture of how VCs screen ideas built adjacent to dominant platforms, precisely because a product that is "a feature, not a company" is the profile most exposed to exactly this kind of absorption. Neither of these sources quantifies how many ideas get shelved for this reason; that number, almost by construction, cannot be directly measured. But the pattern they describe, developers modulating ambition based on a platform's demonstrated appetite, not its shipped output, is the mechanism Leyden's entry-deterrence estimate is standing in for.

"You know those handcars, the little machines that people stand on and pump to move along on the train tracks? That's Karelia.", Steve Jobs to Dan Wood, 2002

Why the Threat Outlasts the Feature

The uncomfortable implication for platform owners is that Sherlocking's deterrent value depreciates the moment it becomes routine rather than surprising. Apple's early 2000s entries were novel enough to generate outrage and press coverage, Wood's account of the Jobs call is still cited two decades later. By the WWDC 2023 and 2024 cycles, journalists were compiling annual "everything Apple sherlocked this year" lists as a genre unto itself, and affected developers, per NPR's reporting, mostly responded with carefully diplomatic statements welcoming competition rather than public objection. The visible act has been normalized into an expected cost of building near a major platform; what hasn't been normalized, and what nobody can fully price, is the ongoing tax on categories that look like the next one on the list. For platform owners, this is a genuinely useful governance lever, chilling third-party investment in categories you'd rather enter yourself, without spending engineering resources to actually enter them. For regulators and complementors alike, it is the part of platform power that no shipped-product antitrust case or feature-parity lawsuit will ever fully capture, because the harm is measured in things that were never tried.