SC
Strategy Map›Layer 3 · Defensibility›Competitive Advantage Examples

Competitive Advantage Examples

What is a competitive advantage?

A competitive advantage is whatever keeps customers choosing a company after rivals have seen what it does and tried to copy it. It lasts when it rests on a barrier, such as switching costs, network effects or scale, that a competitor cannot simply buy its way past.

Here are 110 companies from our case studies that beat a named rival, grouped by the barrier that kept them ahead. Switching costs were the main barrier in 53 of the 122 rivalries, more than any other kind. 26 more winners never built a barrier at all; they are at the end.

Switching costs (53)

The customer expects to lose value by changing to an alternative for the next purchase. Concept page: Switching costs.

WinnerBeatWhat kept it ahead
AdobeQuarkSwitching costs of a trained team working across an interoperable suite
EvidenceIn 2003 Adobe itself named conversion, training and print-shop infrastructure as the barriers keeping Quark customers; by 2023 CMA respondents named retraining as the main barrier to leaving Adobe, and said any change must cover the full suite, not one app
AmadeusSabreSwitching costs
EvidenceMoving an airline's reservations is a multi-year project (Southwest planned its move over several years after two years of work on the first phase; Air Canada replaced a 30-year-old system); larger airline contracts run three to seven years; Amadeus boarded 2,248.6 million passengers in 2025, up 3.8%.
AppFolioBuildiumSwitching costs of the manager's ledger and payment rails
EvidenceLeaving means reconciling trust accounts, rebuilding owner reports and re-setting payments; managers absorbed 2023 eCheck fees rather than switch, and units still rose 13%.
AtlassianRally SoftwareSwitching costs of configured workflows and installed Marketplace apps
EvidenceFiling describes products entrenched in how organizations run (2016); more than 60% of customers and over 80% of the largest run at least one Marketplace app (2019); cohort revenue grew after first purchase (FY2016); leaving means rebuilding fields, workflows and app functions (Oscar migration, 2025)
Bentley SystemsIntergraphSwitching costs in DOT DGN plan standards and project archives
EvidenceDOTs reopen archived projects without translation; translators proved unreliable (2012); Caltrans requires DGN plans whatever the design tool, AutoCAD files not accepted (2014); 98% account retention (2019)
BlockPayPal HereSwitching costs of a seller's register, data and loans running on Square's payment flow
EvidenceSeller cohorts retained over 110% of net transaction revenue (2015); 44% of Square gross profit from sellers on four or more products, and those mid-market sellers retain 15x better than single-product sellers (2022); loans repay from Square card sales (2014-25)
BloombergTelerateSwitching costs of a desk's counterparty chat and analytics
EvidenceMessaging became the standard for client communication and contracting, making switching costly (Bakker 2025); a boutique's traders self-paid $500+ monthly to keep Bloomberg after a switch (2013); 15 banks funded Symphony for $66M (2014)
BrazeIterableSwitching costs from an SDK compiled into customers' apps and journeys built on its event stream
EvidenceOver 90% of customers run the Braze SDK and 80% stream data out through Currents (March 2026); an Iterable-to-Braze move takes 8-16 weeks to rebuild every template and journey and re-instrument the app (2026)
Cadence Design SystemsMentor GraphicsSwitching costs of the Virtuoso custom-design flow, held in place by foundry kits written in SKILL
EvidenceAbout 90% of analog design used p-cells, mostly SKILL on Virtuoso (2007); a rival vendor called designers 'locked into a prison' (2007); foundries kept one kit for Cadence, another for everyone else (2012)
ClouderaMapRSwitching costs that held the installed clusters, not new workloads
EvidenceOlson: petabyte data 'has so much inertia that it's painful to move' (2012); ARR growth 'primarily due to existing customers' and churn about 10% (2020); 190 customers above $1M ARR (2021)
Cornerstone OnDemandSaba SoftwareSwitching costs
EvidenceThree-year enterprise-wide subscriptions with no cancellation provisions except for Cornerstone's default; average annual dollar retention of 95% from 2002 to 2011, and recurring revenue from existing clients higher at the end of every year than at the start from 2001 to 2011; net annual dollar retention of 105.7% in 2018 and 104.6% in 2019; 95.1% in 2020 after the calculation changed to include all customers and incremental sales.
CrowdStrikeCybereasonSwitching costs
Evidence97% gross retention in fiscal 2025 after the July 19, 2024 faulty update crashed Windows systems; 67% of customers on five or more modules running on one agent.
CrowdStrikeSecureworksSwitching costs
EvidenceAfter a faulty Falcon sensor update crashed Windows systems on July 19, 2024, third-quarter gross retention was over 97%, down less than half a point; dollar-based gross retention was 97% at fiscal 2025 year end, with 67% of customers on five or more modules.
DatadogNew RelicSwitching costs from dashboards and alerts configured across teams
EvidenceDashboards, monitors and training shared across infrastructure, APM and logs (2018); leaving means recreating them and retraining engineers (2026 accounts)
DayforceADPSwitching costs from payroll, tax and employee records run on one database
EvidenceInitial terms of three to five years with evergreen renewal and three-to-nine-month implementations (2018); cloud revenue retention 95% or above 2015-17 and 98.0% gross retention in 2024; Ossip calls the tax engine a growing moat (2026)
Dun & BradstreetThe Bradstreet CompanySwitching costs
EvidenceGSA told the GAO in 2012 that D&B effectively had a monopoly and that moving to a government-owned number would mean changing data systems, replacing DUNS data and rewriting procurement rules, at a cost GAO said could be prohibitively high; GSA completed the move only in April 2022. Commercially, 96% annual revenue retention in 2024 and 17 of the top 25 clients served for more than 20 years.
Epic SystemsCernerSwitching costs of the enterprise record, reinforced by the expertise network of Epic-trained staff
EvidenceEpic 'lost a customer once, and they came back a year later' (Faulkner, 2026); 'we don't really lose customers' (2024); net gain of 176 hospitals in 2024 while Oracle Health lost 74; Mayo's replacement cost $1.5B across ~52,000 staff (2017-18); certified analysts, consultants and clinicians trained on Epic lower the cost of staying and raise the cost of retraining for any rival
Epic SystemsVeradigm (Allscripts)Switching costs
EvidenceMoving a large system onto Epic is a multi-year, phased project (Northwell, 2023); once there, customers stay: Epic added 153 acute care clients and lost none in 2023, and won nearly 70% of hospitals in 2024 decisions.
FiservMetavante (M&I Data Services)Switching costs
Evidence61% of US banks had used the same core provider for more than 10 years (ABA 2022); only about 2% leave voluntarily each year; contracts average 5.6 years; Fiserv ran the core for 42% of banks and 31% of credit unions.
HubSpotMarketoSwitching costs from records and workflows on HubSpot's own CRM
EvidenceCustomers on several hubs rose from 34% (2017) to 60% (2021) as net revenue retention climbed from 88.6% at IPO (2014) to 115% (2021); leaving means rebuilding fields and workflows (Recharge, undated)
InstructureBlackboardSwitching costs
EvidenceGross revenue retention of 95%, 96% and 95% at year-end 2019, 2020 and 2021; net revenue retention 107%, 117% and 109%; Canvas at 41% of US and Canadian institutions and 47% of enrollments at the end of 2023, against Blackboard's 17% and 18%.
JFrogSonatypeSwitching costs
EvidenceArtifactory is the single source of truth for an organization's packages, and integrations, workflows and stored binaries built on it make it sticky; net dollar retention 133% (2020) and 119% (2025); 1,168 customers above $100,000 ARR and 74 above $1 million (2025); Enterprise Plus 57% of fourth-quarter 2025 revenue.
KaseyaN-ableSwitching costs
EvidenceMSP technicians work in the RMM all day and moving off a PSA or backup vendor is disruptive under multi-year contracts; Kaseya sells RMM, security and backup as one per-endpoint subscription (2024) and reached $1.5B of ARR in 2024.
KlaviyoMailchimpSwitching costs from years of stored commerce history and the flows built on it
EvidenceEvery order and browse event stored per profile since the store connected (2023 filing); flows cannot be exported, only lists, segments, templates and reports, so leaving means rebuilding flows and forms (2025 exit guide; 2026 agency migration guide); gross revenue retention 88% (June 2023)
Materialise3D SystemsSwitching costs
EvidenceDeferred revenue from annual software sales and maintenance of €13.136M at end-2015; Parts on Demand, a long-time Magics user, connected its new production system to Magics and used command-line access to the full Magics SDK; software revenue fell 6.8% to €40.907M in 2025, half the Manufacturing segment's 13.2% drop.
MongoDBRethinkDBSwitching costs of applications written to the MongoDB API, fenced by the SSPL
EvidenceMicrosoft (2016) and AWS (2019) cloned the API so apps keep 'the same MongoDB application code, drivers, and tools'; SSPL (2018) held AWS to the 3.6 API
OktaOneLoginSwitching costs plus a platform network effect: an integration network vendors build to because Okta's enterprise customers use it
EvidenceThe network grew from 3,000 apps (2013) to 5,000+ (2016), 7,000+ (2021) and 8,000+ (2026). Outside vendors submit and certify their own integrations through a self-service portal (in use by 2020). Some large buyers make vendors certify with Okta before they can sell. Okta pitches vendors on reaching 17,000+ customers, including two thirds of the Fortune 100, and its S-1 describes more customers bringing more vendors. Customers run 89 to 101 apps each through Okta, and gross retention held in the mid-90s after the 2023 breach and stayed steady through fiscal 2026
OpenTextDocumentum (EMC)Switching costs in compliance records repositories
EvidenceMaintenance renewals over 90% (2004), 94% (2022), 91-96% (2025-26); firms that moved users to SharePoint kept one OpenText instance for compliance (Gartner, 2015); leaving means a full-fidelity records migration, rebuilt SAP integrations and repaired document links (2021)
PalantirRaytheonSwitching costs from operations modeled in the Palantir ontology, reinforced by program-of-record status
EvidenceGotham became DCGS-A's data fabric (2021); Army consolidated 75 contracts into one Palantir enterprise agreement (2025); Maven Smart System ordered to program of record (2026); commercial customers reorganize operations on the ontology, net dollar retention 139% (Q4 2025)
PaylocityPaycorSwitching costs
EvidenceAnnual revenue retention above 92% in every fiscal year the cited filings report (2011–13, 2018–20, 2024–26) though most contracts could be ended on 60 days' notice or less (term agreements of about two years were added later); employers switch payroll mostly at the start of a calendar year; HMC kept ADP for 25 years before moving to Paylocity in 2018. Limit: OWL Services consolidated Paylocity, Paycor and QuickBooks businesses onto ADP after a ten-company merger.
PegasystemsAppianSwitching costs
EvidenceMost Pega clients have historically renewed, its software runs mission-critical workflows, and it expected $2.1B of revenue from backlog at the end of 2025; Appian's founder said in 2016 that in 2012 customers who had installed the acquired BPM rivals' systems kept them, so Appian could not win those lost deals back.
PelotoniFIT (ICON Health & Fitness, maker of NordicTrack)Switching costs in the installed base: retention held up but weakened as the base shrank
EvidenceRetention held up but weakened: average net monthly churn of 0.65% in fiscal 2019, 1.41% in the June 2022 quarter and 2.2% in the June 2026 quarter; subscription revenue of $436.6 million at a 73.6% gross margin in the June 2026 quarter funded first full-year GAAP net income of $63 million. The lock-in protects existing owners and does not add new ones: paid subscriptions fell 8.8% to 2.553 million in fiscal 2026, and the owned manufacturing, delivery and exclusive channel of links 1-2 were largely dismantled in 2022, leaving a smaller fleet of first-party microstores.
Procore TechnologiesPlanGridSwitching costs of the shared project record, reinforced by partners who already work in it
EvidenceBudgets, commitments, invoices and years of documents sit in one record (2019-25); gross revenue retention 95% (2025); win rate above 60% against top rivals (2024); a Georgia contractor that left for a cheaper tool in 2024 returned and expanded in 2025; owners mandate Procore down to contractors and subs (2026)
PTCSDRC (Structural Dynamics Research Corp.)Switching costs
EvidenceMoving PLM data is among the biggest obstacles to switching platforms, and Gentherm's move to Windchill took two years for 15 years of data (2018); PLM choices pull CAD choices with them, as when Daimler moved from Dassault's CATIA to Siemens NX to match its Teamcenter data system (2011); 95% of PTC's fiscal 2025 revenue was recurring across more than 30,000 customers.
QualtricsSurveyMonkeySwitching costs from company-wide experience programs
EvidenceIntegrations with CRM and ticketing (2016); 57,000 Walmart manager dashboards (2018); net retention above 120% for eight quarters to 2020; rivals advise layering on, not replacing (2026)
Rockwell AutomationModiconSwitching costs
EvidenceLogix gives customers one software platform to learn (Bloomberg, Nov 2017); PLC-5 devices still in operation after ControlLogix replaced them; ARC user survey, November 2001: Allen-Bradley 60%-70% of North American OEMs and end users, Siemens 5%/19%, Modicon 5%/4%, Siemens much larger in Europe; Logix North American sales up over 25% in Q1 FY2026. The defense is evidenced for North America.
RubrikCohesitySwitching costs in the backup history and recovery runbooks held in the platform (qualified)
EvidenceRecovery depends on backup history and metadata in Rubrik's format (2024 S-1 letter); leaving means running or migrating the old backup chain, which Cohesity built AI file translation to do for Veritas backups (2024); typical three-year contracts; net retention 150% (Jan 2023) to over 119% (2026)
SalesforceSiebel SystemsSwitching costs of configured CRM and installed apps, reinforced by a trained Salesforce workforce
EvidenceCustomers configure fields, workflows and AppExchange apps on the shared core (2005 on); a specialist partner leaving the platform, Veeva, needs a program running to 2030 with 25+ migrations done and ~300 left (2025); the ecosystem was 4x Salesforce's size (2019) and 21M Trailblazers (FY2024) supply the skills employers hire for
SamsaraMotive (formerly KeepTruckin)Switching costs in large multi-product accounts
EvidenceDevices sit inside multi-year per-asset contracts, hardware about 25% of initial contract value (2023); leaving means pulling and rewiring gateways and cameras and retraining staff (2025); 96% of 00,000+ customers run two or more products (2026); Motive puts an enterprise rollout at close to three months (2026)
ServiceNowBMC RemedySwitching costs of the enterprise workflow system of record
EvidenceCMDB, custom apps and non-IT workflows share one data model (2013-17); 98% renewals (Q4 2025) even after renewal discounts were withdrawn (2015); exits run for years (Stanford left Remedy over 2018-19)
ServiceTitanJobberSwitching costs of the contractor operating system
EvidenceCustomers run calls, dispatch, pricebook, payments and reporting on it; leaving means redoing data, accounting and phone migrations and retraining. Gross churn under 1% a month (2015); gross dollar retention above 95% (2022-24 quarters, FY2026).
SnowflakeAmazon RedshiftSwitching costs of migrated workloads and the pipelines built around them
EvidenceCapital One consolidated analytics, marketing and petabytes of logs on one platform across its lines of business (2017-20); DoorDash engineers built and maintained ETL jobs and dataset tuning around it (2021); net revenue retention of 158% (July 2020) and 177% (Q4 FY22). Leaving means rebuilding pipelines, permissions and dependent reports.
SpotifyRdioSwitching costs of personal libraries and taste history, with scale terms from labels
EvidencePremium churn fell from 7.5% to 5.1% (Q4 2015-17) as Apple Music entered; majors cut their share from 55% to 52% (2017); Spotify blocked library-transfer apps
SprinklrSalesforce Social StudioSwitching costs of enterprise-wide social governance
EvidenceAbout 95% of customers renewed each year and existing customers' spend grew to 150% (2015); thousands of accounts, approval flows and audit trails run through one system (Starwood 3,000+ accounts, Nasdaq audit trails); Cisco re-evaluated and stayed; 141 customers above $1M averaging over $3M each (FY2026)
StripeBraintreeSwitching costs
EvidenceBilling used by more than 300,000 companies with nearly 200 million active subscriptions (2024); the Revenue suite on track for a $1 billion run rate in 2026; Intercom's move onto Stripe Billing took a three-month assessment, a three-phase plan that ran from November 2023 to October 2025 to move all subscription customers, and wired Billing into its tax, accounting and CRM systems.
ToastRevel SystemsSwitching costs from hardware, card processing, payroll and loans run through one system
EvidenceChurn almost entirely from closures and no loss to a cloud rival (Dec 2015); leaving means replacing terminals, processing, configuration, training, payroll and any open Toast Capital loan without stopping service (2019-26); a 2023 fee reversal shows limits on pricing power
Ultimate SoftwareLawson SoftwareSwitching costs
EvidenceAn annual revenue customer retention rate for recurring revenue cloud customers of about 97% at the end of 2016 and about 96% at the end of 2017 and 2018; UltiPro served as customers' central system of record for employee data; $522.1M of signed contracts awaited go-live at the end of 2018, and 2017 go-lives took longer than planned.
VeevaOracleSwitching costs of validated regulated records and a connected product suite
EvidenceVault holds each document's audit trail from creation to archive, validated for FDA rules (2015); CRM, Vault and Network sold as one integrated suite (2016); commercial customers own four products on average (2024); about 80% of life-sciences CRM (2025)
WaystarOlive AISwitching costs
EvidenceWaystar's 2024 10-K cites the significant time and resources it takes to switch healthcare payments providers and integrations with over 200 channel partners that make it harder to replace; contracts run two or three years with automatic renewals; net revenue retention 110.1% in 2024; 83% win rate in competitive decisions 2022 through 2024.
Weave CommunicationsSolutionreachSwitching costs of the office phone line and the front-desk routines built on it
EvidenceFiling: 'The deployment of the Weave phone system at each of our customers increases stickiness and customer loyalty' (2021); once integrated, Weave 'became fundamental to how that office ran', so churn stayed low (Ralston, 2021); 94% gross and 103% net dollar retention (2021); staff resisted removing it after a trial (Austin, 2025). Leaving means moving the phone line, texting, reminders and payments at once.
WorkdayOracleSwitching costs of the HR-and-finance system of record
EvidencePayroll, benefits, finance and identity integrations and years of configured processes run through one instance; Yale's finance go-live alone changed or retired 170+ systems (2017); implementations take 6-18 months (2026); gross revenue retention about 97% (FY2026)
XeroMYOBSwitching costs of accounting practices standardized on the shared ledger
Evidence20,000 to 30,000 accounting firms working their clients' books in Xero (2017); customers and accountants loyal, with low churn and a high degree of stickiness (2025); Xero leads cloud accounting in New Zealand, Australia and the UK (2025)
ZendeskFreshworksSwitching costs of configured support operations in expanded accounts
EvidenceTriggers, macros, SLAs, integrations and reports are rebuilt by hand on exit (2026 migration guide); Freshworks' CEO, winning bake-offs, heard only 'some conversations' of Zendesk customers wanting to move (2022); seat expansion drove 122% net expansion with 37% of ARR in $250K+ accounts (2021)

Catch-up barriers (16)

A rival can match today’s spending without matching the assets built over years. Concept page: Accumulated assets & catch-up barriers.

WinnerBeatWhat kept it ahead
Ancestry.comMyHeritageAccumulated records and DNA network
Evidence27 billion records and more than 18 million people in the DNA network by 2020; more than $60 million committed in 2013 to digitize about a billion FamilySearch records; 14 million kits sold by November 2018 against MyHeritage's 2.5 million profiles, with larger databases giving each tester more matches.
ConfluentAmazon MSKCatch-up barrier: Kafka's creators' operating expertise built into a cloud-native engine and managed ecosystem
EvidenceCommitter-led support with 60-minute severe-issue response (2021); Kora engine runs tens of thousands of clusters in 70+ regions with 30x faster scaling than self-managed Kafka (2023-24); MSK matched pay-as-you-go storage and faster scaling only in November 2024; 1,521 customers above $100K ARR and $624M Cloud revenue (2025); IBM paid $11B enterprise value (March 2026)
DigitalOceanLinodeAccumulated tutorial library that owns developer search
EvidenceAbout 6,000 tutorials and 30,000 Q&As (2021); roughly 20 times Linode's organic search traffic (2026); Linode users read DigitalOcean guides for Linode servers (2022)
DoximitySermoCatch-up barrier: verified, current profiles of most U.S. physicians
EvidenceProfiles claimed by 80% of physicians (2021), 85% (2026); all top 20 pharma buy; 125 customers over $500k give 83% of revenue (FY2026)
DynatraceAppDynamicsCatch-up barrier: incumbent rivals had to rebuild to match the automatic full-stack platform
EvidenceDynatrace's rebuild took 'four years, maybe five' (Van Siclen); Cisco launched AppDynamics Cloud in 2022, replaced by Splunk Observability Cloud in 2024; Dynatrace highest in execution, furthest in vision (Gartner, 2024)
EquifaxTrueworkAccumulated assets (catch-up barrier)
Evidence813 million current and historic employment records from payroll data supplied by over 4 million organizations at end-2025, up from nearly 330 million records and 10,600 organizations in 2017; employers are generally not charged and contributor turnover is not significant.
Flatiron HealthCancerLinQCatch-up barrier: a decade of clinician-curated oncology records with a regulatory track record
EvidenceAbout 3.4 million curated patient records with an audit trail to source documents (2024); a decade of human-abstracted data used as the benchmark to validate AI extraction (2025); FDA collaboration since 2016, renewed for five years in 2023. ASCO judged 'hundreds of millions of dollars' were needed to structure comparable data and sold CancerLinQ (2017, 2023).
IQVIASyneos HealthAccumulated data assets, with switching costs on live trials
EvidenceMore than 1.2 billion non-identified longitudinal patient records from about 150,000 data suppliers, assembled over decades and linked by in-house tokenization (2025); IQVIA withheld consent for its data in Veeva applications, which Veeva said hurt its sales (2024); R&D Solutions contracted backlog $17.13B (2018), $27.2B (2022), $32.7B (2025).
KLAPDF SolutionsAccumulated assets
EvidenceInspection markets KLA opened in the 1970s and 1980s, held at about 70% of wafer inspection and 80% of reticle inspection in the early 1990s; every major chipmaker used its products by 2001; 73.8% of metrology and inspection in 2025, more than four times Applied Materials.
NetskopeBitglassAccumulated assets: a private network and one platform a rival must rebuild
EvidenceNewEdge built data center by data center from February 2019 to more than 120 data centers in more than 75 regions (2025), with one console, engine, client, network and gateway across products; gross retention 96% and net retention 118% as of July 31, 2025.
PlaidYodlee (Envestnet Yodlee)Accumulated, maintained bank connections and network data
Evidence12,000 bank integrations built by brute force (Perret, 2025); 6,121 integration repairs in 2025; 52% of broken connections repaired automatically by another app in the network; more than one in two people in the US with a bank account linked through Plaid; fraud and credit products trained on network data. Defense against standardized, priced bank APIs (CFPB 1033 rule, JPMorgan fees from 2025): onboarding network effects and network-wide repair on top of the raw connection.
RedditDiggCatch-up barrier: self-moderated communities and their archive, compounding by network effects within each community
Evidence100,000+ active subreddits and 16B comments (2024) licensed to Google and OpenAI; communities that left in the 2023 blackout stayed small elsewhere; Digg's 2026 relaunch by Rose and Ohanian closed in two months, its CEO calling network effects 'a wall'
SynopsysMentor GraphicsAccumulated assets and catch-up barriers
EvidenceDecades of process-specific learning built with customers and foundries: Synopsys engineers work inside TSMC, Samsung, Intel and GF during process development; Ghazi explains the two-firm market by the cumulative knowledge a newcomer would need; PrimeTime sign-off accepted by virtually all major semiconductor manufacturers by 2002; two- to three-year subscription licenses bundle continuing updates. Design Compiler was the market-leading logic synthesis tool in fiscal 2003.
TempusFoundation MedicineCatch-up barrier: the largest linked molecular and clinical oncology library, with follow-up that only accrues over time
EvidenceMore than 900,000 records with matched clinical and genomic data, a database over 50 times The Cancer Genome Atlas and 200+ petabytes (2024); records keep earning as outcomes accrue: the 2018 cohort of about 7,500 patients produced $66.2M through 2023; Insights net revenue retention 126% (2025); 2025 revenue about $1.27B; market cap $11.15B at end-2025 against Foundation's $5.3B valuation in Roche's 2018 buyout
TripadvisorLonely PlanetAccumulated stock of traveler reviews that rivals cannot quickly reassemble
Evidence50M+ reviews (2011) grew to more than one billion reviews and opinions (2025), with 31.1M new reviews in 2024; Google took Tripadvisor's reviews to fill Google Places (2011), then bought Zagat (2011) and Frommer's ($22M, 2012) for content of its own
ZoomInfoD&B HooversCatch-up barrier: a years-built people dataset kept fresh by researchers and a contributory network
Evidence300+ researchers reverifying contacts every 90 days (2019), 15 acquisitions folded into one database (by 2022), and a contributory network that every new customer strengthens (2020, 2025 filings); a rival must rebuild the research staff, the acquired datasets and the contributor base at once

Network effects (15)

Additional relevant participants can make a product more useful to its existing users. Concept page: Network effects.

WinnerBeatWhat kept it ahead
AirbnbHomeAwayMarketplace liquidity of individual hosts, with guests who come directly
Evidence79% of hosts signed up directly and 23% of new hosts began as guests (2019); 69% of revenue from repeat guests (2019); 91% of traffic direct or unpaid (2020); most listings only on Airbnb; 430M reviews (2020); global short-term rental share 28% to 44% (2019-24)
athenahealthPractice FusionPayer-rules knowledge compounding with network claims volume
EvidenceEach denial became a rule for all clients: 100+ rules a month (2007); 'industry's largest' payer-rules database (2016). 93% first-pass vs est. 70%; 97% renewals despite 90-day exits
AutodeskIntergraphDWG exchange standard and the AutoCAD-trained workforce
EvidenceOver a million users and billions of drawings (1993); nearly 70% of the North American Windows CAD installed base, ~1.4M seats, and any rival engine must read AutoCAD files because users share drawings (FTC, 1997); 14 rivals, Intergraph among them, formed OpenDWG to read and write DWG (1998)
Booking HoldingsDespegarMarketplace liquidity
EvidenceMore than 1,115,000 properties by the end of 2016 and 1,049 million room nights in 2023; gross bookings rose from $68.1B (2016) to $150.6B (2023); first among Brazilians' travel sites by January 2022.
EtsyDaWandaMarketplace network effects: makers and buyers rivals could not pull away
EvidenceDaWanda's founder called Etsy 'the best choice for our community' and referred 70,000 sellers there (2018); Amazon Handmade drew 13,547 sellers to Etsy's one million (2017); after the 2022 fee rise to 6.5%, about 1% of sellers struck and revenue rose 10.6%
LinkedInMonsterNetwork effect: the professional pool recruiters must search
EvidenceRecruiting products compete on usefulness that depends on 'the number and engagement of our members' (S-1, 2011); Hoffman: 'As LinkedIn grew, so did our competitive edge, as more and more members invested activity and data into our network' (2013); Talent Solutions $1,877M, about 63% of 2015 revenue; a recruiter without LinkedIn Recruiter is 'hamstrung from day one' (Acquired, 2016); Microsoft paid $26.2B for 433M members (2016).
LiveRampOracle Data CloudProtocol network effect: RampID as the shared key that brands, publishers and platforms all accept
EvidenceRampID distributed through direct integrations to more than 500 partners (FY2026); 400+ publishers and 70 demand and supply platforms on ATS, including 70% of the US Comscore 20 (May 2021); Howe: 'the bigger we get ... the harder it is for anyone to replicate that value' (2019); Oracle's OnRamp, the only competing onboarder, closed in 2024, with LiveRamp named 'a huge gainer'
MastercardDiscover (Discover Network)Payment network effects
Evidence$10.6 trillion of gross dollar volume and 175.5 billion switched transactions in 2025; the Second Circuit found merchants could not refuse Visa or MasterCard despite significant price increases; about 100 million merchant locations against Discover's 70 million; Visa and Mastercard more than 80% of US card transactions in 2023, Discover about 5%.
MicrosoftAppleApplications barrier to entry (Windows developer ecosystem)
EvidenceWindows supported over 70,000 applications and held above 90% of Intel-compatible PC operating systems for a decade (court findings, 1999); users would not switch to the Mac OS even after a sustained OS price rise; IBM's OS/2 Warp stalled at about 2,500 applications and IBM stopped courting developers in 1996
OpenTableLivebookings (Bookatable)Marketplace liquidity
EvidenceMore than 31,000 restaurants and 14.7 million seated diners a month in late 2013; about one-third of North American reservation-taking restaurants in 2012; Jordan says restaurants that tried free rivals came back within days because the diners stayed on OpenTable.
PayPalBillpoint (eBay Payments)Payment network effects
EvidenceTwo-sided network of 439 million active accounts across about 200 markets (2025); accepted at 77% of the top 500 US internet retailers in 2019, about 65 points ahead of Amazon Pay, the next most accepted wallet; volume kept growing after eBay moved its own payments to Adyen from 2018.
TableauQlikExpertise network effect: the largest pool of trained analysts, and employers who hire for Tableau
EvidenceEmployer demand for Tableau skills up 1,581% from 2011 to 2015 (Burning Glass); 17,000+ at Tableau Conference and over 1 million visualizations on Tableau Public by 2018; Public profiles used as hiring portfolios
TinderMatch.comMarketplace liquidity: the largest local pool of young singles
EvidenceLargest Match Group brand by MAU and most downloaded dating app in North America by 2015, with 9.6 million daily users and more than 1.4 billion profile swipes a day, 86% of users under 35; 10.0 million payers in the fourth quarter of 2023 and $1.92 billion of direct revenue that year, against $691 million for Match and the other Evergreen & Emerging brands combined. Tinder's own payers fell 8% in Q4 2023 after pricing changes.
UiPathBlue PrismExpertise network effects
EvidenceNearly 1.0 million automation professionals in the UiPath community and over 800,000 Academy enrollees by early 2021, against 250,000 registered Blue Prism ecosystem users in October 2020; 98% gross retention in early 2022. Customers stayed while expansion slowed: net retention fell from 145% to 109% by July 2026 and ARR growth from 59% to 12%.
VisaDiscover (Discover Network)Payment network effects
Evidence39.8% of the US credit card market in fiscal 2004 against Discover's 7% and MasterCard's 30.2%; in calendar 2024, $13,433 billion of payments volume on 4,805 million cards against Mastercard's $8,014 billion on 3,146 million, both counting credit, debit and prepaid; Diners Club and Discover, counted on credit cards only, $253 billion; market value of $681.6 billion in October 2026.

Scale economies (13)

Unit cost falls with volume; the leader's cost advantage is the barrier. Concept page: Scale economies.

WinnerBeatWhat kept it ahead
AffirmAfterpayScale economies in funding and underwriting
EvidenceFirst securitization July 2020; AAA senior notes (2024); over $22B funding capacity from 150+ capital partners and 20+ ABS deals over $10B (Dec 2024); $36.7B GMV and 23M active consumers (FY2025) spread fixed underwriting, capital-markets and platform costs
AmazonBarnes & NobleScale and scope economies in a fulfillment network shared by Amazon and its sellers, filled by Prime demand
EvidenceFulfillment centers grew from 13 (2005) to 109 (2015) as FBA sellers stowed inventory in them; third-party sales grew from $0.1B to $160B at 52% a year (1999-2018), credited by Bezos to FBA and Prime; Amazon held about half of U.S. print book sales to Barnes & Noble's fifth (2018)
CarvanaShift TechnologiesScale economies in a national reconditioning and delivery network that took a decade and about $10B to build
EvidenceUnits +33% (2024), +43% (2025) through built plants as per-unit costs fell; rivals face '13 years and $10 billion invested' (2026) and $30-35M per 40k-unit site.
ChewyWag.com (Quidsi, owned by Amazon)Scale economies in a pet-only fulfillment and buying operation (qualified candidate)
EvidenceSales volume, Autoship participation and low seasonality lower fixed and variable cost per unit, and larger purchases improve supplier terms (S-1, 2019); seven centers reach 80% of the US overnight; $80.5M quarterly net income (Q2 2026); Wag.com closed unprofitable (2017)
DoorDashGrubhubScale economies and local marketplace density (qualified candidate)
EvidenceOrder volume spreads a shared platform cost, contribution profit per dollar of order value rose from 3.7% to 4.7% between 2023 and 2025, and merchant, courier and consumer density reinforce each other in each market. Uber Eats at 25% and a near tie in New York show the lead is not exclusive.
FortinetSonicWallScale economies in custom security chips (qualified candidate)
EvidenceChip design 'needs a very big, long-term investment' (Xie, 2021); 36.8% of firewall, UTM and VPN units and the most shipments nine years running (IDC via Fortinet, 2021); chips forward-ordered 'millions versus tens of thousands' (2023); a rival starting a chip program would spread the same fixed cost over far fewer units (reasoned, no unit-cost data)
HelloFreshBlue ApronScale economies
EvidenceProcurement savings and economies of scale lifted contribution margin from 23.0% of revenue in 2017 to 27.4% in 2018 and 28.5% to 28.7% in 2019, with fulfillment automation from 2018 to 2020; the US business, about half of US home-delivered meals in late 2019, turned profitable on adjusted EBITDA that year; group revenue reached EUR 7.66B in 2024.
Hinge HealthOmada HealthScale economies in automated MSK care (emerging)
EvidenceCare-team hours 97% below traditional PT (2025); 47% more members on flat care-team cost, 83% gross margin, record head-to-head win rate (2025)
IQVIAScience 37Scale economies and accumulated data assets
EvidenceContracted R&DS backlog of $32.7 billion at the end of 2025 on $16,310 million of revenue; more than 1.2 billion non-identified patient records and about 56 petabytes of data from about 150,000 suppliers (2021); its CEO said in 2021 that the more technology it deployed, the fewer people it needed and the more margin it made, while customers would pay no more.
NetflixBlockbusterScale economies: fixed programming costs spread over the largest membership
EvidenceStreaming rights bought for fixed fees (2008 on) and exclusive series (2011 on) cost the same however many members watch; spread over 21.4 million US streaming members (2011) and over 70 million members in 190 countries (2016), a smaller rival pays more per member for a comparable catalog
NubankItaú UnibancoScale economies: the lowest cost to serve and cheap deposit funding spread over the largest customer base in Brazil
Evidence131M customers, 113M in Brazil (62% of adults); cost to serve $0.8 per active customer a month; efficiency ratio 19.9% vs Itaú's 36.9% in Brazil; $41.9B deposits funded at 87% of the interbank rate (Q4 2025)
OneMain FinancialOportunScale economies
EvidenceOver 20% share of about $100B of nonprime personal loans (2023); $4.6B raised in 2023 while some competitors struggled for funding; operating expense ratio under 7% against Oportun's 15%, with opex per loan flat from 2019 to 2023 (Findell, 2025).
Tempur Sealy (Somnigroup)CasperScale economies in making, branding and selling mattresses
Evidence$7,476.5 million of 2025 net sales and $1,306.0 million of adjusted EBITDA across its manufacturing, Tempur-Pedic, Sealy and Stearns & Foster, and Mattress Firm stores; in 2020 Casper's own filing said larger rivals had greater brand recognition and fulfillment infrastructure and could acquire customers at lower cost.

Brand and trust (4)

Accumulated reputation creates buyer preference by reducing uncertainty or increasing perceived value. Concept pages: Branding, Trust.

WinnerBeatWhat kept it ahead
CertaraSimulations Plus (GastroPlus)Catch-up barrier: a regulatory track record and validated library rivals must rebuild
EvidenceSimcyp in 80.5% of PBPK-supported FDA novel approvals 2019-2023 against 5.2% for GastroPlus; 375+ label claims informed in place of clinical studies (2024); EMA's first and only PBPK platform qualification (2025) lets sponsors skip re-establishing the model's credibility; library upkeep of 40+ person-years a year (2013); software renewal 90% (2019-20) and software revenue up 18% in 2025 while services grew 3%
DuolingoBabbelBrand default plus streak switching costs: the name people search for, and a daily count learners will not break
Evidence'Duolingo' drew nine searches to each for 'learn Spanish' (2021) and about two thirds of the UK language-app market on about $100K of European ad spend (2021). About 15 million learners held 365-day streaks at end-2025 and 15.4 million revived lost streaks in June 2026; a rival can copy the streak but not a learner's count. DAUs rose 23% to 58.7 million in Q2 2026 after chatbots arrived
OnAllbirdsBranding
EvidenceA premium running reputation built through selective specialty retail held its price as the brand spread to everyday buyers: gross margin rose from 54.3% in 2020 to 62.8% in 2025, credited to full-price selling, on net sales that grew from CHF 425.3 million to CHF 3,014.0 million.
ZillowTruliaDefault-destination brand with scale economies in audience acquisition
EvidenceTwo-thirds of traffic direct with almost no ad spend (2011); about 80% direct and 2x rivals' daily app users (2025) after CoStar outspent it about 4x (2023).

Demand ownership (6)

Control of the customer relationship gives an intermediary bargaining power over suppliers. Concept page: Demand ownership.

WinnerBeatWhat kept it ahead
Booking HoldingsExpediaDemand ownership over Europe's independent hotels, reinforced by marketplace network effects
EvidenceHotels cited its 'larger customer base and higher website traffic' (2013); 71% of European OTA hotel bookings (2023); Spain fined it €413M over a 70-90% share and parity terms (2024)
CompassOpendoorDemand ownership: the agent network that holds the seller relationship
EvidenceNearly 90% of US sellers and buyers work with agents despite iBuying (Compass S-1); Compass retained over 90% of principal agents in 2020 and 96.8% quarterly in Q4 2025; the Anywhere purchase (closed January 9, 2026) was expected to take its network from about 40,000 to about 340,000 agents worldwide, with a UBS estimate of about 18% combined share. Opendoor, which built its offer to bypass agents, said in 2025 that a meaningful percentage of its acquisitions came through agents and rolled out partner agents in every market.
Credit KarmaCredit SesameDemand ownership: lenders depend on its member reach (qualified candidate)
EvidenceTwo of the five largest card issuers shared approval models to reach members (2018); 100+ financial partners, 100M+ members, nearly $1B revenue (2019); Credit Sesame about 15M members (2021)
GoodRxSingleCareDemand ownership: the default place to check a drug's cash price, with codes kept on file for refills
Evidence80%+ repeat activity (2020); codes saved in pharmacy profiles so refills need no re-presenting (2025); significant unpaid traffic from word of mouth, clinicians and press (2020, 2025); top-three PBM share of revenue fell from 61% (2018) to 22% (2025) as more PBMs supplied prices
GoogleKayakDemand ownership
EvidenceMetasearch sites and travel suppliers depend on Google's results page for traffic. Kayak's CEO said in 2024 that Google dominates search and traffic acquisition, and in 2025 that Google is by far Kayak's biggest third-party channel, with some lost free traffic replaced by paid Google traffic; Booking Holdings wrote the Kayak brand down by $457 million in 2025, citing higher customer acquisition costs. Google's 2024 test removing hotel features from European search cut traffic to hotels by more than 10%.
ZillowRedfinDemand ownership
EvidenceThe most visited US real estate site (Comscore, 2021); 100 million unique visitors in October 2023 against Redfin's 35 million; 204 million average monthly unique users in Q4 2024 on $2.236 billion of 2024 revenue, paid mostly by agents and other professionals buying access to that audience, from 2016 at auction prices and from 2018 as a fee due at closing (Flex); in February 2025 Zillow paid Redfin $100 million to become the exclusive multifamily listings provider on Redfin's rental sites.

Scope economies and bundles (8)

Shared infrastructure makes each additional product cheaper than a standalone rival's. Concept pages: Scope economies, Product tying & default bundles.

WinnerBeatWhat kept it ahead
AmazonStitch FixPrime bundle
EvidenceTry-on was included in Prime at no extra charge from 2017; Prime had 200 million members worldwide in 2020; Amazon sold an estimated $68 billion of US apparel and footwear in 2023, 41.4% of online apparel (Wells Fargo). Amazon's own styling and try-on features stayed small: it ended Try Before You Buy in January 2025 because the program reached only a limited number of items and customers.
AtlassianPagerDutyProduct tying and default bundles
EvidenceOpsgenie's alerting and on-call capabilities are sold only inside Jira Service Management since June 2025, so customers buy them with the service desk; PagerDuty's fiscal 2026 10-K says larger competitors bundle competing products with other software, often at a reduced price; Atlassian reported accelerating Service Collection growth and cross-sell in its fourth quarter of fiscal 2026.
CrowdStrikeCylanceOne installed agent carrying many products, costly to remove
Evidence147% net retention at the 2019 listing; half of customers on six-plus of 32 modules by 2026; retention above 97% after the 2024 outage
InstacartAmazonScope economies, with shared inventory learning
EvidenceThe same retailer relationships, transaction data and advertiser tools serve marketplace, retailer storefront and advertising offers. Rival entry can purchase some shared assets; retailer exclusivity is not asserted.
MicrosoftBoxProduct tying and default bundles
EvidenceContent storage, Teams and Copilot ship in one Microsoft 365 license on over 450 million commercial paid seats (reported January 2026); Teams provisions a SharePoint library for every team, and Copilot works on the documents stored there; Box's own filing names bundled enterprise license arrangements and lower-priced competitors as a source of pricing pressure.
SoFiEarnestScope economies: one member base, deposit funding and tech stack shared across products
Evidence51% of new products from existing members; deposits 156 bps below warehouse, ~$713M a year saved (Q2 2026)
SoFiLendingClubScope economies: one member base and one bank across many products
Evidence13.65 million members holding 20.2 million products at the end of 2025; 40% of new products in the fourth quarter opened by existing members; $37.5 billion of deposits that cost 181 basis points less than warehouse lines.
WorkdayGreenhouseProduct tying and default bundles
EvidenceRecruiting shares one record with Workday's core HR system, which more than 60% of the Fortune 500 used by early 2025; Workday's share of Fortune 500 applicant tracking rose from 22.6% (2019) to 39.2% (2025) and over 40% (2026).

Counter-positioning (3)

A superior business model the incumbent cannot copy without damaging its existing business. Concept page: Counter-positioning.

WinnerBeatWhat kept it ahead
CloudflareAkamaiCounter-positioning: free traffic Akamai's per-traffic model could not match, lowering Cloudflare's costs and training its defenses
EvidenceCDN and DNS free since 2010 on flat subscriptions, not usage billing (2019 letter); the free tier runs on idle capacity and its scale makes ISPs partners that lower colocation and bandwidth costs (2019); the system learns from every request (2011, 2014, 2019). Akamai's revenue stayed usage-based with unit-price declines at renewal (2015). Cloudflare fronted 26.2% of all websites to Akamai's 0.7% (Sept 2026).
HashiCorpAWS CloudFormationCounter-positioning: a cloud-neutral workflow the leading cloud would not offer
EvidenceAxon moved workloads from AWS to Azure on Terraform while Microsoft backed it to pull workloads from the market leader (2017-18); AWS co-built a generated Terraform provider with HashiCorp on its Cloud Control API instead of making CloudFormation cross-cloud (2021); Terraform 11.3% extensive use among developers vs Pulumi 0.8% (2023)
RobinhoodE*TradeCounter-positioning against commission-funded brokers (2015–2019)
EvidenceFor four years incumbents kept per-trade commissions that were part of their revenue; E*Trade guided to $6.60–7 per trade for 2019 with no advertised-rate cut (Jan 2019). When it matched $0 in October 2019 it gave up an estimated $75M a quarter (commissions about 17% of revenue) and agreed to sell to Morgan Stanley in February 2020. By then Robinhood had passed E*Trade in customers (2018) and took over half of US investing-app downloads in 2020.

Cornered resources and other barriers (4)

Barriers that rest on something rivals cannot get: a patent, a license, scarce property or a system customers are slow to replace. Concept pages: Cornered resource, Incumbent system inertia, Intellectual property, Scarce property.

WinnerBeatWhat kept it ahead
AMC EntertainmentMoviePassScarce property: ownership of the screens
EvidenceAMC operated 630 North American locations and held the #1 or #2 market share in 21 of the 25 largest US metro areas; it refused to discount or share revenue with MoviePass, and A-List was profitable in the fourth quarter and full year 2019 after price increases of 10% in ten states and 20% in five.
Guardant HealthFoundation MedicinePatent-protected lead in liquid biopsy: Foundation licenses Guardant's digital sequencing
EvidenceGuardant held about 75% of advanced-cancer liquid biopsy vs about 21% for Roche/Foundation (Leerink, 2018); Guardant sued over FoundationOne Liquid (2020); Foundation paid $25M plus royalties for a license to Guardant's digital sequencing patents (Jan 2022)
Palo Alto NetworksCheck Point SoftwareIncumbent inertia: Check Point's high-margin model kept it from matching the rebuild and then the spending (2007-2022)
EvidenceCheck Point bought NFR for intrusion prevention in 2006 but shipped a full next-generation firewall only in 2011; Shwed called Palo Alto Networks a niche player with little chance at its 2012 IPO; Palo Alto's R&D passed Check Point's in 2014 and was about 3x by 2020; its revenue passed Check Point's in 2017; by late 2022 it held about 24% of firewall sales to Check Point's 9% while Check Point kept a 45% operating margin and about $1.3B a year of buybacks, and it still spent $1.4B on buybacks in 2025 on $2.7B of revenue
PayPalKabbageCornered resource: the lender that holds the merchant's sales flow
EvidenceWorking Capital collects from the merchant's PayPal sales, which merchant cash advance providers cannot do (Aite, 2013); customers renew or re-access more than 90% of the time; more than $30 billion of loans and cash advances to more than 420,000 business accounts since 2013 (2025). PayPal tied its 2023 credit deterioration, with portfolio net chargeoffs of 18.8%, to risk limits widened on its Business Loans term product, which is not limited to PayPal merchants, and cut those originations.

Won without a lasting barrier (26)

These companies beat their rival, but our research found no barrier the rival actually felt. Each row says what won it instead.

WinnerBeatWhy it won
CompassAGNT / eXpeXp recruited many agents cheaply through revenue share. Compass paid star agents with venture capital, held them first with clawback contracts and then with an all-in-one agent platform, and bought whole brokerages until it became the largest brokerage.
CompassOfferpadOfferpad answered the home seller who wanted certainty by buying the house itself, so its capital sat in homes and its results followed home prices; it bought a fifth of Opendoor's iBuyer volume in 2021. Compass kept that seller on the open market with an agent, used $1.6 billion of venture capital to recruit agents and to front preparation costs and bridge loans repaid at closing, came through 2022 with transactions down 6% against an industry down 18% while 89% of buyers and sellers kept using agents, entered Phoenix in 2023 after Offerpad's 2022 retreat and bought Anywhere in January 2026. Offerpad turned to agent referrals, which brought a third of its 2025 requests; in the second quarter of 2026 Compass's brokerage closed 153,009 transactions to Offerpad's 295.
CursorWindsurfCodeium reached developers as a free plugin inside every editor and built enterprise sales early. Cursor took over the whole editor in 2023, filled frontier models' gaps with its own Tab and apply models, retrained them on users' accept-and-reject signals and let the product sell itself. By mid-2025 Cursor had $500M ARR to Windsurf's $82M; Windsurf then lost direct Claude access, its OpenAI sale and its founders to Google before Cognition bought the rest. Cursor keeps its lead by shipping faster, and its users can still switch tools at will.
DarktraceIronNetDarktrace's self-learning AI modeled each customer's own network with no attack data, rules or peer participation, so it could be installed free on a prospect's network and sold in volume through direct sellers and partners; IronNet's differentiator, collective defense, only paid off once many organizations agreed to share sensitive data, and it sold to a few large buyers through its founder's relationships. Darktrace reached 7,437 customers and $415.5 million of revenue in its year to June 2022 and was taken private at a valuation of about $5.3 billion in 2024; IronNet had 88 customers and $27.5 million of revenue in its year to January 2022, ran out of money in 2023 and drew no bidder among 114 approached in bankruptcy.
GitLabGitHub (Microsoft)GitHub sold hosting to developers and charged companies from the first seat to run it themselves; GitLab gave companies a free server for behind the firewall, sold each paid tier to a more senior buyer, and folded CI and the rest of delivery into one application. When Microsoft bought GitHub, GitLab became the platform tied to no cloud, and it kept growing on private company code after GitHub matched its free tier and CI.
Global-eESW (eShopWorld)ESW grew by carrying a few giant US brands, Nike above all, into Europe; Global-e took the seller-of-record model Borderfree had pioneered to many European brands on the stores they already ran, so every country a brand added paid it a fee and a DHL shipping margin. Shopify's checkout and the purchases of Flow and Borderfree then widened that base to $6.6 billion of GMV in 2025, while ESW was described as stagnating.
Hims & HersRoRo (Roman) out-raised Hims and was the larger company in 2020 ($250M sales against $149M), after both launched the same discreet online care within days in 2017. Hims got to profitable unit economics sooner, listed publicly and then outspent on a mass consumer brand and its own compounding pharmacies, whose personalized plans raised revenue per subscriber; by 2024 Hims had passed Ro, at about 2.5 times Ro's outside-estimated revenue.
Hinge HealthCastlight HealthCastlight sold self-insured employers price and quality search, billed on every eligible employee, then became a shelf of pre-integrated programs from other vendors. Hinge Health delivered treatment for one costly condition, counted a member only when engagement was billable, and sold through the benefits platforms employers already used, Castlight's among them; automation then cut its cost per member. Hinge reached a $6.2 billion private valuation in 2021 and $587.9 million of 2025 revenue, while Castlight's revenue stalled between $143.3 million and $156.4 million in 2018-20 and it sold to Vera for about $370 million in 2022. Castlight's core navigation accounts went to Accolade and Grand Rounds and to health plans' free tools, which its 2020 annual report named as its main competition.
Hinge HealthPear TherapeuticsHinge Health and Pear both sold software-delivered treatment and both counted employers and health plans among the buyers they wanted, though they treated different conditions: Hinge back and joint pain, Pear addiction and insomnia. Hinge sold a finished digital physical therapy program to self-insured employers and their health plans, which contracted before any member enrolled, often paid on milestones and were offered guarantees on engagement. Pear reached most patients through a clinician's prescription, a pharmacy and a payer's reimbursement, drew most of its revenue from state Access Agreements funded largely by government money, was paid on 41% of fulfilled 2022 prescriptions, and was sold at auction for $6.05 million in 2023 while Hinge reached 2,929 clients and an $8.45 billion market value.
HomeAdvisor (ANGI Homeservices)Porch GroupHomeAdvisor sent each homeowner's request to a few prescreened pros who paid for every match, so each request it bought with paid search and TV earned money at once; Porch listed 1.5 million pros free and asked a small share to subscribe, found holes in its own economics by 2015 and left the consumer marketplace for software. HomeAdvisor absorbed Angie's List in 2017 and reported $1.13 billion of revenue in 2018 as the largest digital home-services marketplace, without a barrier that kept its acquisition costs down.
InovalonCotiviti (Verisk Health)Verisk Health had a HEDIS engine certified since 2001, the same as Catalyst's, but grew by assembling analytics and later payment accuracy businesses. Inovalon bought Catalyst, which put it inside most plans' quality data, then sold those plans Medicare Advantage risk work done by its own clinicians and sold the pooled data to drug makers, reaching a $6.4 billion take-private.
Livongo HealthProteus Digital HealthLivongo made a habit people with diabetes already had, checking blood sugar, send its data automatically, and billed the employers and health plans that paid for their care for each enrolled member; Proteus asked patients to take a sensor-tagged pill, wear a patch and run a phone app, and came to depend on a drug maker partner. Livongo added conditions for the same clients and reached $170.2 million of 2019 revenue before Teladoc agreed a merger valued at $18.5 billion in August 2020, the month Otsuka bought Proteus's assets out of bankruptcy for $15 million.
LyftUberLyft won by reaching substantial business scale; Uber remained the larger continuing rival. This chain explains Lyft’s entry, survival and recovery without declaring Uber a loser.
MytheresaFarfetchMytheresa bought a curated edit for luxury's biggest spenders, sold it mostly at full price and stayed profitable, then took over YNAP with Richemont's cash in 2024-25. Farfetch aggregated boutiques' stock for a much larger, more discount-driven audience, spread itself across acquisitions and platform deals, never sustained a profit, and was sold in an English administration in January 2024 after its own share-funded YNAP deal collapsed.
NavanAmerican Express Global Business Travel (with Egencia and CWT)Amex GBT sold agency service on top of booking tools travelers bypassed, and grew by buying Egencia and CWT; TripActions built a tool travelers preferred and ran its own agents on the same trip record. When software took over most of that agent work, Navan's gross margin reached 74% and it began taking global programs such as Visa's from CWT.
Oak Street HealthBabylon HealthOak Street built its own primary care centers for low-income Medicare patients, gave each doctor about 500 patients, and took full risk from Medicare Advantage plans on a model that cut hospital admissions about 51% below benchmark; revenue reached $2,160.9 million in 2022 and CVS bought it in 2023 at an enterprise value of about $10.6 billion. Babylon took the same kind of capitated risk, mostly for Medicaid members, through an app layered on networks it bought or borrowed; its 2022 claims expense of $1,017.0 million nearly matched revenue of $1,109.7 million, and its US subsidiaries filed for Chapter 7 liquidation in August 2023.
Oscar HealthNeueHealth (Bright Health)Both sold narrow-network exchange plans at low prices. Oscar ran its own claims and data systems, grew on reinsurers' capital and retreated where it lost money; Bright raced into states it could not see the costs of and ran out of capital.
RampBrexRamp entered in 2020 as roughly the 200th corporate card and sold it as a way to spend less, with its own expense software free; it then sold bill pay, procurement and paid software to the same finance teams while keeping every size of customer and a burn under $2 million a month. Brex, first to the startup card, moved off payment volume toward enterprise software in 2022 and dropped tens of thousands of small businesses; Ramp passed it in volume by the end of 2023, reached $1 billion in annualized revenue in August 2025 and a $32 billion valuation that November, while Brex, at about $700 million, agreed in January 2026 to sell to Capital One for $5.15 billion.
RampParkerRamp and Parker were both founded in 2019 to sell charge cards to growing businesses. Ramp sold a card built to cut spending to any qualifying company, asked for the balance in full each cycle and set limits from the cash a customer held (as described in 2026), and grew to $1 billion in annualized revenue by August 2025; Parker gave ecommerce brands up to 90 days on each purchase, which tied it to warehouse lenders and an issuing bank that pulled away in 2026, and it filed for Chapter 7 that May.
Sprout SocialHootsuiteHootsuite gave individuals a free dashboard of columns and won users by the millions; Sprout built one shared inbox for a team and charged for it from day one, with a salesperson on the trial. Sprout's paying teams got bigger, reached a public listing and $458M in revenue, and in 2023 Hootsuite dropped its free plan to chase the same buyers.
SquarespaceWixWix gave its builder away to anyone and bought a larger user base with years of losses. Squarespace made every customer pay from 2005 for a designed, hosted site, kept the creatives who valued it, and spent their long subscriptions on measured podcast reads, winning the profitable design-led segment rather than the largest subscriber count.
TalkspaceBetterHelpBoth sold therapy by message. BetterHelp kept buying cash-paying consumers with advertising; Talkspace put a credentialed network inside health plans, so insurers paid for its patients when ad costs rose, and a hospital operator bought it for that in-network reach.
TeladocAmwellAmwell sold telehealth software that health systems and plans ran under their own brands, priced by members with access or forecast consultations, and billed its own clinicians' visits separately. Teladoc sold a finished service, its own on-call clinicians, for a monthly per-member fee paid by employers and plans, so revenue grew with covered lives rather than visits and funded a portfolio of care lines sold to the same buyers: $2.53B revenue and 100M+ members (2025) against Amwell's $249.3M.
TwilioNexmoNexmo sold wholesale SMS routes to high-volume apps; Twilio sold any web developer the phone network as a few lines of code, billed by the minute. Developers it recruited one hackathon at a time carried Twilio into their employers, so it became the default and stayed #1, though switching costs protect only its deeply embedded customers.
WizLaceworkLacework reached the cloud three years before Wiz was founded, with agents that learned each workload's behavior and cut alert noise for security and DevOps teams. Wiz needed only read access to a customer's cloud accounts to map every exposure in minutes, sold that view to the largest companies first, reached $100 million of ARR in 18 months and sold to Google for $32 billion, while Lacework's growth stalled near $100 million of ARR and it sold to Fortinet for $152.3 million.
ZscalerNetskopeBoth run cloud-native inline security, so Zscaler did not beat Netskope on architecture. It got there first: it was replacing web-gateway appliances at the largest distributed companies five years before Netskope launched as a SaaS-visibility tool, and it put private access on those seats three years before Netskope did, so by 2026 its ARR is about four times Netskope's.

Related

Hamilton Helmer's version of this list, with examples: 7 Powers. Every barrier type in the taxonomy: the Defensibility layer.