Source-linked AI summary
Scale, Concentration, and Entry Timing in the Shopify App Ecosystem: A Longitudinal Study of Platform Governance and Application Survival
Fabrizio Assabese, Peter-Jan Celis, Giuseppe Destefanis
TL;DR
The paper addresses limited longitudinal evidence on Shopify’s ecosystem structure, competition, platform intervention, and application survival by extending a snapshot study with a seven-year weekly panel and reconstructed listing histories. It finds that late-mover growth advantages persist, failure is often early and predictable, and platform governance can reshape competition asymmetrically.
Problem
Longitudinal evidence is needed to determine how Shopify competition evolves, how platform interventions affect categories, and which applications survive beyond a single snapshot.
Method
The study combines a September 2025 ecosystem snapshot with a weekly panel of 7,708 applications over 366 weeks and semi-annual listing histories reconstructed from the Internet Archive.
Results
Late-mover growth advantages persist across seven years, while most failure is silent and early; category leadership changed in 92% of categories.
Takeaways & Limitations
Shopify app markets remain contestable, but outcomes depend on platform governance and entry conditions, while later entry is a higher-variance strategy rather than a universally dominant one.
Takeaways & Limitations
Install counts combine active and abandoned usage, do not capture monetisation, and panel exit cannot distinguish delisting from loss of detectable storefront presence.
Abstract
from arXiv · showhide
The Shopify marketplace hosts more than 16,000 active third-party applications, serving 2.7 million active merchant stores generating an estimated \$706 billion in annual sales, yet little empirical evidence exists on its structure and adoption drivers. We analyse it using a September 2025 snapshot of all 24,826 applications, a weekly panel of 7,708 applications over 366 weeks (February 2019 to March 2026), and listing histories reconstructed from the Internet Archive. Across 55 established functional categories, over half were low-concentration and 20\% highly concentrated, with larger categories consistently less concentrated. Later entrants grew faster than early movers in 88\% of the 50 analysed categories, persisting across seven years of quarterly re-estimations, with early movers on net losing detected installations whilst late movers gained them. Platform governance reshaped competition asymmetrically: Shopify's entry into chat reversed more than two years of de-concentration; the deprecation of its Product Reviews application, to our knowledge the first measured platform-owner exit from complementor category, released 191,000 installations of which at most a third reappeared as competitor adoption within a year; the 2021 reduction of the platform revenue share produced no detectable change in entry or retention. Failure is largely silent and predictable: the median exiting application peaked at 8 detected installations and disappeared from tracking within 68 weeks of launch, category leadership changed hands in 92\% of categories over seven years, and publicly observable data from an application's first six months predict exit within two years with cross-validated AUC above 0.8. App markets on Shopify remain contestable; who benefits depends on platform governance and entry conditions more than on entry timing.
1 Introduction
Shopify’s app marketplace supports merchant commerce and software entrepreneurship, but its competitive structure and adoption dynamics required evidence beyond a single snapshot. This extended study adds longitudinal data to examine governance, survival, and the durability of entry-timing patterns.
- More than 16,000 active applications extend Shopify stores with functions including reviews, marketing, logistics, subscriptions, and social-media integration.
- A September 2025 snapshot found heterogeneous competitive structures, an inverse relationship between category size and concentration, and faster recent growth for later entrants in 88% of 50 categories.
- A single snapshot cannot show how competition evolves, how platform interventions affect categories, or which applications survive.
- The study adds a weekly panel and archived listing histories to analyse governance events, survival, and early-warning prediction alongside the original RQ1–RQ3 results.The new analyses cover platform governance, application lifecycle and survival, and prediction of exit.
- The extended article verifies the original findings, re-estimates entry timing at 26 quarterly points, and adds robustness checks and a replication package.
2 Related Works
Prior research has examined platform-owner entry, governance, and ecosystem structure, but large-scale quantitative evidence on B2B marketplace competition remains limited. This study addresses that gap with platform-wide measurements of Shopify’s category structure and entry timing.
- Prior platform research links owner entry to changes in complementor innovation, prices, growth, attention, and disintermediation across mobile and retail marketplaces.
- Systematic category-level concentration measurement across revenue or install bases remains scarce in platform ecosystems.
- Large-scale quantitative studies of B2B marketplace structure and post-entry growth dynamics comparable to iOS and Android research remain absent.
- The study contributes HHI and top-five market-share measures across 55 Shopify functional categories, documenting heterogeneous competitive structures.
- Later entrants grew faster than early movers in 88% of 50 analysed categories, while early movers lost detected installations on net.
3 Dataset
The dataset combines a September 2025 Shopify snapshot, weekly Store Leads installation observations, and Internet Archive listing histories. Linking these sources supplies longitudinal adoption and historical listing characteristics while requiring explicit handling of censoring and measurement limits.
- The September 2025 snapshot contains 24,826 applications and 2,701,805 live merchant stores, with application status, categories, pricing, installations, ratings, reviews, and vendor fields.
- The analytical sample retains active applications with at least one installation, yielding 4,213 installed active applications from 16,698 active applications.
- The weekly panel records detected application installations across 7,708 applications and 366 weeks from 18 February 2019 to 1 March 2026.
- Internet Archive captures reconstruct historical listing characteristics because Shopify publishes no historical App Store record.
- Panel applications are merged by handle, with semi-annual listing covariates assigned to weekly observations using the most recent prior snapshot.
- Entry-observed applications are separated from left-censored applications using snapshot creation dates, with 2,247 entry-observed and 5,023 left-censored applications.First panel appearance reflects tracking onboarding rather than market entry.
4 Methodology
The methodology measures ecosystem scale, category concentration, and entry-timing growth, then extends those analyses with longitudinal prediction and robustness-oriented sample restrictions. It uses installation-based market shares and mature, sufficiently populated categories to stabilize comparisons.
- The study analyses ecosystem scale descriptively, concentration with HHI and top-N shares, and entry timing through recent growth metrics rather than cumulative share.
- Market share is computed within each primary category from cumulative application installations.
- Categories require at least 20 active installed applications, and concentration is classified as low below 0.15, moderate from 0.15 to below 0.25, and high at least 0.25 HHI.
- Entry timing uses creation date as a proxy, excludes applications younger than 180 days, and decomposes growth into 90-day velocity and rate.
- Early and late movers are defined as the first and fourth creation-date quartiles within each category.
- The early-warning model uses first-26-week features and logistic regression to predict two-year exit, evaluated with five-fold cross-validated AUC.
5 Results
The Shopify app ecosystem has expanded rapidly while combining substantial diversity with category-specific concentration. Larger categories are less concentrated, and concentrated markets can coexist with late-mover advantages and varied competitive structures.
- Ecosystem Scale: 16,698 of 24,826 applications (67.3%) were active, while 74.8% of active applications recorded zero installations.The zero-install rate may reflect both web-scraping limitations and genuine market-entry barriers.
- Temporal Trends: 4,106 applications were created in 2024, up from 203 in 2015, representing more than twentyfold growth over nine years.The 2025 count is partial and therefore is not directly comparable with complete years.
- Concentration Within Categories: 11 of 55 categories (20.0%) were highly concentrated, 16 (29.1%) moderately concentrated, and 28 (50.9%) low-concentration.High concentration was defined as HHI ≥0.25, moderate as 0.15 ≤ HHI < 0.25, and low as HHI <0.15.
- Category Variation: Social proof had HHI = 0.678 and Instafeed held 81.9% share, whereas upsell and cross-sell had HHI = 0.059 and its leader held 16.1%.These examples show that high adoption or category importance does not uniformly imply concentration.
- Concentration Within Categories: Larger categories had lower mean HHI, indicating that more applications sustained viable market shares as category size increased.This inverse relationship contrasts with an expectation of stronger winner-take-all effects in larger markets.
- Entry Timing: HHI was uncorrelated with entry-timing correlations, while all ten highly concentrated categories in the comparison favoured later entrants on both growth measures.The reported Spearman correlations were ρ = 0.01 for velocity and −0.05 for rate.
6 Discussion
The longitudinal evidence shows that Shopify’s app ecosystem generally de-concentrates and remains contestable, while platform actions and dominant leaders can reverse that pattern. Entry timing benefits surviving late movers but also brings higher silent-failure risk, and early public signals predict exit.
- Market concentration: 53 of 68 categories de-concentrated significantly from 2019–2026, while 13 concentrated and two showed no significant trend.The strongest concentrating cases were chat after platform entry and social proof around a dominant application.
- Platform governance: Platform entry into chat reversed more than two years of de-concentration through the platform application’s own share capture.Platform build decisions can compress the space available to competitors.
- Platform governance: At most a third of Product Reviews’ released installed base reappeared as competitor adoption within a year after platform exit.The category leader’s growth slowed across the event, and most of the free default application’s installed base did not reappear.
- Entry timing and survival: Category leadership changed in 92% of categories over seven years, showing that late-mover advantage can reach the top rather than remain confined to smaller applications.The advantage is persistent across seven years of rolling analysis, although late entry remains higher variance.
- Entry conditions: Category selection trades room to grow against crowding at entry, and the positioning mechanism cannot be established from the available data.Higher concentration appears protective in hazard models until early traction is held constant, when concentration becomes a risk factor.
- Failure prediction: Six months of publicly observable data predicted exit within 104 weeks with cross-validated AUC above 0.8.The early-warning model indicates that applications likely to fail are distinguishable within their first half-year.
- Practical implications: For platform operators, build-and-retire decisions restructure categories within quarters, whereas commission changes showed little detectable effect on entry or retention.The practical implication is that governance deserves scrutiny comparable to pricing.
7 Threats to Validity
The study’s validity is constrained by measurement, coverage, sampling, and generalisability boundaries. These include imperfect installation and exit measures, a commercially active panel subset, uneven archival data, Shopify-specific events, and temporal or data-quality limitations.
- Measurement: Revenue estimates measure merchant e-commerce sales rather than app subscription revenue, so $706 billion contextualises merchant scale but not developer revenue.The figure should not be interpreted as app-market or developer revenue.
- Construct validity: Install counts combine active and abandoned usage, do not capture monetisation, and mechanically advantage older applications.Relative within-category metrics and temporal velocity decompositions mitigate some concerns, but creation date also conflates age with launch sequence.
- Construct validity: Panel exit is disappearance from tracking, which cannot distinguish delisting from loss of detectable storefront presence.Detection can also temporarily lapse when an application changes its storefront code and later reappears under a new signature.
- Internal validity: The panel requires categories with at least 20 active apps and applications aged 180+ days in RQ3, limiting inference for small markets and launch-stage volatility.Early and late movers are selected from the first and fourth creation-date quartiles to maximise contrast while retaining sample size.
- Coverage and censoring: Panel coverage grows from 779 applications in 2019 to 6,307 in 2026 and is skewed toward established applications.Results on panel trajectories are conditional on the tracked population, despite snapshot linkage and entry-observed restrictions.
- Archival reconstruction: Historical covariates depend on uneven Wayback Machine coverage, semi-annual sampling, and a carry-forward assumption that characteristics persist between snapshots.Short-lived price changes may be missed, while repricing is assumed to be infrequent.
- External validity: Governance-event findings concern one platform’s actions and generalise only as hypotheses, while concentration and obsolescence patterns may vary with governance, maturity, and lock-in.The study is based on Shopify, which targets SMEs, uses subscription pricing, and curates its store.
- Temporal validity and data quality: The original growth analysis used a single 90-day window, although the panel re-estimated entry timing at 26 points across seven years.Weekly scraping also misses rapid changes, and traffic-model revenue estimates introduce absolute error while preserving relative patterns.
8 Conclusion
The longitudinal study confirms that Shopify app markets remain contestable while showing that governance events and entry conditions shape outcomes more than entry timing alone. It also finds that application failure is often early, silent, and predictable.
- Most categories de-concentrated as they grew, and the late-mover advantage persisted across seven years rather than reflecting one measurement window.
- Shopify’s governance actions affected competition asymmetrically: platform entry re-concentrated chat, while product-review exit released installations that mostly did not reappear among competitors.The revenue-share reduction left entry unchanged and retention without a detectable response.
- The median exiting application peaked at eight installations and disappeared within sixteen months, while category leadership changed hands in 92% of categories.Later cohorts reached zero visible adoption faster at equal age.
- Six months of publicly observable data predicted two-year exit with cross-validated AUC above 0.8, driven by early installation growth and the first app-store review.
- App markets remain contestable, but beneficiaries appear to depend more on governance events and entry conditions such as freemium pricing and early review traction than on entry timing.The authors identify testing other platforms and explaining resilient early movers as future work requiring information beyond public listings.
Declarations
The authors disclose an employment-related competing interest and describe the independent source of adoption data. They also report a replication package containing scripts, code, and archival outputs, while noting licensing limits on redistributing Store Leads data.
- The first two authors are employed by Judge.me, and adoption data comes from the independent third-party provider Store Leads.
- The panel analyses are fully scripted in the replication package, and event dates derive from public announcements.
- The replication package includes analysis scripts, figure-generation code, archival reconstruction tools, and extracted public-source listing metadata.
- Store Leads data cannot be redistributed under its commercial licence, including snapshot exports, installation counts, and review counts.