Gaming desktop apps take 11% of web bid requests - and fool viewability tools

Jounce Media today published its June 2026 Supply Path Benchmarking Report, revealing that desktop apps - installed Windows and Mac applications that render ads through a browser-like Chromium layer - now account for 11.2% of all web bid requests. That figure places the desktop app category ahead of every single web publisher in the bidstream, including Yahoo at 10.9%, Raptive at 7.0%, and Mediavine at 3.8%. The finding has immediate consequences for programmatic buyers, who in many cases have been purchasing this inventory without any clear signal they were bidding on an application rather than a website.

The report, which draws on Jounce Media's data co-op covering bidstream activity across 2,151,472 websites, 5,488,245 mobile apps, and 11,733 connected TV apps during May 2026, marks the first time the consultancy has formally classified desktop apps as a distinct environment in its client-facing outputs.

What desktop apps actually are

The category is built almost entirely on one technical foundation: the Electron framework. Apps like CurseForge - the largest desktop app by bid volume in Jounce Media's dataset - are constructed using Electron, which wraps a Chromium-based web rendering engine inside a standalone desktop application shell. To a user, these products behave like native applications. They have their own icon in the taskbar, open in a dedicated window, and are installed through a standard setup process. But underneath, ads and most content are rendered by the same Chromium engine that powers Google Chrome and Microsoft Edge.

That dual nature creates a structural problem for the programmatic ecosystem. Each bid request must contain either a site object or an app object. The site object covers traditional web inventory. The app object covers non-browser environments like mobile apps and connected television. Desktop Electron apps fall awkwardly between the two. According to Jounce Media, every Electron app in its dataset presents inventory using a web object - meaning bid requests declare a domain like cforge.overwolf.com and rely on DSP buyers to understand that this domain is an alias for a standalone application, not a conventional website.

CurseForge is the largest single source within the desktop app category. Gamers use it to browse and install community-created modifications - known as mods - for titles including Minecraft and World of Warcraft, tracking live game statistics and altering gameplay mechanics from within the application. Overwolf acquired CurseForge from Twitch in 2020 and monetizes the ad inventory through a mix of direct campaigns and programmatic auctions. Beyond CurseForge, Overwolf provides ad monetization services to more than 200 additional desktop apps under an exclusive representation model comparable to what Raptive and Mediavine provide for web publishers.

According to Jounce Media, 241 total desktop apps are currently active in the bidstream. Of those, 215 are powered by Overwolf. That concentration means Overwolf-managed apps represent 68% of all desktop app bid requests, with the next seven apps - including Blitz, Wand, Feather, Medal, Modrinth, Bluestacks, and U.GG - collectively accounting for more than 30% of the remaining volume. Approximately 30 additional applications operate outside the Overwolf stack entirely.

How user agents disambiguate the environment

The most widely adopted method for signaling that an inventory source is a desktop app rather than a website is the user agent string, carried in the device.ua field of each bid request. Standard browser user agents - the strings identifying Chrome, Edge, Firefox, or Safari - are shared across millions of websites, so they provide no meaningful signal about any individual publisher's environment. A bid request for cnn.com and a bid request for espn.com will carry nearly identical user agent distributions.

Desktop apps, by contrast, should carry app-specific user agent strings that contain a human-readable token uniquely identifying the underlying application. According to Jounce Media, as of late May 2026, all major desktop app developers have implemented this approach. The CurseForge app, for example, includes the token "CurseForge/1.305.1-34041" alongside Chrome version and Electron version identifiers. The Blitz app surfaces "Blitz/2.1.573". Wand uses "Wand/12.25.1". These strings do not overlap with browser user agents or with each other, creating what Jounce Media describes as a strong signal that the associated domain is an alias for a standalone application rather than a conventional website.

Using this uniqueness criterion, Jounce Media has now added a fifth environment classification - Desktop App - to its Property Scores file, which previously covered web, mobile app, and CTV. The same classification is now present in the company's Seller Profiles dashboard, Exchange Profiles dashboard, and Data Transfer service. Buyers and technology platforms can use this signal to either target or explicitly exclude desktop app inventory depending on their campaign objectives.

Concentration and scale in context

The 11.2% figure carries a specific methodological context. Jounce Media's data co-op pools media metrics from global programmatic buyers across every active web, mobile app, and connected TV supply path. The desktop app share is measured as a proportion of web bid requests - meaning the comparison group is limited to inventory presented with web objects. Within that dataset, no single web publisher comes close to 11.2%. The share is larger than Yahoo's entire web footprint as measured by the same methodology.

According to Jounce Media, the average RTB-enabled publisher monetizes through 29.2 directly integrated sell-side technology platforms and authorizes 19.5 of those partners to initiate resold auctions. Rebroadcasting supply chains account for 44% of display auctions and 26% of video auctions. Across the 100 Bellwether portfolios that Jounce tracks as trusted sources of premium supply, those publishers collectively control inventory capturing 71% of DSP spend.

The desktop app category exists largely outside the Bellwether framework. Overwolf itself scored unfavorably in Jounce Media's 2025 Sentiment Survey, which measures buyer trust across programmatic supply sources. Now.gg, the publisher of the Bluestacks app, was identified as the least trusted publisher in that survey. That skepticism, according to Jounce Media, reflects real quality concerns that the May 2026 research surfaced in detail.

The background refresh problem

Desktop applications have characteristically long user sessions. A gamer tracking stats in Porofessor or installing Minecraft mods in CurseForge may keep the application open for hours. That session duration creates both an opportunity and a risk for ad monetization: more time on-screen means more potential impressions, but it also creates incentives to refresh ad placements aggressively - including in states where the app is not actually visible to the user.

Jounce Media benchmarked ad refresh behavior across the top three Overwolf apps and the seven largest non-Overwolf apps in four distinct states: in view (the app is open and fully visible), occluded (the app is open but covered by another window), minimized, and closed. As of late May 2026, every major desktop app in the sample only refreshes ads in the in-view state. That is the current picture. The prior picture was substantially worse.

In early May 2026, Jounce Media's evaluations found background refreshing behavior across five apps: Blitz, Wand, Feather, Modrinth, and Bluestacks. When these applications were occluded, minimized, or even after the user pressed the close button - which in some cases moved the app to the system tray rather than terminating it - ads continued to fetch and render, typically every 30 seconds. According to Jounce Media, the company contacted each of the five developers and indicated it intended to publish the findings. Over the following four weeks, all five disabled background refreshing. Wand and Blitz did so shortly before publication. For Bluestacks, the change produced a 77% drop in bid request volume compared to the April 15 to May 14 baseline period. Feather and Modrinth saw similar declines exceeding 70%.

The arithmetic is direct. If disabling background refresh reduces impression volume by more than 50%, then more than half of previously sold impressions must have been generated when zero pixels were visible to the user. According to Jounce Media, background refreshing was active in five apps that historically represented approximately 25% of all desktop app supply.

Why viewability metrics failed to detect this

The more troubling aspect of the finding is that standard viewability measurement appeared to confirm high quality on these same apps throughout the period when background refresh was active. According to Jounce Media's data from the 90 days preceding the report, viewability vendors were able to apply their measurement technology to more than 80% of impressions. Among measurable impressions, at least 70% were reported as viewable. Bluestacks recorded 97.6% reported viewability. Wand registered 95.3%. Modrinth came in at 89.8%. Digital Ingot at 87.2%. Blitz.gg at 73.3%.

These figures look, on their face, like strong premium inventory metrics. The reason they are not is rooted in a technical limitation specific to the Electron framework.

Viewability measurement on Chromium-based environments typically relies on two browser APIs. The Intersection Observer API determines whether an ad is positioned within the visible viewport. The Page Visibility API reports whether the application window itself is visible or hidden. Together, they allow a verification vendor to determine both that an ad slot is in view and that the user is actually looking at the screen.

The problem emerges when desktop apps enable background refresh. To keep ads fetching and rendering in the background, developers must disable a setting in the Electron framework called backgroundThrottling. According to Electron's own developer documentation, cited in the Jounce Media report: "If backgroundThrottling is disabled, the visibility state will remain visible even if the window is minimized, occluded, or hidden."

The result is that a viewability vendor receives two signals: the Intersection Observer reports that the ad is positioned in the viewport (accurate), and the Page Visibility API reports that the viewport is visible to the user (inaccurate, because backgroundThrottling has been disabled). Vendors relying solely on these app-provided signals will report high viewability for sessions where the user has minimized the application or covered it with another window.

According to Jounce Media, some sophisticated viewability products appear to have correctly identified low viewability in apps with background refresh enabled, possibly through novel techniques that go beyond the standard Chromium API signals. The firm encourages buyers with historical desktop app exposure to audit their viewability data against the findings to assess measurement accuracy. This connects to a broader pattern of viewability metrics being used as proxies for quality in contexts where they can be manipulated - a dynamic previously documented in the context of MFA supply and attention measurement research.

Jounce Media frames this explicitly: "When the metrics are too good to be true, they often are." The firm notes that these superficial metrics have previously steered buyers toward MFA supply, and that the same mechanism is now directing spend toward low-quality desktop app inventory. That pattern has downstream effects across the supply chain. DSPs and their customers bid, win, and pay for impressions that no consumer can see. SSPs onboard publishers that do not meet baseline marketplace quality standards, exposing themselves to clawback risk. Publishers that avoid background refreshing - including Overwolf, Medal, and U.GG, according to Jounce Media - lose market share as a direct consequence of their more responsible practices. Supply path optimization tools have struggled to detect this category of quality issue because the metrics themselves appear clean.

Three practices the industry can align on

According to Jounce Media, based on discussions with every major desktop app developer, three practices would go a long way toward building buyer confidence in the category.

The first is environment clarity. Desktop apps should submit bid requests with a dedicated user agent that unambiguously identifies the underlying application. The domain used to monetize the app should be dedicated exclusively to that app and should not double as an ad-supported website.

The second is refresh practices. Ad refresh should only occur when at least 50% of the ad slot's pixels are in view - the MRC viewability threshold for standard display. No bid requests or win notifications should be issued when the application is occluded, minimized, or closed.

The third addresses viewability signals directly. MRC-accredited viewability vendors should validate that ads only qualify as viewable when the application viewport is genuinely visible to the user. If vendors detect apps that trigger false positives due to disabled backgroundThrottling, their reporting should mark those impressions as unmeasurable rather than viewable.

Other notable findings from the June report

Beyond the desktop app focus, the June 2026 report includes several other data points relevant to the broader programmatic market.

In the exchange coverage rankings, Magnite holds 96.0% overall supply coverage across its dataset, with 99.3% in the CTV segment specifically - a figure consistent with its March 2025 benchmarking position. PubMatic ranks second overall at 89.7%, and Index Exchange and Google tie for third at 85.8% each.

Among May's notable exchange movements, Criteo's coverage rose 4.1 percentage points to 67.8%, driven primarily by a new integration with Fox to monetize the Tubi TV streaming service. Teads grew 1.6 percentage points to 32.3% following a new partnership with Scripps covering content on The Roku Channel, Samsung TV Plus, and Vizio Watchfree+. Rise jumped 3.4 percentage points to 20.2% after launching a Mediavine partnership, with Minute Media recording a 2.8-point gain from the same relationship.

In sustained supply shifts, DMG Group migrated all traffic from dailymail.co.uk to dailymail.com in late April. The content and ad experience are unchanged, but the domain change requires updates to targeting and inclusion lists. MTN TV, a streaming service for Apple TV, grew from near-zero bid request volume in Q4 2025 to a top-30 CTV supply source - but more than 99% of its bid requests are initiated by unauthorized resellers, and Jounce classifies the inventory as high risk.

For cyclical surges projected in June, the report highlights MLB-related properties as the largest seasonal opportunity. milb.com is forecast for a 137% increase above its January baseline, MLB mobile apps for 107%, and mlb.com for 54%. SportsEngine apps managed by NBCUniversal are projected at a 239% surge. These figures represent historical seasonal patterns, not guaranteed outcomes.

Timeline

• 2009 - Overwolf founded in Tel Aviv.
• 2020 - Overwolf acquires CurseForge from Twitch, bringing the app and its mod-browsing community into the Overwolf ad monetization platform.
• March 9, 2025 - Jounce Media's Supply Path Benchmarking Report confirms Magnite leads with 99% CTV supply coverage, establishing the series' scope as a key industry reference.
• June 2024 - DoubleVerify research documents a 19% year-over-year increase in MFA impression volume and finds MFA sites often show artificially high viewability metrics - a parallel dynamic to the desktop app issue Jounce later identifies.
• July 15, 2025 - IAB Europe discussion surfaces limitations of viewability as a standalone quality proxy, with panelists pointing to the need for multi-metric quality frameworks.
• August 7, 2025 - Zillow pilots containerized RTB with Index Exchange in a test specifically designed to access site-level signals - including refresh rates - that standard bid requests do not carry.
• January 14, 2026 - IAS launches automated supply path optimization with Total Visibility, addressing the structural opacity of programmatic supply chains.
• April 10, 2026 - DoubleVerify exposes the AutoBait MFA network, showing how high viewability metrics can mask ad fraud at scale.
• April 13, 2026 - IAS opens beta for low-quality AI content blocking, extending quality controls to a category of inventory where automated metrics had been unreliable.
• April 15 - May 14, 2026 - Jounce Media establishes the baseline period for measuring bid request volumes across desktop apps that engage in background refreshing.
• Early May 2026 - Jounce Media's evaluations identify active background refresh behavior in Blitz, Wand, Feather, Modrinth, and Bluestacks, confirming that ads were serving with zero pixels visible to users.
• May 2026 (over four weeks) - All five developers contacted by Jounce Media disable background refreshing following disclosure. Bluestacks records a 77% drop in bid request volume. Feather and Modrinth each record drops exceeding 70%.
• Late May 2026 - All major desktop app developers have implemented app-specific user agent strings, enabling environment disambiguation in the bidstream.
• June 1, 2026 - Jounce Media publishes its June 2026 Supply Path Benchmarking Report, formally classifying desktop apps as a distinct environment and disclosing the background refresh findings.

Summary

Who: Jounce Media, a digital marketing consultancy that operates a data co-op pooling programmatic media metrics from global buyers, published the June 2026 edition of its monthly Supply Path Benchmarking Report. The report focuses on Overwolf, the operator of CurseForge and more than 200 other desktop apps, alongside independent desktop app developers including the publishers of Blitz, Wand, Feather, Modrinth, and Bluestacks. DSP buyers and their brand and agency clients are directly affected as holders of budget that has been allocated to this inventory category.

What: Desktop apps - Windows and Mac applications built on the Electron framework that render ads via Chromium - account for 11.2% of web bid requests, making the category a larger programmatic supply source than any single web publisher. Five apps within the category were discovered to have been refreshing ads in the background - when occluded, minimized, or closed - generating impressions with zero pixels visible to users. Standard viewability vendors reported inflated scores for these apps, in some cases above 97%, because the Electron backgroundThrottling setting disabled the Page Visibility API that viewability measurement relies on. After Jounce contacted the developers, all five disabled background refreshing. Jounce Media has also added a new Desktop App environment classification to its Property Scores file and dashboards.

When: The report was published on June 1, 2026. The background refresh behavior was identified in early May 2026. The developers disabled the behavior over the four weeks before publication. The bidstream data underlying the supply landscape findings was collected in May 2026.

Where: The desktop app category operates across Windows and Mac computers globally. Bid requests are presented to DSPs via web objects with domains that function as aliases for installed applications. CurseForge, the largest single app in the category by bid volume, is accessible through a domain within the overwolf.com namespace.

Why: The classification gap between web and app environments in the OpenRTB standard - which requires either a site object or an app object in each bid request - leaves desktop apps without a clear technical home. That ambiguity has persisted for years without a formal industry response, allowing background refresh behavior to go undetected. Jounce Media surfaced the issue, contacted affected developers directly, and is publishing the findings to prompt buyers, SSPs, and MRC-accredited viewability vendors to establish clearer standards for an environment that now represents more than one in nine web bid requests.