Sentry: light pages, early contact
Sentry is error monitoring. A site embeds its SDK so that when a script throws an exception in a real visitor's browser, the stack trace, the browser version and the sequence of actions that led to it are sent to Sentry's servers, where a developer can see it. The whole point is to catch failures the developer never reproduced, which means the SDK has to be loaded and listening before anything else can break. This index found Sentry on 219 of the 5,294 sites measured, 4.1% of the corpus and 16th of the 88 companies broken out here by reach. The pages it sits on are among the quieter ones in this index, and they start early anyway.
Six companies on the median page — rank 76 of 88
The median site carrying Sentry contacted 6 separate companies over 42 requests, 1 of them recognised as a tracker by this index's vendor catalogue. Against the 88 companies broken out here, 6 companies ranks 76th, where rank 1 is the highest. Only twelve of the 88 rank below it on that measure. Its median tracker count ranks 62nd and its request count 48th.
That is not what a company's reach usually predicts. Sentry is 16th of 88 by number of sites, so it is widely deployed, but the pages deploying it are engineering-led properties rather than advertising-led ones. Six companies is roughly the footprint of a page that runs a CDN, a font host, an analytics tool and a monitor.
96.3% of those sites made contact before a banner could be answered
211 of the 219 sites reached someone inside the opening window — the interval that closes where a visitor's answer to a banner first becomes possible. That share ranks 20th of the 88 companies here, with rank 1 the highest — near the top of a set whose median page it sits below on every load measure.
The median first contact landed at 659 ms and the fastest single site in the group at 120 ms. The median number of companies first contacted after that window closed is 0, measured directly rather than inferred from the gap between two medians. Light and early, with nothing following.
What this does not establish is when Sentry itself arrived. Early contact is stored as a number per site, not as a set of names, so no arrival on those 211 pages can be pinned on Sentry. Only the co-presence is measured.
OneTrust on almost a third of Sentry's sites
Consent software was detected on 109 of the 219 sites. OneTrust was seen alongside on 70 of them, an IAB TCF framework banner on 19, Cookiebot on 8, Osano on 4, Usercentrics on 4 and CookieYes on 2. One page can show two of these at once, which is why the counts are sightings and not a division of the 109.
OneTrust also shows up in the company co-presence data, on 69 of the 219 sites (31.5%), which is high for this corpus. OneTrust is the enterprise end of consent management — the tool a company buys when it has a legal department. Its concentration here, alongside the IAB advertising framework's relative scarcity, describes the kind of site that runs Sentry: a product or a platform with a compliance budget, not a publisher selling inventory.
What Sentry travels with
Google was present on 126 of the 219 sites (57.5%). Then OneTrust on 69 (31.5%), Google Tag Manager on 37 (16.9%), Cloudflare Analytics on 25 (11.4%), jsDelivr on 22 (10%), Wistia on 20 (9.1%), Microsoft Ads on 16 (7.3%), Amazon CloudFront on 16 (7.3%), Cloudflare on 15 (6.8%) and the Meta Pixel on 13 (5.9%).
Wistia is the tell. It is a video host aimed at marketing teams, and its appearance this high in a co-presence list points at software companies with product tours and demo reels. The list has no ad exchanges in it at all.
Software companies, games and one clear split
spotify.com contacted 6 companies, all 6 inside the window, first contact at 550 ms behind OneTrust. appsflyer.com: 5 companies, all 5 inside, at 625 ms. onelink.me: 5 with all 5 inside, at 687 ms. unity3d.com reached 7 with 6 inside at 992 ms, amplitude.com 10 with 6 at 854 ms, zendesk.com 8 with 3 at 1,308 ms. tumblr.com is the lightest at 2 companies, both inside the window, at 770 ms behind a TCF banner. canva.com: 3, all inside, at 1,254 ms with no banner detected. roblox.com reached 3 with 1 inside the window at 685 ms.
Three of the twelve show the opposite pattern, with nothing at all arriving early. xiaomi.com contacted 9 companies, none inside the window, first contact at 2,834 ms. mi.com: 10 companies, none early, at 2,374 ms. epicgames.com reached 6 with none early at 2,087 ms, behind OneTrust. A first contact past two seconds means the page did its own work first — this index records the delay, not the reason for it.
| Site | Companies | Before consent | First contact |
|---|---|---|---|
| roblox.com | 3 | 1 | 685 ms |
| spotify.com | 6 | 6 | 550 ms |
| unity3d.com | 7 | 6 | 992 ms |
| xiaomi.com | 9 | 0 | 2834 ms |
| tumblr.com | 2 | 2 | 770 ms |
| epicgames.com | 6 | 0 | 2087 ms |
| canva.com | 3 | 3 | 1254 ms |
| appsflyer.com | 5 | 5 | 625 ms |
| mi.com | 10 | 0 | 2374 ms |
| amplitude.com | 10 | 6 | 854 ms |
| zendesk.com | 8 | 3 | 1308 ms |
| onelink.me | 5 | 5 | 687 ms |
| Tool | Sites |
|---|---|
| OneTrust | 70 |
| IAB TCF consent tool | 19 |
| Cookiebot | 8 |
| Osano | 4 |
| Usercentrics | 4 |
| CookieYes | 2 |
The six most common are listed. The remaining 2 of 109 sites with a banner carry something outside that six. One tool is recorded per site, so these counts never overlap.
Seen on the same pages
Companies most often contacted by the same sites. Co-presence only — the crawl records what arrived, not who called whom.
- GoogleTogether on 126 of these sites
- OneTrustTogether on 69 of these sites
- Google Tag ManagerTogether on 37 of these sites
- Cloudflare AnalyticsTogether on 25 of these sites
- jsDelivrTogether on 22 of these sites
- Microsoft AdsTogether on 16 of these sites
- Amazon CloudFrontTogether on 16 of these sites
- CloudflareTogether on 15 of these sites
Questions
What does Sentry do?
It collects crashes and errors from real browsers and servers so developers can find and fix them. This index classifies it as error monitoring, and not as tracking.
Why does an error monitor need to load early?
It cannot report a failure it was not present for. An SDK that loads late misses everything that broke before it. This index measures when a page first contacts any third party, not the loading order of individual scripts, so it cannot show where Sentry sat in that sequence.
Did Sentry fire before consent on these sites?
The corpus counts how many companies were reached early on each site but never records which ones, so that cannot be answered from this data. What it supports is that Sentry was present on 211 sites where some company was contacted inside the window.
Are Sentry's pages heavy?
No. Six companies at the median ranks 76th of the 88 companies broken out here, where rank 1 is the highest, and 42 requests ranks 48th. It is one of the lighter groups in this index.
How many sites is the Sentry figure based on?
219 of the 5,294 sites measured, 4.1% of the corpus. That is 16th of the 88 companies broken out here by reach, while the pages themselves rank 76th by companies contacted.