Downdetector is a website and app that detects outages in online services by counting complaints. People who cannot load a bank app, a streaming service or an email inbox report the problem, and Downdetector compares the volume of those reports with what is normal for that service at that hour. When the count climbs well above the norm, the site flags an incident and draws a chart that journalists, operators and the public read as evidence of a failure. It exists because companies are slow, and sometimes reluctant, to confirm that their own systems are broken.

The service was founded in April 2012 by Tom Sanders and Sander van de Graaf, according to Wikipedia, and was bought by Ookla, the company behind Speedtest.net, in August 2018. Ookla itself changed hands in 2026. Accenture announced on March 3, 2026 that it would acquire Ookla from Ziff Davis, and the deal closed on June 17, 2026, according to a note on Accenture's newsroom.

How detection works

The raw material is the problem report. A visitor to a service's page on Downdetector, such as the page for Outlook or Instagram, clicks to say something is wrong and usually picks a category. During the Microsoft disruption of November 12, 2026, 33% of reporting users cited the app, 32% the website and 18% trouble receiving messages, according to WJXT. Those splits help separate a broken login server from a broken mobile release.

Direct reports are not the only input. Wikipedia describes reports "collected from various sources, including the page for each website on Downdetector itself and Twitter." A December 2025 press release from BigPanda, an incident-management software company, lists "social media and user reports (via Downdetector)" among the signals it ingests. Reuters, in its outage copy, describes Downdetector as tracking outages "by collating status reports from a number of sources."

The second element is the baseline. Some services draw complaints all day; a global email platform with hundreds of millions of users will always have someone reporting a fault. Downdetector's website states that it only reports an incident when the number of reports is "significantly higher than the normal volume for that time of the day," according to The Hill. Comparing against the same hour matters, because complaint volume follows daily patterns of use. A threshold calibrated on 3 AM traffic would fire constantly at 9 AM.

Ookla said in December 2024 that the platform drew on "over 25 million monthly reports from individual users, real-time analysis, and verification of outage reports," according to a press release on its 2024 outage review. Wikipedia puts coverage at more than 12,000 services across 45 country sites. Neither figure has been independently audited, and the thresholds themselves are not published.

Who uses it and how it is sold

Downdetector has two audiences. The consumer site is free and carries advertising. A Pro plan went live in Australia in September 2026 at A$13.99, removing adverts and adding push alerts for up to ten favourite services, according to ISPreview, with other markets "coming soon."

The second audience pays more. An enterprise tier, priced privately, offers API access, up to five years of outage history, monitoring of up to 100 services and custom alerts, according to the same report. Telecoms operators, cloud providers and large consumer brands buy it to learn about a fault from their customers before their own dashboards turn red, or to watch competitors and suppliers. The data also flows into third-party tools: BigPanda announced an integration with Downdetector on December 16, 2025, available to enterprises holding a Downdetector licence.

Accenture's interest sits on this side of the business. Manish Sharma, Accenture's chief strategy and services officer, said in the acquisition announcement that "Downdetector identifies incidents faster," pairing it with Speedtest, Ekahau and RootMetrics as network data products. The connectivity division generated $231 million in revenue in 2025, around 16% of Ziff Davis's sales, according to The Register. Accenture paid $1.2 billion for the package; Downdetector's share was not disclosed.

Origin and evolution

Downdetector began as a small consumer site in 2012 and expanded country by country, with a separate site for each market. The Ookla acquisition in 2018 placed it alongside Speedtest, which measures connection speeds through user-initiated tests, more than 250 million a month according to Accenture. The two products are complementary: Speedtest shows what a network delivers, Downdetector shows what users say is broken.

Under Ookla the service moved from a consumer curiosity to an industry reference. Its annual and half-year reviews ranked the largest outages by report volume. For the first nine months of 2024, the Facebook outage of March 5 led with 11.1 million reports and the CrowdStrike failure of July 19 followed with nearly 5 million, according to a summary published by the Internet Society. Those rankings travel widely because no other public source measures user-perceived failure across thousands of services on one scale.

The ownership chain then shifted twice in nine months: a Ziff Davis division until mid-2026, and part of Accenture, a consulting firm, from June 17, 2026. The paid consumer tier followed three months later.

Why it matters to marketers

Outages now hit advertising directly. During the AWS failure of October 19 and 20, 2025, a DNS problem in the US-EAST-1 region affected 142 AWS services for about 15 hours, and PPC Land reported that Downdetector received more than 6.5 million reports by The Guardian's count and more than 8 million by Morning Brew's. Ad platforms, measurement tags and clean rooms built on that infrastructure failed alongside consumer apps. Downdetector's chart was, for many teams, the first public evidence of how wide the damage ran.

For media buyers, a spike on a platform's Downdetector page is often the earliest outside signal that delivery or reporting anomalies are not self-inflicted. For brand and communications teams, the same page is a public record of a company's failure, cited by Reuters and local broadcasters within minutes. Downdetector is also a publisher in its own right, selling advertising against pages whose traffic surges when something breaks.

The limits matter for ad tech in particular. Advertising failures are often invisible to consumers. When Google Ad Manager and AdSense faltered from January 13, 2026, publishers reported revenue declines of 50% to 90% while pages and ads still appeared to load for readers. Nobody files a Downdetector report because a match rate fell.

Limitations and disputes

Reports are not users. Headlines routinely turn "11.1 million reports" into "11.1 million people affected." The counts measure people motivated to report, which skews towards consumer-facing services, mobile users, and markets where Downdetector is well known. Business-to-business tools and back-end systems are underrepresented.

Attribution. Users report the service in front of them, not the component that broke. The October 2025 AWS failure surfaced as complaints about Snapchat, Fortnite and banks, and the CrowdStrike count of July 2024 was spread across services that relied on it or on Microsoft, according to the Internet Society summary. Downdetector shows where users perceive failure, not where the fault lies.

Inconsistent figures. On November 12, 2026, Reuters cited more than 97,000 Outlook reports at 2:07 PM ET, while WJXT cited nearly 119,000 outage reports by 1 PM. Neither explained the difference, which may reflect cumulative versus peak counts or different service pages.

Speed is contested. Accenture markets faster detection, yet a crowdsourced system cannot see a fault until users notice. On November 12, Microsoft said its Exchange Online incident began at 16:17 UTC; reports about Outlook began climbing at about 17:25 UTC, according to Downdetector's own post on X, a lag of about 68 minutes.

Single point of failure. During the Cloudflare outage of November 18, 2025, Downdetector itself was temporarily knocked offline, according to Futurism. Its dependence on the same shared infrastructure it monitors prompted a run of parody "Downdetector's Downdetector" sites, according to PC Gamer.

Neutrality under new ownership. Accenture sells services to telecoms operators and enterprises whose outages Downdetector publicises. No editorial or methodology change has been announced, but the commercial overlap is new.

Not the same as

Official status pages, such as Google's Search Status Dashboard, are written by the provider. They confirm incidents the company acknowledges; Downdetector records what users report, whether or not the company agrees.

Status aggregators such as IsDown and StatusGator collect and republish official status pages, sometimes adding their own user reports. Their primary input is what vendors publish.

Network telemetry tools such as Cloudflare Radar measure traffic directly from infrastructure, which makes them better for detecting country-level shutdowns and routing failures but blind to an app whose servers respond with errors.

Speedtest, Downdetector's sister product, measures connection performance through tests users run. It answers "how fast," not "is it broken."

Recent developments

The most recent test came yesterday. Microsoft's Exchange Online disruption on November 12, 2026 spread to reports from 74 countries and territories within about an hour, according to Downdetector, and affected connectors at OpenAI and Anthropic that rely on Microsoft 365, according to Reuters. The episode showed both the platform's reach and its lag behind provider-side detection.

It also followed a year of change: a new owner, a paid consumer tier and deeper integration into enterprise monitoring tools. Whether Accenture keeps the public site as a free reference, or treats it mainly as a data feed for paying clients, has not been stated.


Timeline

  • April 2012 - Downdetector founded by Tom Sanders and Sander van de Graaf.
  • August 2018 - Ookla, owner of Speedtest.net, acquires Downdetector.
  • September 8, 2022 - Downdetector logs more than 2,600 Twitter reports around 1:40 PM ET after the death of Queen Elizabeth II.
  • March 5, 2024 - Facebook outage generates 11.1 million reports, the largest of 2024's first nine months.
  • July 19, 2024 - CrowdStrike failure generates nearly 5 million reports.
  • December 2024 - Ookla cites more than 25 million monthly reports in its 2024 outage review.
  • October 19-20, 2025 - AWS US-EAST-1 outage draws between 6.5 million and 8 million reports, depending on the source.
  • November 18, 2025 - Downdetector itself goes offline during the Cloudflare outage.
  • December 16, 2025 - BigPanda announces a Downdetector integration.
  • March 3, 2026 - Accenture announces the $1.2 billion acquisition of Ookla from Ziff Davis.
  • June 17, 2026 - Acquisition closes.
  • September 2026 - Downdetector Pro launches in Australia at A$13.99.
  • November 12, 2026 - Microsoft Exchange Online outage generates reports from 74 countries and territories within about an hour.

Summary

Who: Founded by Tom Sanders and Sander van de Graaf, owned by Ookla from 2018 and by Accenture since June 17, 2026. Users submit reports; journalists, operators, brands and enterprise clients read them.

What: A crowdsourced outage tracker that flags an incident when problem reports for a service rise significantly above the normal volume for that time of day, covering more than 12,000 services.

When: Launched in April 2012, ranked major outages through the 2020s, and sold to Accenture in a $1.2 billion deal announced March 3, 2026. Its latest high-profile reading came with Microsoft's November 12, 2026 outage.

Where: On country-specific websites in 45 markets and in mobile apps, with data sold through an enterprise API and partner tools such as BigPanda.

Why: It gives an independent, near-real-time public view of service failures that providers may confirm late or not at all. Its counts measure reports rather than affected users, lag behind the onset of faults, and miss failures that consumers cannot see, including most ad tech breakdowns.