Downdetector guide
How Does Downdetector Work? A Plain-English Guide
How does Downdetector work? It counts user reports. People report problems on downdetector.com and in the Speedtest by Ookla app, and Downdetector adds public posts. When reports jump far above the normal level for that time of day, it flags a likely outage.
Updated
How does Downdetector know when something is down?
It does not test servers. It listens to people, in four steps.
- Step 1
People report
Users tap "I have a problem" on the site or app.
- Step 2
Posts are scanned
Public posts saying a service is down are counted too.
- Step 3
Compared with normal
The count is checked against the usual level for that hour.
- Step 4
A jump is flagged
Far above normal means "user reports indicate problems."
This is crowdsourcing. Many small signals from many people add up to one clear picture.
Where does Downdetector get its data?
| Source | What it adds |
|---|---|
| User reports on the website | The main signal, one click per report |
| Reports from the app | More reports from phones |
| Social media posts | Public posts that say a service is down |
| Real-time checks | Filters that remove fake or repeat reports |
Crowdsourced outage reports are fast because people notice problems right away. Companies often take longer to confirm them.
Does Downdetector use Twitter reports?
Yes. Public posts on X, formerly Twitter, are one source. But posts alone do not make an outage. Downdetector combines them with direct reports, so a viral joke does not set off an alarm.
What is the baseline on Downdetector?
The Downdetector baseline is the normal number of reports for a service at a given hour. On the chart, it is the shaded area at the bottom.
It rises and falls through the day. Netflix gets more reports at 9 PM than at 4 AM, because more people are watching. Downdetector builds each baseline from past reports, so a busy evening does not look like an outage.
How does Downdetector calculate outages?
It compares reports right now with the baseline. Downdetector does not publish its exact formula, but it only flags an incident when reports are well above the typical level.
- Near the baseline No problem
- Somewhat above A possible minor issue
- Far above "User reports indicate problems"
How many reports does Downdetector need to show an outage?
There is no fixed number. YouTube may need thousands of reports to stand out. A small local bank may need a few dozen. What matters is the rise above that service's own normal. Our guide on reading Downdetector charts shows how to judge it.
User reports: strengths and limits
Strengths
- Very fast, often before the company says anything
- Catches partial problems, like logins only
- Covers thousands of services, for free
Limits
- Not confirmed by the company
- News and viral posts can cause spikes
- People sometimes report the wrong service
How often is it right? Read is Downdetector accurate? and our guide to false alarms.
Who runs Downdetector?
- 2012
Founded in the Netherlands
Downdetector starts as a small site that counts user reports.
- 2018
Bought by Ookla
The company behind Speedtest adds Downdetector to its tools.
- 2026
Ookla joins Accenture
Accenture buys Ookla, and Downdetector with it.
Read the full story in who owns Downdetector.
A worked example: one evening, one outage
Here is how the numbers might play out for a big streaming service on a weekday night. The figures are made up to show the idea.
| Time | Reports | Normal for that hour | What Downdetector shows |
|---|---|---|---|
| 8:00 PM | 40 | 35 | No problems |
| 8:15 PM | 90 | 36 | Rising |
| 8:30 PM | 2,400 | 37 | Problems |
| 10:00 PM | 120 | 38 | Recovering |
The jump at 8:30 is dozens of times the normal level, so it is flagged. The same 2,400 reports at a busier hour would still stand out, because the baseline moves with the clock.
Why user reports beat server checks for spotting outages
A simple website checker can only ask "does the server answer?" Many real outages pass that test. The homepage loads, but logins fail, payments hang, or videos will not play. Crowdsourced outage reports catch these, because people report what they actually see.
User reports also cover things a checker cannot reach at all, like phone signal, smart TV apps, and game servers. The tradeoff is speed of confirmation: a human report says "something is wrong," not exactly what.
Problem types: the second signal
When you report, Downdetector asks what is wrong, like login, app, or no signal. Those answers become the "most reported problems" breakdown. It turns a vague spike into a clear picture, such as "mostly login problems," which tells you whether you can still use the service if you are already signed in.
What Downdetector does not do
- It does not get data from the companies for its public pages.
- It does not confirm the cause. A spike means many users are affected, not why.
- It does not show your address. The map works at city and region level.
- It does not charge services to be listed, and it does not need their permission.
Knowing these limits is the key to reading it well. Our guide to reading the chart puts it into practice.
Put it into practice
Now that you know where the reports come from, see where they cluster on the Downdetector outage map, and add your own reports or get alerts with the Downdetector app. If the site itself will not load, here is what to do when Downdetector is down. And to test one website from your own connection, use our free website down checker.
Frequently asked questions
Does Downdetector check if websites are online?
Not the way a website checker does. Downdetector’s main signal is people saying they have a problem. That is why it can spot problems like broken logins that a simple online check would miss.
Does the company being tracked tell Downdetector about outages?
No. Companies do not feed outage data to Downdetector’s public pages. Downdetector works it out from user reports. Some companies pay for Downdetector’s business tools to watch their own services. See Downdetector Enterprise.
Why does Downdetector show reports when there is no outage?
Every big service has a few unhappy users at any time. Those reports are normal and are part of the baseline. Downdetector only flags a problem when reports are far above that normal level.
Can fake reports trigger an outage on Downdetector?
It is hard. Downdetector says it checks reports in real time to filter out fake ones. A real outage also needs many reports from many places at once, which is tough to fake.
Is Downdetector official?
It is an official product of Ookla, but it is not run by the companies it tracks. When Downdetector says "user reports indicate problems," that is its own reading of user reports, not a statement from the service.
How quickly does Downdetector spot an outage?
Often within a few minutes of the first reports, because people check it as soon as something breaks. That is usually faster than official status pages, which can take 15 to 60 minutes to post.
Does Downdetector track services in every country?
Downdetector runs sites for many countries, each with its own list of local services and its own baselines. This guide focuses on the US site at downdetector.com.
Independent guide. Not affiliated with Downdetector or Ookla.