Here’s What “Up to 1000Mbps” Actually Means in Seattle
A Data-Oriented Interpretation.
If You’re Like Me…
You probably use the internet for work and play, such as downloading large files containing assets to use in a project, or uploading a new project to be shared with the world. Like me, you’ve had issues with your ISP. And like me, you’ve considered paying your ISP more with the expectation that they will make everything go faster. But exactly how much faster does your service get per dollar spent? Are we getting more “bang for our buck” the more bucks we spend? If we look at their pricing options, it would appear that way. Who wouldn’t pay an extra $5 to get an extra 600Mbps? But, with all of the attention ISPs have gotten over the years, we now know better than to just take their word for it. So let’s use some cold hard data to find the truth.
Here’s What I Want to Know
Knowing if an ISP is giving us the speeds we deserve is simple. We just need to answer these 3 questions:
- What and how much are people paying for?
- What do these ISPs say they’re offering?
- What do we get for what we pay for?
Here’s What We Know
The city of Seattle has made their broadband test data public, which is what we’ll be using to help us answer our questions.
This is what the data looks like:
- 4068 unique reports
- 3119 reports with a usable amount of data
- Reports range from as early as Feb 2016 to as recent as Jan 2018
Available information on each report:
- actual_download: Recorded download speed at time of test
- actual_upload: Recorded upload speed at time of test
- advertised_download: User reported download speed they are paying for
- advertised_upload: User reported upload speed they are paying for
- connection_type: Type of connection to internet (wired/wireless and single user/multiple users)
- cost_of_service: User reported cost of ISP service
- date_pretty: Formatted date and time of test
- isp: Name of ISP company
- isp_user: Name of user
- min_rtt: latency from user to test server
- timestamp: Time of the test, in Unix timestamp format
- seattle_blkgrpce10: Not sure what this is, will not be used in the study.
The data contained many missing values and had erroneous self-reported values (someone was reported paying less than $25 for 1GBP/s speeds, which isn’t offered by any ISP). Additionally, the types of packages people are subscribed to (internet only vs bundles) are not reported, so for the sake of simplicity I’m going to assume everyone buys internet-only packages. In order to keep the data for advertised_download as consistent as possible, I found actual prices for different tiers of service and substituted the values in based on the self-reported category cost_of_service. Here is the list of prices I found:

The Most Popular ISP
The most popular provider is Comcast with 1967 different users in the study. In second place we have CenturyLink with 920 different users and ‘AS11404 vanoppen.biz LLC’ (A.K.A Wave) as the runner up with 232 different users. Since these three companies comprise of over 75% of all data points, I’m going to limit our study to just these three ISPs.
So here’s how much people are paying, by price range:
Here are the speeds we get for each price range:
Now let’s take a look at the average speeds by ISP:
While searching for internet pricing options, I noticed that ISPs did not advertise upload speeds, so all values for advertised_upload should be taken with a grain of salt.
It’s abundantly clear that the speeds we’re actually getting are nowhere near the advertised limits. Ironically, the fastest overall speeds on average (download + upload) are with Wave, which isn’t even offered in the city of Seattle anymore.
What if…
What if internet rush hour is causing reports to be slower?
Internet rush hour is the time period in which the majority of internet users are all online at the same time. And according to Wikipedia, during this time frame, users commonly experience slowness while browsing or downloading content. In order to control for this, let’s check what the most frequent times are for testing speeds.
Now lets check actual values and averages to determine if we’re seeing a significant enough decrease in service due to internet rush hour.
There does seem to be a bit of a decrease in speeds during peak hours but overall it doesn’t appear to be very significant. Most notably, the premium users (100+) don’t seem to be affected at all by internet rush hour.
What if people are just not sitting close enough to their wireless routers?
It’s very possible that users are not testing their speeds under the best circumstances. To control for this, let’s use the connection_type category to compare speeds of each category. Let’s find the actual speeds by connection type, average speeds by connection type, and then compare that to the speeds as advertised by the ISPs:
“Up to 1000Mbps” is just another way of saying 125Mbps, assuming you’re paying for the highest tier of service.
Our ISPs might as well just chop the last digit off each of the advertised speeds because that would be much closer to reality than what they currently have listed. Is only getting 10% of the advertised maximum speeds on average not false advertising?
What if people only speed test when their internet is slow?

My code/thought process/methodology.
If you have any feedback or critiques, please feel free to share them with me. If you found this informative, or just appreciate a good meme in an otherwise dry subject, mash that 👏 button. 🙏 Thanks y’all.
