Private longitudinal analysis

Internet speed test history and performance tracker

Monitor internet speed over time without an account. Turn repeated local tests into transparent charts, medians, spread, time-of-day summaries, and guarded trend signals.

Up to 100 local resultsMedian and p10-p90No background testingPortable JSON and CSV

Private local history

Your performance timeline starts with one test

Completed tests are stored only in this browser. Build a timeline across days, then return here for medians, spread, time-of-day patterns, and guarded recent-versus-earlier trends.

Run a speed testImport history JSON

No account, cloud upload, or automatic background testing. The tracker can analyze only completed tests saved on this device.

Published methodology 1.0

How the history analysis works

Only locally measured runs with valid core metrics and full recorded download and upload transfer confidence enter the aggregate. Shared or degraded-confidence results remain visible as exclusions rather than silently changing the trend.

1-2 eligible runs

Snapshot history

One or two eligible runs can orient the user but cannot describe a repeatable pattern.

3+ eligible runs

Repeated history

At least three eligible runs support descriptive spread and trend summaries, but the stronger multi-session checks are not all satisfied.

6+ eligible runs

Stronger multi-session history

At least six eligible runs span two or more local calendar dates and at least six hours.

Median, p10, p90, and consistency

Each core metric uses a median plus linearly interpolated p10 and p90 values. Download spread is (p90 - p10) / median, expressed as a percentage. The bands below are descriptive, not confidence intervals.

Steady sample: The download p10-to-p90 spread is no more than 10% of the download median.
Variable sample: The download p10-to-p90 spread is more than 10% and no more than 25% of the download median.
Highly variable sample: The download p10-to-p90 spread is more than 25% of the download median.

Recent versus earlier medians

With six or more eligible runs, the chronological sample is split into an older half and a recent half. Median download, upload, idle latency, and idle jitter must cross both the published absolute and relative comparison guards before they are labeled improved or regressed.

A trend is an association within these saved browser measurements. It does not prove a cause, statistical significance, an outage, or access-line performance.

Inspect the comparison guards

Local time-of-day windows

The tracker groups timestamps using the browser's current UTC offset. Travel and daylight-saving changes can affect how older timestamps are grouped.

Overnight

00:00-06:00

Morning

06:00-12:00

Afternoon

12:00-18:00

Evening

18:00-24:00

A stronger multi-session history requires at least 6 eligible runs across 2 or more local dates and at least 6 elapsed hours.

Controlled collection protocol

Build a more comparable history

1
Keep the test path stable
Use the same capable device, connection type, provider label, and endpoint when practical.
2
Reduce unrelated load
Pause unrelated transfers and record any condition that could change the browser path.
3
Test at planned times
Collect across multiple dates and time windows instead of testing only when the connection feels slow.
4
Read cautions with the trend
Use eligible counts, exclusions, endpoint and condition changes, medians, and spread together.

Primary methodology sources

Federal Communications Commission

Eleventh Measuring Broadband America Fixed Broadband Report

The program uses repeated observations and aggregate reporting for download, upload, latency, and loss measurements.

Open source

Internet Engineering Task Force

RFC 2330: Framework for IP Performance Metrics

Defines repeatability, methodology, and careful interpretation as core properties of useful internet performance measurements.

Open source

SwiftSpeedTest

Speed test comparison rules 2026.07.24

Supplies the published absolute and relative guards used for recent-versus-earlier median trend classification.

Open source

Frequently asked questions

Internet speed history questions

How can I track my internet speed over time?: Run completed SwiftSpeedTest measurements on the same browser and return to this page. The tracker can analyze up to 100 locally stored results with medians, p10-p90 spread, time-of-day buckets, and a guarded comparison between the earlier and more recent halves of the selected sample.
Does this internet speed tracker run in the background?: No. Browser and device restrictions mean it cannot test while the page is closed, the tab is suspended, or the device is asleep. It analyzes completed user-initiated tests rather than claiming continuous outage monitoring.
Where is my speed test history stored?: History stays in local browser storage on the device and browser that ran the test. SwiftSpeedTest does not require an account or upload the history to a cloud profile. You can export JSON or CSV and import a valid JSON export on another browser.
What makes a result eligible for the trend analysis?: The analysis includes locally measured results with a valid timestamp, finite non-negative core metrics, and full recorded download and upload transfer confidence. Editable shared results and incomplete-confidence transfers are disclosed and excluded.
Why use medians and p10-p90 spread?: The median reduces the influence of one unusually high or low observation. Linearly interpolated p10 and p90 values summarize the central 80 percent of the recorded sample. They describe these browser results and are not statistical confidence intervals.
How many speed tests do I need for a useful history?: One or two eligible runs are a snapshot. Three or more support a repeated descriptive sample. The stronger multi-session label requires at least six eligible runs across at least two local calendar dates and a six-hour elapsed span.
Can the history prove my ISP caused a slowdown or outage?: No. A browser result includes the device, browser, local network, route, and selected endpoint. Trend labels describe an association within the saved sample; they do not establish causation, statistical significance, an outage, or access-line performance.
How should I collect comparable speed test history?: Use the same capable device, connection type, provider label, and endpoint when practical. Pause unrelated traffic, record conditions, and test at planned times across multiple dates instead of testing only when the connection feels slow.

Your performance timeline starts with one test

Completed tests are stored only in this browser. Build a timeline across days, then return here for medians, spread, time-of-day patterns, and guarded recent-versus-earlier trends.

No account, cloud upload, or automatic background testing. The tracker can analyze only completed tests saved on this device.

How the history analysis works

1-2 eligible runs

Snapshot history

One or two eligible runs can orient the user but cannot describe a repeatable pattern.

3+ eligible runs

Repeated history

At least three eligible runs support descriptive spread and trend summaries, but the stronger multi-session checks are not all satisfied.

6+ eligible runs

Stronger multi-session history

At least six eligible runs span two or more local calendar dates and at least six hours.

Median, p10, p90, and consistency

Steady sample: The download p10-to-p90 spread is no more than 10% of the download median.
Variable sample: The download p10-to-p90 spread is more than 10% and no more than 25% of the download median.
Highly variable sample: The download p10-to-p90 spread is more than 25% of the download median.

Recent versus earlier medians

A trend is an association within these saved browser measurements. It does not prove a cause, statistical significance, an outage, or access-line performance.

Inspect the comparison guards

Local time-of-day windows

The tracker groups timestamps using the browser's current UTC offset. Travel and daylight-saving changes can affect how older timestamps are grouped.

Overnight

00:00-06:00

Morning

06:00-12:00

Afternoon

12:00-18:00

Evening

18:00-24:00

A stronger multi-session history requires at least 6 eligible runs across 2 or more local dates and at least 6 elapsed hours.

Build a more comparable history

Keep the test path stable

Use the same capable device, connection type, provider label, and endpoint when practical.

Reduce unrelated load

Pause unrelated transfers and record any condition that could change the browser path.

Test at planned times

Collect across multiple dates and time windows instead of testing only when the connection feels slow.

Read cautions with the trend

Use eligible counts, exclusions, endpoint and condition changes, medians, and spread together.

Primary methodology sources

Federal Communications Commission

Eleventh Measuring Broadband America Fixed Broadband Report

The program uses repeated observations and aggregate reporting for download, upload, latency, and loss measurements.

Open source

Internet Engineering Task Force

RFC 2330: Framework for IP Performance Metrics

Defines repeatability, methodology, and careful interpretation as core properties of useful internet performance measurements.

Open source

SwiftSpeedTest

Speed test comparison rules 2026.07.24

Supplies the published absolute and relative guards used for recent-versus-earlier median trend classification.

Open source

Internet speed history questions

How can I track my internet speed over time?

Run completed SwiftSpeedTest measurements on the same browser and return to this page. The tracker can analyze up to 100 locally stored results with medians, p10-p90 spread, time-of-day buckets, and a guarded comparison between the earlier and more recent halves of the selected sample.

Does this internet speed tracker run in the background?

No. Browser and device restrictions mean it cannot test while the page is closed, the tab is suspended, or the device is asleep. It analyzes completed user-initiated tests rather than claiming continuous outage monitoring.

Where is my speed test history stored?

History stays in local browser storage on the device and browser that ran the test. SwiftSpeedTest does not require an account or upload the history to a cloud profile. You can export JSON or CSV and import a valid JSON export on another browser.

What makes a result eligible for the trend analysis?

The analysis includes locally measured results with a valid timestamp, finite non-negative core metrics, and full recorded download and upload transfer confidence. Editable shared results and incomplete-confidence transfers are disclosed and excluded.

Why use medians and p10-p90 spread?

The median reduces the influence of one unusually high or low observation. Linearly interpolated p10 and p90 values summarize the central 80 percent of the recorded sample. They describe these browser results and are not statistical confidence intervals.

How many speed tests do I need for a useful history?

One or two eligible runs are a snapshot. Three or more support a repeated descriptive sample. The stronger multi-session label requires at least six eligible runs across at least two local calendar dates and a six-hour elapsed span.

Can the history prove my ISP caused a slowdown or outage?

No. A browser result includes the device, browser, local network, route, and selected endpoint. Trend labels describe an association within the saved sample; they do not establish causation, statistical significance, an outage, or access-line performance.

How should I collect comparable speed test history?

Use the same capable device, connection type, provider label, and endpoint when practical. Pause unrelated traffic, record conditions, and test at planned times across multiple dates instead of testing only when the connection feels slow.