Text & web utilities

Wikipedia daily pageview trend analyzer

Plot daily Wikimedia pageviews for 30 or 60 complete UTC dates with explicit access and agent filters.

Your settings

Demonstration records are loaded initially. Fetching connects directly to Wikimedia.

Live data requests go directly to Wikimedia. No request is made until you choose Fetch.

Your result

Ready when you are

Enter your inputs and select Analyze loaded data.

Report an issue

Explore the Wikipedia analysis suite →

Try it, understand it

Explore the result

These are explicitly labeled demonstration records. Choose Fetch on the form to replace them with current Wikimedia data.

Worked example comparison
ExampleOutputWhat to notice
Three sample daily countsViews across returned daily records600100 plus 200 plus 300 totals 600; the requested window remains partial.
A sixty-day demonstration windowViews across returned daily records600The three sample records remain explicitly partial, not a sixty-day live series.
Different sample filter metadataViews across returned daily records90The separate synthetic user-classified series totals ninety.

Check your understanding

Are pageviews unique readers?

No. They are request counts under the selected access and agent filters.

Are missing days filled as zero?

No. Missing records remain missing and the response is marked partial.

Behind the tool

How this tool works

Request thirty or sixty complete UTC calendar dates ending yesterday, using the legacy per-article endpoint’s inclusive date boundaries. The selected access and agent filters remain part of the result. Returned points retain their actual date spacing; missing dates are not filled with invented zeros. The table shows daily values and the chart joins returned points for visual comparison without predicting future traffic.

THE OUTPUT RULEReturned-view total = sum of returned daily counts; average = total / number of returned dates

A worked example

The demonstration has three returned daily counts: 100, 200 and 300. Their total is 600 and their mean over those returned dates is 200. It is explicitly partial within a thirty-day requested window. Last-versus-first change is 200%, which compares two endpoints rather than measuring a fitted trend.

Good to know

Pageviews under a stated request filter are not unique readers, edit counts, article quality or a forecast. All-agents can include automated traffic. Data publication can lag, and redirects or renamed articles can have separate historical series. A missing or partial response does not establish zero demand.

A pageview count is not a count of unique people

One reader can generate several requests, and the agent filter changes which classes of requests are included. All-agents and user-classified traffic therefore answer different questions. The chosen filter is reported with the daily table rather than being hidden behind a generic popularity number.

The Wikipedia internal link grapher counts a bounded set of outgoing links, not attention. It cannot establish readership from the number or placement of graph nodes. Likewise, a pageview spike does not by itself establish article accuracy, importance or the reason people arrived.

Use complete UTC dates and retain missing-data information

The request ends yesterday so the current incomplete UTC day does not masquerade as a full day’s traffic. The legacy endpoint includes both date boundaries, making a thirty-date window start twenty-nine days before its final date. Requesting an extra endpoint would produce an off-by-one comparison.

If a date is absent from the response, the tool does not invent a zero. The displayed mean uses returned days and the result is marked partial. Endpoint percentage change also depends strongly on the chosen first day; it is not a fitted trend or a forecast. The Wikipedia category explorer supplies a separate organizational view that should not be confused with a time series of attention.

A few questions, answered

Are pageviews unique readers?

No. They are request counts under the selected access and agent filters.

Are missing days filled as zero?

No. Missing records remain missing and the response is marked partial.

Find your calculator

Start typing to explore the library.