Research Guides
Define Typical Before Summarizing Article Traffic
“The typical article receives 500 visits” sounds precise, but it leaves an important question unanswered: what does typical mean? A few unusually popular pages can make a portfolio's arithmetic average look very different from the experience of its middle-ranked page.
This worksheet helps an editor describe a collection of website measurements without hiding its shape. It is a manual analysis method, not a claim about automatic statistical reporting in Colquhoun.shop.
1. Define what one observation represents
Before calculating, state the unit: one article's visits over the same reporting window, for example. Record the source, metric definition, included pages and exclusions. Do not mix an article's lifetime total with another article's weekly total and call the result a comparable collection.
Preserve the original rows and keep missing measurements distinct from measured zeros. Write the inclusion rule before choosing the number that will appear in the headline.
2. Name the summary measure
NIST defines the mean as the sum of observations divided by their count and the median through the ordered middle observation or pair. Its guidance explains that extreme tail values can strongly affect the mean, whereas the median depends on ranks. Different measures answer different questions about a distribution. NIST: Measures of Location
Do not label an unspecified calculation “average” in one report and silently switch its method in the next. Name the method in the table heading or nearby note so comparisons remain interpretable.
3. Show a small example before writing the conclusion
Suppose five fictional articles receive 80, 90, 100, 110 and 2,120 visits during an identical window. Their total is 2,500, their arithmetic mean is 500 and their median is 100. All three values are valid descriptions of this invented dataset, but they describe different things.
The total answers how many visits the collection received. The mean spreads that total evenly across five articles. The median identifies the middle observation after ordering the five values. Saying only “a typical article got 500 visits” conceals how concentrated the total is.
A more informative note is: “These five articles received 2,500 visits in total; the mean was 500 and the median was 100, with one article accounting for 2,120.” This does not claim the next article will perform like either summary.
4. Investigate extremes without automatically removing them
Check the unusually large row against the original report. Confirm that it is the intended page, period and metric rather than a duplicated record or a collection total accidentally included as a page.
If the value is genuine and belongs in the defined collection, do not delete it merely to make the numbers look more consistent. Record the concentration. If it is an error, correct it with an audit trail and recalculate every affected summary.
If stakeholders need separate groups, define those groups for a reason they can understand, such as article age or content type. Do not invent categories after seeing the result solely to obtain a preferred headline.
5. Publish the calculation context
- Collection: pages, window and selection rule.
- Measurement: source, metric and missing-data treatment.
- Summaries: count, total, named mean and named median where useful.
- Shape: concentration or spread that changes the interpretation.
- Checks: extreme rows investigated and corrections made.
- Limits: what this collection cannot establish about future results.
Keep the worksheet and calculation available to the authorized reviewer. Reproduce the numbers from the saved rows before approving the final wording.
The objective is not to declare the median universally better or the mean misleading. It is to choose and label the summary that fits the question, while leaving enough context for a reader to see what one number leaves out.