Website numbers tell only part of the story

Before interpreting growth, we need to know what is counted, in relation to which set, and which question the figure still leaves unanswered.

Digital anthropology
A counter and three partial windows leave part of the scene outside.
A counter and three partial windows leave part of the scene outside. AI-generated illustration: a conceptual scene, distinct from the example on the page.

The number rises from one hundred to two hundred. The doubling is exact, but the judgement we are tempted to add may be premature. Are these page views, visits, people or actions? Is the observation period the same? What did we want to understand before the chart suggested satisfaction?

Statistics make some traces of a reader’s passage available. They can help locate a problem or formulate a hypothesis. The leap starts when we ask those traces alone to tell us whether a text was understood, a page was useful or the work deserves to continue.

Keeping the denominator fixed

For this page I built an explicitly fictional example. It contains no actual MAVID traffic. The first case imagines one hundred views and fifty sessions; the second has two hundred views and one hundred sessions within the same time interval. The average remains two views per session.

Three cases with fictional numbers

Fictional example, with no audience data or actual traffic.

Three cases with fictional numbers
Fictional caseViewsSessionsViews per session
A100502
B2001002
C200504

In the third case, views are still two hundred but sessions are fifty: the average becomes four. The page-view total matches the second case while the relationship changes. This is not enough to choose a better situation. It reveals why the total alone does not answer a question about how the site is traversed.

Umami’s metric documentation distinguishes events and sessions. Reading a dashboard requires starting with the tool’s actual definitions before assigning a personal meaning to its labels.

The reading a number does not contain

Four pages viewed might indicate a route rich in links, or a difficult search. A single page might answer the question someone arrived with. The same traces leave different stories possible.

A metric becomes useful when we know which question it can help clarify. To understand why content is hard to find, I would examine routes and entry points. If I care about comprehension, a page count cannot replace a question addressed to readers or an appropriate trial.

Absences need caution too. An unrecorded event may result from implementation choices, a blocker or behaviour the tool does not observe. I cannot turn every gap in a chart into a lack of interest. The boundary of measurement comes first; interpretation follows.

Two trays of counters keep the total while changing the grouping.
Two trays of counters keep the total while changing the grouping. AI-generated illustration: a conceptual scene, distinct from the example on the page.

Comparing periods that make sense

An increase may accompany the publication of many new articles or arrivals through an external link. Comparison with a quiet period then describes different situations. The difference remains true, but the reason for looking at it changes.

In a small project, a few added or missing visits can shift a percentage considerably. Saying a change is one hundred per cent preserves the calculation and may conceal its scale. I prefer keeping the starting and ending numbers together, alongside the period and the unit counted.

The sample makes these elements visible and fictional. I have not used automated visits to construct growth, present a demonstration screen as the site’s results, or derive audience forecasts from the example’s chosen numbers. The comparison tests a line of reasoning rather than displaying performance.

The decision that remains mine

The two AI-generated illustrations show partial windows and groups of tokens. They materialise the limits of a total; they are not charts of collected data. The table preserves the exact values from which the calculations start.

A figure may bring me back to a page, prompt a link check or lead me to clarify an entry point. It should not decide in advance that every article must resemble the one with most visits. An archive loses part of its usefulness if it removes everything seldom searched for that nevertheless answers a precise question well.

Counting can guide the work without becoming the criterion that justifies it. The dashboard gives back part of the website. Deciding what to publish and how long to leave it available still requires a reason that does not fit entirely inside that window.

Two trays of counters keep the total while changing the grouping.
Two trays of counters keep the total while changing the grouping. AI-generated illustration: a conceptual scene, distinct from the example on the page.