A page about an obscure cassette might get forty visits in a month and still do more useful work than a post shared widely enough to get four hundred. Perhaps readers follow it to the label’s archive, return with a correction, or use it to identify a recording. Pageviews count loads, not what happened next—or whether anyone found what they came for.
That distinction matters for a journal covering Japanese noise, drone, electroacoustic work and small DIY releases. Search demand can be limited, documentation incomplete, and readers may follow references rather than browse several posts in a sitting. Start with what a piece is meant to do, then look for evidence that bears on that purpose.
Give each article a purpose before choosing a metric
A review, a discography note and an explanation of a recording technique need different tests. A review might help someone decide whether to seek out a release. A discography note might settle an ambiguous title or format. An account of tape manipulation might help a musician understand the process. Put that purpose in an editorial spreadsheet so you have something to refer back to when the numbers come in.
- Discovery: Is the article reaching people looking for its subject? Check relevant search queries and entrances, not visits alone.
- Use: Are readers reaching the details likely to answer their question? Look at interactions with track lists, notes, citations or related passages where you can measure them.
- Continuity: Does the piece lead to another relevant part of the archive? Look at meaningful onward paths, not total clicks.
- Response: Has it prompted a specific correction, question or submission? Keep a qualitative record alongside the numbers.
These are signals, not four targets every article has to meet. An investigation into an orphaned file may do its job by marking what remains unknown, even without an onward click. For a practical guide, whether readers can find and use the examples matters more.
Separate reach from attention
Start with entrances and users, keeping the definitions and time window consistent. A pageview can include several loads by one person; an entrance means the session began on that page. A brief spike after a discussion may reflect real interest, or people opening a tab and leaving it unread. Neither count tells you what they understood.
Engaged sessions, active time and scroll depth help, but they are easy to overread. Analytics systems define engagement differently. Someone may leave a page open while listening to a referenced recording; someone else may find a catalogue number in ten seconds and leave satisfied. A 75 percent scroll event tells you the browser reached that point, not that every preceding paragraph was read.
Read those figures against the page itself. If readers tend to leave before an essential source note near the bottom, its placement may need attention. The same drop after a long transcript, useful only to some readers, means less. Compare a long interview with other interviews rather than with a short release notice. Check mobile and desktop separately if a table or audio embed might behave differently.
Use a small set of deliberate events
Track actions whose purpose you can explain: opening a footnote, playing an audio embed, downloading a permitted reference document, or following an internal link to a related technique. Count both the action and, where possible, the people who had a chance to take it. Ten footnote clicks mean something different if twenty readers reached the note than if two thousand did.
Do not mistake interaction for endorsement. A play does not tell you whether someone finished the excerpt or liked it. A citation click might show curiosity—or signal that the claim needs clearer sourcing. Collect no more personal data than the editorial question requires, and check that events fire correctly before interpreting them.
Look for paths, not just exits
Leaving a page is not necessarily a sign of dissatisfaction. A reader may have found the date they needed, followed an external citation, or put down the phone to listen. On a specialist archive, a single page may serve as a reference, which makes exit rate especially easy to misread.
Onward paths are more revealing when the pages have a real connection. If a discussion of sound as physical material leads readers to an account of DIY electronics in Japanese experimental music, they may be interested in how the sounds were made. That does not prove either article persuaded them. Check which links they use, whether the links are visible on mobile, and whether the destination meets the expectation set by the anchor text.
On-site searches can expose missing connections. If readers repeatedly search for an artist’s alternate spelling after landing on a profile, a name variant may help. A no-results search might reveal a cataloguing gap rather than a weak article. Look at the actual query first; one typo is no reason to overhaul a page.
Read small numbers without pretending they are large
Specialist subjects produce thin data. With eight visitors in a week, a rise from two tracked interactions to four is not a stable trend. Put raw counts next to percentages, and use longer comparison periods when the publishing schedule and seasonality allow. Note any newsletter mention, artist repost, festival programme or newly indexed page that changed where visitors came from. Otherwise, ordinary fluctuations can look like progress or decline.
A simple monthly record is often enough:
- Write down the article’s intended use and its publication or revision date.
- Record entrances by broad source, including raw counts.
- Add one or two relevant behaviour measures and define them.
- Log substantive responses: corrections, rights concerns, reader questions or useful references sent in.
- Note distribution or technical changes that might explain a shift.
Keep the qualitative log selective. A message identifying an uncredited field recording is not the same as a generic compliment; it may call for a correction and a clearer provenance note. Silence proves nothing about the research. Some readers never write to editors, and some pages are found years after publication.
Compare against the article’s own history
For an archival entry, a before-and-after comparison can tell you more than a ranking against the site’s most visited post. If a corrected title and clearer opening bring in searches that match the recording, the edit may have helped people find it. Check the queries and the page before taking credit: an unrelated mention elsewhere could explain the increase.
For a new piece, give launch-day circulation time to settle before judging its routine use, and keep a short note on where it was shared. An article on an obscure electroacoustic method might receive only a handful of visits each month and still be a dependable reference. A broader headline written to chase traffic would make a different promise to those readers.
Make editorial decisions from evidence, not a score
One performance score built from visits, scrolls, clicks and time conceals the ways those signals disagree. High search entrances followed by quick exits might mean the opening misses the query—or answers it immediately. Low traffic and repeated, specific corrections might point to a small, knowledgeable readership and sourcing that needs work. Ask what else could explain the pattern before changing anything.
Make the smallest change the evidence supports: bring a buried catalogue note into view, repair an audio control, add a documented alternate spelling—or leave a page alone if it serves its narrow purpose. Record the edit and revisit the same measures later. If searches suggest a release came in a format the page does not mention, verify that format against reliable documentation before adding it. A search term is a question, not a source.