A visitor opens a Finnish thrash demo entry, checks its discography reference, and finds the recording they came for. Another lands on the homepage and leaves. Both register as page views, but the number says little about whether either visit was useful. What matters depends on the page: helping someone find a recording, check a source, understand a scene connection, or place a show in its historical context.
For an independent music archive, analytics should help editors make evidence easier to find, not turn every article into a traffic contest. Begin with a few questions about what visitors do, and use figures that answer those questions without claiming more than they can show.
Assign each page a job before choosing a number
A release entry, a long-form essay, and a concert chronology call for different measures. On a release entry, visitors may need the format, track listing, and documented date. An essay may lead them from an interpretation to its supporting examples. A chronology should make a year or venue easy to find. Visits show demand for these pages, but not whether the pages do their jobs.
For a representative page in each category, write down an action you can observe. “Appreciate the significance of Finnish thrash” is an editorial aim, not something analytics can measure directly. “Open a cited release entry from an essay” can be counted, though it still says nothing about whether the visitor agrees with the essay.
- Discovery pages: Track arrivals from search and the share of visitors who continue to a relevant archive entry.
- Reference entries: Track use of links to related releases, bands, or documented shows.
- Long-form articles: Track whether readers reach source notes or follow a clearly identified internal reference.
- Archive search: Track searches that return no results and whether visitors refine them.
These are starting points, not universal targets. Someone may find a precise release date on one page and leave satisfied. Treating that visit as a failure would reward extra clicks rather than better reference material.
Separate reach from useful engagement
Use traffic metrics to understand demand
Users, visits, page views, and landing pages answer different questions. A user count estimates distinct browsers or devices, not individual people. Visits describe sessions; page views count page loads or views. Landing pages show where visits began. For a small archive, that report can reveal whether people tend to arrive through a band entry, a piece about a recording, or the homepage.
Compare similar periods rather than reading too much into a spike. A festival mention, a newly indexed page, or an outside discussion may briefly raise traffic. If an older discography entry draws more visits, check where those visits came from and what happened on neighboring pages before concluding that interest in its subject has grown. Search traffic may reflect the wording of a query, not a band’s historical importance.
Choose engagement signals tied to a task
An internal-link click can show a move from context to evidence. A history of demo circulation, for instance, might point readers to how Finnish thrash demos found their listeners. Counting that move tells you more than counting every click on the article. It also gives you a reason to check whether the destination delivers what the link promises.
Scroll depth suggests that someone reached material near the bottom of a long article; it cannot establish careful reading. Time on page has a similar weakness: a visitor may leave a tab open while doing something else. Use either measure to investigate a problem, not certify quality. If few people reach a source note containing essential qualifications, bringing those qualifications into the main text may be more useful than chasing a higher scroll-depth figure.
Build a modest metric set
A dashboard need not display every figure the platform offers. For each important page type, choose one measure of reach, one or two measures of task completion, and one measure of friction. Write down each definition so the next editor knows what a rise or fall actually represents.
| Editorial question | Metric to inspect | What it cannot establish |
|---|---|---|
| Which entries draw new visitors? | Landing-page visits by page category | Whether the entries satisfy those visitors |
| Can visitors locate a release? | Archive searches with no results | Whether the requested release exists |
| Do essays connect readers to evidence? | Clicks on relevant internal references | Whether readers accepted the evidence |
| Does mobile navigation work? | Task-completion events by device type | The cause of a difference without testing |
Name events after actions, not conclusions. archive_search_submitted records something concrete; reader_engaged builds an assumption into the label. Document whether counts include repeat actions by one visitor and whether automated traffic is excluded. If the archive changes analytics tools, consistent definitions matter more than making the new dashboard look like the old one.
Segment carefully, especially when numbers are small
A site-wide average can hide a broken path. If mobile visitors use archive search less often than desktop visitors, compare the interface at both screen sizes. If people arriving at a release entry rarely open its band overview, look at the wording and placement of that link. Split figures by page type, device, or entry source only when the groups are large enough for a sensible comparison.
Small archives often have sparse data. Three clicks one month and six the next is a doubling, but it is also just three additional clicks. Show counts beside percentages and look across several periods. Nor does a rise after an edit prove that the edit caused it: a new incoming reference or a shift in search visibility might explain the change.
Search queries need particular care. A no-results search might point to an alternate band spelling, a missing release format, or an interface typo. It might also ask for a recording the archive cannot document. Use query logs to investigate cataloguing and navigation, not as grounds for inventing an entry. The archive’s credibility depends on that distinction.
Keep measurement proportionate to privacy and editorial purpose
More visitor data does not automatically lead to better decisions. Aggregated page reports and a few navigation events may answer most editorial questions. Avoid sending names, email addresses, or full free-text submissions to analytics unless there is a clear need and an appropriate privacy process. Check, too, whether search-term reporting could retain personal information someone typed into the search box.
Consent rules and retention obligations depend on the tools and jurisdictions involved. Make sure the site’s privacy information describes what it collects, and revisit analytics settings when the site changes. A small measurement plan is easier to audit and maintain.
Traffic should not decide which history deserves coverage. An obscure cassette may draw few visits while filling a real gap in the documented record. Analytics can reveal that readers struggle to find its entry; deciding whether it belongs in the archive still calls for source assessment, context, and editorial judgment.
Turn a finding into a specific edit
Review figures at a pace that suits the archive’s publishing schedule. A monthly check may catch broken links between entries or unexpected search failures; a longer comparison offers a fairer view of article performance. When you make an edit, note the date and the problem it was meant to solve. That record helps you read the later figures.
Suppose searches for a band name repeatedly return no results, although its entry exists under a spelling used on a particular release. Verify that spelling against the release material, add the documented alternate term to site search, and test both versions on a phone and a desktop browser. At the next review, compare no-results searches for those terms with the total searches for the band. It is a narrow test, but it can tell you whether the cataloguing change helped people find the existing record.
