A review picks up 300 page views after a newsletter goes out. Normally, it gets only a handful in a week. The spike tells you the newsletter brought traffic to the page—not that anyone finished the review, liked the record, or wants more writing like it. Reading analytics starts with separating what happened from what you think it means.
For an independent music site, the useful questions are fairly modest: Which pages do people find? Where do visits seem to come from? Are older pieces still being reached? Is a technical problem getting in the way? Basic analytics can help, as long as you read the numbers alongside publication dates, referral context, and the limits of your tracking setup.
Know what each number counts
Analytics tools define things differently. Check the glossary and configuration of the one your site uses; these common measures are a starting point, not interchangeable measures of readership:
- Page views count recorded page loads. One person opening the same article twice may produce two views.
- Users or visitors estimate distinct people or browsers according to the tool’s identification rules. One person using two devices may be counted twice; privacy settings may prevent recognition altogether.
- Sessions or visits group activity within a time window set by the tool. A returning reader can generate multiple sessions.
- Referrals and channels classify where a visit came from: search, another site, social traffic, a campaign link, or direct/unknown traffic. The classification is not always right.
- Events record actions you have configured, such as an outbound-link click or a file download. If an event was never configured, its absence tells you nothing about whether the action happened.
Name the measure behind a claim. “The review had 300 views” is safer than “300 people read the review.” Even a recorded view does not guarantee that the page displayed fully, let alone that someone spent time with the text.
Give comparisons a fair window
Compare equivalent periods before deciding a change matters. A page published yesterday has had less time to attract search traffic than one published six months ago. An archive essay about cassette distribution might draw a slow trickle of visits, while a festival announcement gets most of its traffic before the event. Their totals answer different questions.
For a weekly check, use the same seven-day span each time and note what was published or distributed. When comparing articles, look at both a fixed early window—say, the first 30 days after publication—and a later period. Keep the dates beside the numbers. A newsletter send, an artist’s mention, or temporary homepage placement may explain a rise without suggesting any change in audience taste.
Percentages can be especially misleading on a small site. Five visits becoming ten is a 100% increase, but it is still only five additional visits. Show the counts too, and resist calling a small fluctuation a trend. For pages with little traffic, several months of observations are often more useful than a daily chart.
Compare pages with similar jobs
A short news item, a long listening guide, and a contact page are doing different work. A contact page may succeed if a few relevant visitors find the email address; a guide may keep proving useful through search visits long after publication. Group like with like when asking editorial questions—older reviews with older reviews, for instance, or introductory explainers with other explainers. Individual exceptions will still matter more than a league table of pages.
Trace the route, not just the total
Open a page-level report and check its traffic sources. Search visits to an older field-recording essay may suggest it answers a continuing query. A sudden cluster of referrals from one music forum points to a particular discussion. Direct or unknown traffic is murkier: it can include bookmarks, typed addresses, untagged email links, or visits with no available referrer information.
Search-query reports, where available, may expose a gap between what a page promises and what visitors seem to want. But queries are often incomplete or withheld, and a query that brought someone to the page cannot tell you whether they found an answer. Read the essay before changing its title or structure. An accurate, distinctive piece need not be rewritten to pursue a vaguely related search phrase.
On-site navigation offers another clue if your setup records it. Readers moving from an introduction into a specific review show that at least one route through the archive is being used. If they do not make that move, the link might be buried, the review might not be relevant, or the tool might have missed the transition. The report gives you somewhere to look, not a diagnosis.
Handle engagement measures carefully
Engagement time, bounce rate, and exit rate invite more certainty than they deserve. Definitions vary by platform, and some measures depend on whether the tool recorded another event. Someone could open a long interview, read it, and leave without clicking. Depending on the setup, that visit may look less engaged than a quick series of clicks through several pages.
A high exit rate on a complete article is not necessarily bad; the reader may have found what they came for. A high exit rate partway through a multi-page feature, coupled with reports of broken navigation, calls for a closer look. Test the page on a phone, check how it loads, and see whether the headings and links make sense. A single engagement score cannot judge the writing.
Check what the instrument misses
Analytics gives you a partial record. Consent settings, ad blockers, browser restrictions, bot filters, and tracking errors can all change reported traffic. So can a configuration change, even when the audience has not changed at all. Before reading meaning into a sharp break in the chart, check whether the tracking code, consent banner, site theme, or analytics settings changed around that date.
Automated traffic deserves attention too, especially if a small site sees an abrupt surge. Many visits in a short interval with no plausible referral, or a cluster tied to an unexpected location, may raise questions; neither proves bots were involved. If a spike would affect a decision, compare the analytics report with available server logs or hosting reports. They may disagree simply because they count different things.
Privacy matters. An independent music site rarely needs invasive tracking to answer basic editorial questions. Aggregate page and referral information is usually enough to spot an inaccessible article or a useful route into the archive. Treat inferred identity, location, and individual listening habits with caution.
Turn a report into a bounded decision
Keep a short log with four fields: the question, the measure and date range, relevant context, and the next check. For example: “Are older reviews being found? Search visits to reviews published more than a year ago, measured over the last three months; one review was linked in a recent newsletter; inspect the remaining reviews for broken pages and unclear titles.” That records what the evidence supports without deciding that a quiet review is unwanted.
Before editing, decide what result would change your mind. If an article seems to get search visits but few onward clicks, inspect its navigation, make any warranted change, and compare a reasonable period afterward. If the sample is tiny, leave the result open. An archival review can document an obscure release even if it never draws much traffic.
This week, pick one older article. Note its views, approximate visitor count, leading traffic source, and measurement dates. Then open it on a phone and follow the first relevant internal link. A broken link gives you something definite to fix. If it works, keep the numbers on record without inventing a reason for them.