How We Measure Article Performance: The Analytics That Actually Matter

If you run a content site in 2026, the analytics dashboard will hand you a number for everything. Pageviews, sessions, bounce rate, average session duration, pages per session, traffic sources, device breakdowns, scroll depth, click-through rate, conversion rate — every platform from GA4 to Plausible to Fathom ships a dozen metrics by default, and every third-party tool adds another dozen. The trap is not that any of them are wrong. The trap is that most of them are directionally correlated with each other, and if you do not decide which ones actually measure whether your content is working, the dashboard will pick for you. It will pick pageviews.

I am writing this article because I promised transparency when I described how Pickuma works. I have written about how we use AI in our editorial workflow and why we chose developer tools as a niche. This piece closes the loop: given that we publish, how do we know whether the articles are any good?

The Metrics We Track (and the Ones We Ignore)

Let me start with the list of metrics I look at every week, and then explain why each one earned its spot and what it told us over six months of data.

The four I treat as primary signals: engaged time-on-page, affiliate click-through rate, return visitor ratio, and review depth correlation to conversion. These four together tell me whether someone read the article, whether it moved them toward a tool decision, whether they trusted the site enough to come back, and whether the effort we put into deeper reviews translates into actual outcomes.

The four I ignore: raw pageviews, social share counts, bounce rate as reported by GA4, and pages per session. Each of these sounds useful and each of them will lead you toward the wrong editorial decisions if you optimize for them.

Let me go through both lists.

Engaged Time-on-Page

Google Analytics defines engaged sessions as those lasting longer than ten seconds, or having a conversion event, or having two or more pageviews. That is a floor so low it qualifies as tripping hazard. Ten seconds is not reading. Ten seconds is loading the page, scanning the headline, and scrolling once.

What I actually look at is the distribution of time-on-page for each article. GA4 reports it in buckets — under ten seconds, ten to thirty, thirty to sixty, over sixty, over three minutes. For a 1,200-word review article, the over-three-minutes bucket is the one that signals an actual reader. Those are people who opened the article, started reading, and kept going. When that bucket drops below eight percent of total pageviews, the article is either the wrong topic, the wrong introduction, or ranking for a search intent that does not match what the article delivers.

Across six months, articles in our “how-to” and deep-dive review categories averaged 18 percent three-minute reads, while broader comparison articles averaged 12 percent. The difference is about three percentage points more than I expected, and it has changed how I prioritize topics.

Affiliate Click-Through Rate

This one is straightforward to define but easy to misinterpret. An affiliate CTR is the number of clicks on an outbound affiliate link divided by the number of unique pageviews. If 500 people land on a review and 35 click through to the tool’s website, the CTR is seven percent.

What makes this metric useful is tracking it over time, per article, relative to its own baseline. A review might launch with a 5 percent CTR, drift up to 8 percent after we refine the call-to-action placement, then settle around 6 percent as the article ages and search intent shifts. None of those numbers are inherently good or bad. The question is whether the trajectory makes sense relative to the changes we made and the tool’s positioning in the market.

What the CTR does not tell you — and this is where most publishers go wrong — is whether the reader converted into a paying customer after clicking. Unless you have access to the affiliate partner’s conversion data, you are measuring interest, not revenue. I treat CTR as a signal of whether the article successfully communicates why a tool is worth investigating, not whether it closed the sale.

On Pickuma, our average CTR across all affiliate links sits around 4 to 6 percent. Reviews that directly solve a named, specific problem — “I need to deploy a Next.js app without Kubernetes” — tend to generate CTRs in the 8 to 12 percent range. Reviews that compare five tools generically land around 2 to 4 percent. The difference is the same story as the time-on-page data: specificity converts.

Return Visitor Ratio

This is the metric I pay attention to when I want to know whether the site is building trust. A return visitor is someone who visited Pickuma at least once before in the trailing 28-day window and came back. GA4 reports this as a percentage of total users.

If every article on your site is optimized for search traffic and nothing else, return visitor ratios trend toward zero — you are a one-and-done answer box, and Google is your entire distribution channel. That is not inherently bad, but it is fragile. An algorithm update, a new competitor buying your exact keywords, or a shift in how Google surfaces review content can wipe out your traffic overnight. A growing return visitor ratio means people are bookmarking the site, remembering the name, or coming back because one article led them to another.

Pickuma’s return visitor ratio has moved from about 6 percent in the first month to 14 percent by month six. That is still low by the standards of a community-driven site, but the upward trend is what I track. The articles that generate the most return visitors are not the search-traffic-heavy comparison posts — they are the deep reviews and the meta pieces like this one, where the takeaway is not “which tool should I pick” but “I want to read more from this person.” That signal changes how I think about content mix.

Review Depth vs. Conversion Rate

This is the metric that took the longest to observe because it requires enough articles with enough affiliate conversion data for the pattern to emerge. The question is simple: does writing a longer, more detailed review actually produce more affiliate conversions than writing a shorter, checklist-style comparison?

The answer, from our data, is yes — but not in the way I expected. Depth alone does not predict conversion. Word count shows no meaningful correlation once you control for the topic. What does correlate is whether the review answers the three questions every developer asks before trying a tool: what is the actual setup time, what breaks when you push it past the demo use case, and what does the pricing page avoid telling you. When a review answers all three, the affiliate CTR is consistently higher — sometimes double — compared to reviews that cover features and benchmarks without the friction points.

I measure depth qualitatively rather than by a formula. Each review article gets a simple check: did we actually use the tool long enough to hit a real limitation, and did we document that limitation in the article? If yes, the article almost always outperforms the ones where we describe what the tool is supposed to do without testing whether it does it.

How We Use GA4 for Content Analytics

Google Analytics 4 is a sprawling tool, and it is easy to drown in it. My approach is deliberately narrow. I have three reports I look at weekly, and I ignore everything else.

The first is the Pages and Screens report, filtered to the trailing seven days, sorted by views, with the comparison toggle set to the previous seven-day period. This tells me which articles are getting traffic and whether that traffic is moving up or down. I do not make editorial decisions from this report — traffic changes are downstream of search rankings, seasonality, and link rot, none of which I can control directly. I use this report to flag anomalies: an article that lost 60 percent of its traffic in a week is worth investigating. An article that gained 200 percent is worth understanding and amplifying.

The second is a custom Exploration report I built for engaged sessions by page path. GA4 Explorations let you drag dimensions and metrics into a free-form table, and I keep one pinned that shows article URL, engaged sessions, average engagement time, and event count — all filtered to the “engaged session” segment. This is where I look for the disconnect between traffic volume and actual reading behavior. An article with high pageviews and near-zero three-minute reads is ranking for something but not delivering. An article with modest traffic and a high engagement rate is doing the thing we actually care about.

The third is the Traffic Acquisition report, which I use once a month to check whether our channel mix is diversifying or consolidating. Right now, about 70 percent of Pickuma’s traffic comes from organic search, 15 percent from direct, and the remainder from referral and social. I track direct traffic as a share of total because it is the closest proxy GA4 offers for brand recall — people typing the URL or clicking a bookmark. If that number trends up over six months, the site is building a relationship independent of Google’s ranking algorithms. If it stays flat, every article is effectively rented distribution.

How We Determine Whether a Review Article Is Working

A review article on Pickuma has three jobs: answer the developer’s specific question, help them decide whether the tool is worth trying, and earn enough trust that they come back for the next decision. The metrics I described above each map to one of these jobs — time-on-page to question-answering, CTR to the try-it decision, return ratio to trust.

I evaluate whether an article is working by looking at all three together, not by picking a single threshold. A review with a 6 percent CTR and a 20 percent three-minute read rate is working — people are reading and clicking. A review with a 2 percent CTR and a 25 percent three-minute read rate is complicated. It means people are reading the whole thing and deciding not to click. That could mean the article is doing its job honestly — warning people away from a tool that would waste their time. Or it could mean the article buried the call-to-action or failed to make the case for why trying the tool is worth the effort. I investigate those cases individually.

The reviews that unambiguously work are the ones where people read, click, and then — this is the hardest signal to get — come back to the site within the same week. Those sessions show up in GA4 as “engaged sessions” with a second article view, and they strongly suggest the first review led to a tool trial that went well enough for the developer to trust us with the next decision.

Over six months, about 40 percent of the review articles we have published fall into the working category by these criteria. Roughly 35 percent are in the complicated bucket — good engagement, low CTR. The remaining 25 percent need rewriting or repositioning. I consider that a healthy distribution for a young site, because it means we are not publishing only the articles that are easy to make perform. Some of the most valuable articles are the ones that tell someone not to use a tool, and those will never convert at high rates — but they build the trust that makes future conversions possible.

What We Learned from Six Months of Data

If I had to reduce the last six months of analytics to three lessons, they would be these.

First, traffic volume is a distraction during the first year. A site this young does not have the domain authority to compete for high-volume keywords, and trying to force it leads to publishing articles optimized for search engines rather than readers. The articles that built Pickuma’s returning audience were the ones written for a specific person with a specific problem, not for a keyword. Those articles got less traffic on day one but more traffic on day 90, because the people who found them stayed.

Second, the number of affiliate clicks per article is a better signal of editorial quality than the number of pageviews per article. A single article with 1,500 pageviews and 90 affiliate clicks did more for the site than three articles with 3,000 pageviews each and 10 clicks. When I started tracking click volume rather than traffic volume as the success metric, the editorial calendar changed. Fewer articles, more depth, better outcomes.

Third, scroll depth is more honest than time-on-page, but also harder to act on. GA4 reports scroll depth as an event, and the numbers are grim — fewer than 30 percent of visitors scroll past the 75 percent mark on any given article. But scroll depth does not tell you why someone stopped scrolling. Did they get the answer they needed in the first half? Did they lose interest? Did the page load slowly and they bounced before the bottom rendered? Time-on-page, for all its flaws as a GA4 default metric, gives you a second dimension — if someone stopped scrolling at 50 percent but stayed for four minutes, they probably found what they needed. If they stopped at 50 percent and stayed for 20 seconds, the introduction did not hook them.