Article effectiveness score

Article effectiveness score is a composite measure of how well a single help centre or knowledge base article resolves the question it is meant to answer. It combines three signals into one number: did readers find the article, did it stop them raising a ticket, and did they say it helped. An article with 5,000 views that deflects 80 percent of related questions and earns a 90 percent helpful rating scores far higher than one with the same traffic that helps nobody.

The score matters because help content is leverage. A good article answers the same question thousands of times with no agent involvement, which lowers ticket volume and frees the team for complex work. A weak article does the opposite. It ranks, gets read, fails to resolve the question, and the reader raises a ticket anyway, so you pay twice.

Unlike raw page views, the effectiveness score tells you whether content is working, not just whether it is being seen. A page can be popular and useless. The score is designed to separate those two states so you can invest your editing time where it changes outcomes.

There is no single industry formula, so the score is a weighted blend of the signals you can reliably measure. Choose weights that reflect what the article is for. A troubleshooting guide should lean on deflection, while an onboarding explainer might lean on helpfulness. The inputs below are the common building blocks.

1. 1

   Deflection rate

   The share of article views that did not result in a ticket on the same topic within a short window. This is the strongest signal that the article actually resolved the question.

2. 2

   Helpful rate

   The share of readers who rated the article helpful through a thumbs up or yes prompt. It is subjective but captures clarity that deflection alone can miss.

3. 3

   Search click rate

   The share of relevant searches that clicked through to the article. A perfect article nobody can find still scores low, so findability belongs in the blend.

4. 4

   Weight and combine

   Assign each signal a weight that sums to one, multiply, and add. Keep the weights consistent across articles so scores are comparable.

A single effectiveness score tells you an article is underperforming but not why. The cause could be poor findability, outdated content, or a confusing layout, and each sits with a different owner. A metric tree breaks the score into the drivers that produce it, so a low score points to a specific fix rather than a vague instruction to improve the article.

KPI Tree lets you connect each branch to the team that owns it. Findability sits with the SEO and search owner, content accuracy sits with the subject expert, and the helpful rating sits with the content editor. When a score drops, the change is pushed to the accountable owner, so the person who can fix the outdated screenshot is the one who hears about it, not the whole team.

Because the score is a custom blend, the right benchmark is internal consistency rather than an external standard. The ranges below assume a score normalised to 100 and give a practical read on when an article is pulling its weight and when it needs work.

Improving the score means lifting whichever signal is dragging it down, not editing at random. Read the tree first to find the weak branch, then apply the matching lever below.

1. 1

   Ranking articles by views

   A high traffic article can have a low effectiveness score. Sort by score, not by popularity, or you will pour editing time into pages that are already working.

2. 2

   Trusting helpful ratings alone

   Self reported helpfulness has low response rates and a positive bias. Anchor the score on deflection, which is behavioural, and treat ratings as a supporting signal.

3. 3

   Counting deflection too loosely

   If your deflection window is too long, unrelated tickets count against the article. Tie deflection to tickets on the same topic within a short, fixed window.

4. 4

   Changing the weights mid stream

   Reweighting the formula makes scores incomparable across time. Lock the weights so a rising score reflects a better article, not a moved goalpost.