Did they actually read it?

Numbers without context mislead you

Is 100 km/h fast? It depends entirely on whether you're on a motorway or a residential street. A number in isolation rarely means anything — only context makes it useful.

The same is true in publishing analytics. Take these two articles:

Article A

View time: 40 sec · Scroll depth: 70% · Article length: 300 words

Article B

View time: 50 sec · Scroll depth: 70% · Article length: 600 words

The misleading result

Same metrics — but Article A (300 words) likely performed much better than Article B (600 words).

Context changes everything. That's exactly why Consumed exists.

Two good metrics. One shared blind spot.

Scroll depth

Shows how far readers scroll — but not how they scrolled. A reader may rapidly scan to the bottom looking for related articles. Scroll depth alone can't tell you if they read a word.

View time

A reader spending 40 seconds on a 300-word article is impressive. The same 40 seconds on a 1200-word feature is actually quite poor. View time without knowing article length is close to meaningless.

Both metrics try to answer the same question: did the reader actually read the article? At Kilkaya, we approach that question more directly.

Introducing Consumed

Instead of juggling several ambiguous numbers, Consumed asks one simple binary question: did the reader consume this article — yes or no?

Consumed

Spent enough time relative to article length and scrolled at least 60% of the article.

Not consumed

Didn't meet either or both criteria — even if they spent some time or scrolled a bit.

But is 60% actually good?

A 60% Consumed rate on a 300-word article is very different from 60% on a 2000-word investigative feature. Short articles are naturally easier to consume — in some cases, the first screen view alone puts 50% of the article in sight.

Kilkaya analyzed thousands of articles to understand the relationship between word count and Consumed rate — and built a formula that calculates what you should expect for any given article length.

92%Very short (under 200 words)78%Short (200–400 words)58%Medium (400–800 words)40%Long (800–1500 words)25%Very long (1500+ words)

Meet Diff Consumed — the number that actually matters

The difference between actual and expected Consumed — Diff Consumed — tells you whether an article genuinely engaged readers beyond what you'd expect. It's the fairest way to compare a 200-word news brief with a 2000-word feature.

A model that evolves with you

Right now, Expected Consumed is based on article length alone — and that already creates a far fairer baseline. But the model is designed to go further. Additional factors can be layered in to reflect how your readers actually behave.

Section or topic — Sports vs. politics vs. culture have different reading patternsSubscriber vs. free — Paying readers typically engage differentlyEditorial intent — A breaking news brief has different expectations than a long featureCustom parameters — Any signal your CMS provides can be incorporated

The goal is not to lower the bar. The goal is to create a fair comparison that reveals which stories genuinely capture readers' attention — regardless of whether they are short updates or long, in-depth features.