Read-through depth

Time on page tells you how long someone stayed, not how far they got. Three minutes on a twelve minute article can mean a careful read of the first quarter, or a skim of the whole thing. Read-through separates the two: for every page view, Pulse records the deepest point of the article that entered the screen, as a number from 0 to 100.

What is measured

The measurement is relative to the article, not to the whole document. Reaching 100% of a document would mean scrolling past the footer, the comments and the related posts, which a reader who finishes the last paragraph and closes the tab never does.

The tracker looks for the article in this order:

1. the element matched by data-content-selector, if you set one
2. the first <article> element
3. the first <main> element
4. the document itself

The height is read again on every measurement, so lazy-loaded images, web fonts and embeds that change the page after load are taken into account.

Pointing the tracker at your content

Most themes need no configuration. Add data-content-selector when your theme either has no <article> and no <main>, or wraps more than the article body in them (comments, author box, related posts):

<script
  defer async
  src="https://tracker.hakanai.io/hakanai.min.js"
  data-site="YOUR_SITE_ID"
  data-content-selector=".post-body">
</script>

The value is a CSS selector, matched with querySelector. The first match wins. If the selector matches nothing, or is not valid, the tracker falls back to the order above rather than failing.

Without the attribute on a theme whose <article> also contains the comments, the low thresholds stay accurate and the "Reached the end" figure reads lower than reality.

Which views are excluded

Some page views carry no depth at all. They are excluded from the panel, from both the numerator and the denominator, which is why the panel states how many views it computed on:

• Pages shorter than the screen. Nothing scrolls, so every threshold would be reached just by opening the page. Counting them would push every bar up for no reason.
• Views from an older tracker. The script is cached by your visitors' browsers. After a release, page views keep arriving without the field for as long as those caches live. Coverage climbs back on its own.

A depth of 0 is a different thing from no depth at all: it means the view was measured and the reader never scrolled.

Reading the panel

The panel sits on the post page, under the traffic sources.

• Reached 25 / 50 / 75 / the end. The share of measured views that got at least that far. The bars can only decrease, since each view is counted at the deepest point it reached.
• Reached the end counts views that got to 90% of the article. The cut-off is below 100 on purpose, to absorb the byline, the tags and the share row that themes put after the last paragraph.
• Median depth. Half the views got further than this, half got less far.
• Completion. The same figure as the last bar, promoted to the top of the panel.
• vs site median. The gap between this article's median and the median of the whole site over the same period, as a relative difference.

Below 30 measured views the panel shows when collection started instead of a distribution: one view would move a bar by several points, and the shape would not mean anything yet.

Billing

Read-through costs no additional event. The depth travels on the page exit signal the tracker already sends, and your quota is only charged when the page view itself is created.