How is the internal PageRank score calculated?

[dev_mike]

I'm technically curious — what algorithm does LinkJuice use to calculate internal PageRank? Is it the classic iterative PageRank or something different?

[Kate M.]

LinkJuice uses the standard iterative PageRank algorithm with a damping factor of 0.85. Each iteration propagates scores through the directed graph until convergence (delta < 0.0001).

Starting scores are distributed uniformly across all nodes. The damping factor models the probability of a random surfer continuing to follow links vs. jumping to a random page.

The scores are then normalised so the mean equals 1.0, making it easy to compare pages relative to the site average.

[Oliver]

Totally agree with the above. I'd also look at the outgoing links from your strongest pages. If your homepage links to 200+ things it loses focus. Cutting that list to 30-40 links concentrating on commercial pages made a noticeable difference for us.

[Dev]

Just adding a data point: we did this analysis for an 80,000-page e-commerce site and the run time was under 2 minutes in Chrome. No issues at all. The orphan detection alone saved us probably 15 hours of manual audit work.