Are random links a sufficient replacement for link spread?

In short, using random links is not a sufficient replacement for link spread when considering an alternative to Graphite's Internal Links API.

Here's why:

Computing the expected number of incoming links to a page, suppose we have 100 pages, the probability of selecting each page is 1/100. If we select 10 outgoing links for each page, you may expect that in this setup that each page gets 10 links.

However, the true number is drawn from a binomial distribution with p=0.01 and n=100*10=1000. This is equivalent to flipping a weighted coin where the chance of heads is 1% 1,000 times. A page will only get exactly 10 incoming links 12.6% of the time. 45.7% of the time it will get fewer than 10 incoming links. 12.9% of the pages will get 6 or fewer internal links.

If the goal is that each page has greater than k links, where k is the number of links outgoing on each page, using random links is not an ideal replacement based on the percentages given above.

If the goal is to reduce lift in terms of effort of implementing the API, talk to our team to learn more about the ease of implementing the Internal Links API.