Everyone always told me that networking is a good thing. But I’m a nerd; I didn’t really believe that until I made the spreadsheet to prove it. Now I’m convinced. Specifically, if you want the rank of your blog to climb towards the left end of the long tail, you want to find a way to join a group of other bloggers who are advantaging each other in some way. But the lesson has implications far beyond blogs.
(full disclosure: one of the reasons that I’m interested in the subject is that I’m coordinating a network of entrepreneur-written blogs called My Way. Nice to know what benefits the member blogs might gain in terms of readership and what they have to do as members to gain that advantage.)
To start with the conclusion: if a subset of blogs were to organize themselves into a network and they were the only such network in the blogosphere (ain’t gonna happen!) and if they each favored each other only slightly in terms of links, in the absence of other factors (also ain’t gonna happen!), those blogs would rise directly to the top of the blogosphere.
With 100 iterations of my model, here’s how the affinity blogs rise in rank (remember low rank is good):
And, in a universe where the average blog has 10 inbound links, here’s how the average affinity blog does over the same 100 iterations:
Looks like groups are good, huh?
Here’s how the simulation was done:
The simulated world starts and ends with a universe of exactly 150 sites. No new sites are created during the simulation and none leave. Each site has exactly 10 outlinks (which may point back to itself). In the beginning the links are “fairly” distributed so that each site also has 10 inlinks (links pointing to it).
In each iteration, the links are redone. There is a one percent probability that a new link will point randomly to one of the 150 sites (call this the "stumble-upon" percentage). There is a 99 percent probability that a new link’s destination will be chosen randomly from the set of links in the previous iteration. The effect of that last rule is that sites with more links than the average in one generation are likely to have more links than the average in the next generation (“the rich get richer”). But it’s all fair, Everyone starts out even and only luck determines who gets more as time goes on.
But luck inexorably leads to a long tail distribution. Here’s the initial state (everyone equal):
After only 10 iterations, severe inequality has set in:
And after 100 iterations we have a true long tail distribution:
Now, if you’ve been paying attention and you’re a true nerd, you’re jumping up and down and waving your arms and yelling that this was NOT a fair simulation: ten of the sites ganged up and made an affinity group so of course there is now uneven distribution of inbound links.
Well, you’re are partly right so I reran the simulation with NO affinity group, everyone truly independent and equal. Here’s the graph after 100 iterations, still a long tail but not quite as skewed:
What did the affinity group do to advantage themselves and each other? They slightly modified the rule for assigning a new link. With a 20% probability, each time one of the group members repointed one of her or his links, it was pointed to a random member of the group (without regard to the current rank of that member). With an 80% probability, group members set their links the same way as everyone else (described above).
Lesson: a little preference goes a long way. Note that something like this would probably happen in the real world if group members just read each others blogs before they read anything else.
Now that I think the model works, I’m doing three things. I’m uploading it (Download longtailmaker4.xls) so you can look at it, play with it, and tell me what’s wrong with it. Just run the CREATE macro and you’ll be guided through setting parameters. It can run a long time.
And I’m planning to run with different parameters and get an idea what group sizes are efficient and the effect of different degrees of preference. I’ll let you know the results.
And I’m going to see if this knowledge helps the blogs in My Way gain readership.
Power rules series which produce the kinds of curves shown above have been recognized for more than a century. Chris Anderson’s book The Long Tail applies this math to observations of web-enabled phenomena and his blog continues the discussion.
My post probably should have been entitled “What Rotarians Already Know”.