What We Learned from 11 PBN Experiments

In March 2017, there were changes in Google’s algorithm that greatly affected PBNs as a ranking strategy.

When I was running Diggity Links (now Rank Club), we would test every single domain before adding it to our network.

If a domain’s link caused a positive ranking increase when linking to a test domain, it was added to the network.  Otherwise it went in the trash.

Then, all of a sudden, PBN test success rates dropped down to disturbing levels.

Depending on the type of domain (expired vs backorder/dropcatch vs auction) pass rates got as low as 20% while, at the same time, as many as 50% of domains were toxic.

Yes, they would actually hurt whatever they linked to.

It was quite shocking but an interesting problem to solve, nonetheless.

What had Google changed?  And how could we turn this around in our favor?

My R&D director Rob Rok (also now at Rank Club) and I sat down and designed 11 test experiments involving over 300 PBNs.

This is what we found.

You’re about to learn the exact filters and factors we found that increased the pass rate and power of our PBNs.

How we Test

slide from the affiliate lab

Borrowed from affiliatelab.im

Our test methodology is very straight forward.

It all boils down to the testcases we select which will be the target URLs which we’re going to send our test links to.

We find target URLs with the following qualities:

  • Haven’t had links in years
  • They rank on page 2-4 for some non-competitive keyword
  • They have no links at all, or have never had a target anchor text sent

A good example would be a URL ranking for “garth brooks tickets nashville 2015”.

When you select testcases like this, you remove two important factors that can easily botch your test environment:

  • Optimization: If you over-optimize with the wrong anchor text, then you won’t get a good result whether or not the PBN is good or bad.
  • External Variables: If the target URL is getting links from elsewhere, how do you know if it was your domain that caused a shift in rankings?

All that remains to be tested is: is the domain toxic or not?

You can read more about our testing plan here.

Here’s some examples of passing results (link was placed on the date in the black box):

positive

Neutral results:

neutral

Toxic results:

toxic

The Baseline

I mentioned before that PBN pass rates suddenly were terrible.

Just how terrible?

baseline

Depending on the domain type, there would be different pass rates, flat and toxic rates.

But the impact was written on the wall.

The current way we were doing things was inefficient.  Auction domains (which had the best pass rates) didn’t even have a 50% chance of being an asset.

It was time to hit the lab.

Now that we had the baseline numbers (our control group), we needed to test various theories and see if we could beat those numbers.

11 PBN Test Experiments and their Results


Theory #1: All Domains Need to Sit Unused for a While Before Being Linked From

In 2016, I discovered that expired domains can be toxic if linked out from too early.  Does this also apply to backorder and auction domains?

Test Structure

Test 20 backorder domains and 20 auction domains by linking out from them immediately.  Then wait 35 days and repeat.

Look for a difference in pass rates between the two test times.

Conclusion: True

holding period

What we’re looking for here is are situations I’ve marked in green where the domain…

  • Fails the first test (Toxic)
  • Passes the second test 35 days later (Pass)

This situation resembles similar behavior I saw with expired domains where it seems as if some filter is being applied to links created too soon.

Some conclusions to take from this data:

  • Auction domains benefitted more often from waiting-before-linking as compared to backorder domains. This suggests that there’s more going on with backorder domains.  We’ll visit this again later.
  • Simply waiting before linking was not a cure-all. There were still plenty of toxic and neutral domains in both categories.
  • Domains can still pass without waiting. It doesn’t seem to be a requirement to wait, but the trend shows that it helps improve pass rates.

We didn’t solve this whole PBN problem in one fell swoop, but we got some guidance.

Nonetheless, we decided to change our processes to wait before linking out from auction domains because it increases the chance of success.

Early win.

Theory #2: A Domain’s Backlink Profile Resets when it Expires

When a domain has expired, goes to backorder, or auction, then perhaps Google resets the backlink profile and thus it’s link juice.

Test Structure

Test 10 domains of each type of domain.

If any pass, then Google has not put a blanket “reset” all backlinks on the domain type.

Note: A requirement of the expired domains for this test was that they were unused for over a year.

Conclusion: False

Some domains passed (with similar rates to the baseline).

reset link juice

Honestly, I didn’t think this theory was going to be true at all, but you gotta cover all the bases when testing.

Theory #3: Monetization of PBNs Hurts Their Pass Rates

Just like over-commercialization of money sites might affect their ability to rank, over-commercialization of PBNs might affect their ability to pass positive link juice.

I know what you’re thinking…

It’s a stretch but we monetized our PBNs:

sample

Perhaps we should remove the monetization?

Test Structure

Actually this test is completely unneeded.

Our monetization is purely cosmetic.  There’s no actual affiliate cookies being transferred on clicks.

What the Googlebot sees is likely just images.

Conclusion: NA

Perhaps this is “a thing”, but we didn’t have any testcases to test it.

Theory #4: PBNs Require Rankings and Traffic to Pass Value as a Link

The concept makes a lot of logical and conceptual sense.

If a PBN has organic traffic, that means it ranks, which means Google likes it, which means it can be used as a link.

Test Structure

We would gather all 158 PBNs that had passed in our experiments.

We would throw them into SEMRush, Ahrefs, and Similar Web to see if they’re ranking for anything and thus pulling traffic.

Any sites with over 100 visitors per month we would label as “having traffic”.  Less than 100 visitors per month would be labeled as “no traffic”.

Conclusion: False

pbns with traffic

Despite popular opinion, a PBN doesn’t require traffic in order to be a positive link asset.

I wasn’t surprised.

On one hand, I literally wrote the book last year on the benefits of getting links from ranking PBNs.

But I never felt like it was a “must have”.  Rather it is a “nice to have”.

And this Ahrefs study showing that “91% of the internet has no traffic” provides more qualitative support for these results.

I mean, think about it…

Does only 9% of the internet qualify as a link source?

Theory #6: Keep Content Consistent Between the Old Site and the New PBN

Jason Duke, a member of my testing group, had been brainstorming with us.

Jason mentioned that he had personally experienced success from keeping content consistent from one site to the other by leveraging Wayback installations.

At Diggity Links, since we were doing complete WordPress rebuilds, this wasn’t an option for us.

So instead we would…

Test Structure

Take 20 PBNs from each category.

We would look up the domains from archive.org to pull homepage content from the old domain which would become the first post on the new WordPress PBN version.

We named this process “Duking” the PBNs.

We would then test these domains to see if they improved from the baseline.

Conclusion: True

Here’s a comparison of the pass rates per category between the baseline numbers and the Duking results…

duking

Expired domains and auctions domains had a significant benefit from Duking.

The benefit on backorder domains was negligible.  We’ll explore why later.

But as a whole, keeping content continuous seems to be a simple and cheap way to significantly increase pass rates on expired and auction domains.

That said, there are legal implications of copying people’s content, even on an expired site that you now own.

Consult a lawyer, homie.

Theory #7: Phone Verified Google Accounts (PVA) Might Increase Pass Rate for PBNs.

Most “real” websites are connected to GSC, Analytics, Google+ or some other kind of Google property.  Perhaps it’s a trust factor when it comes to links.

Test Structure

Purchase 60 PVA Google accounts and connect them to 60 PBNs (20 of each category).

Observe the improvement (or lack of) from the baseline result.

Conclusion: False

pva

The addition of the PVA accounts made negligible difference to the result.

You could even argue that the results got a tad bit worse, but I’m going to chalk that up to only having 60 PBNs in the experimental group.

Theory #8: Backorder/Dropcatch Domains are Caught on Toxic, Bad-Neighborhood Registrars

My favorite prodigy SEO, Yashar Ghaffarloo, came up with this one in our testing group.

The theory is that most “normal” people would never buy domains on dropcatchers like Xz, Pheenix, Dropcatch, etc.

Perhaps Google just made a simple filter, potentially penalizing some or all links created from certain registrars.

Test Structure

Buy 30 dropcatch’d PBNs from various backorder services.

  • Group A (10 PBNs) – Moved registrars to Godaddy and linked from after 3 days.
  • Group B (10 PBNs) – Same as above but linked after 15 days
  • Group C (10 PBNs) – Same as above but linked after 30 days

Conclusion: True

backorder - toxic registrars

Hot damn.

These are some pretty compelling results.

By simply moving these domains to another registrar, you’re able to take a normally toxic domain and significantly increase its chances of passing.

Strangely enough, the best result is when you link from the backorder PBN very soon after the registrar transfer, but again, this is probably run-of-the-mill experimental variance.

Anyhow, here’s another win to add to the list.

Theory #9: The Amount of Times a Domain Drops and Changes Ownership Affects its Ability to Pass a Positive Link

Let’s say we have a super promiscuous domain that was first registered in 2008, but had been passed around between 20 who.is owners, hosts, and registrars.

Would this perform worse than a domain from 2015 that has had one single owner before it got to you?

Test Structure

Look at all 158 PBNs that had “passed” so far in our experiments.

Chart a histogram of # of owners, host changes, name server changes, etc vs # of how many domains passed.

From here we’ll be able to see if there’s a trend.

Conclusion: True

change histogram

At first glance, you can immediately see that there is a higher chance of a PBN passing a test if that PBN has changed owners/hosts/nameservers/etc less times.

But what really makes this data more compelling is that typically you’re not able to find that many domains with only a few owner changes.  Most domains that you encounter will have at least 10 changes.

Domains with less of an ownership change history have a higher chance of passing.

Add another win for the list.

Theory #10: You can Trick Google into Thinking there Was No Ownership Change through Nameservers

Jason Duke gave us another bright idea.

What if we were to trick Google into thinking that the nameservers didn’t change since the last ownership, and thus the owner didn’t change.

Test Structure

Typical nameserver setups allow you to place 4 different entries for the nameservers.

  • Nameserver 1
  • Nameserver 2
  • Nameserver 3
  • Nameserver 4

What if we were to set them up where the old two nameservers were the two on top, while the actual two nameservers for our hosts were the two on bottom.

  • Nameserver 1: ns1.oldserver.com
  • Nameserver 2: ns2.oldserver.com
  • Nameserver 3: ns1.ourserver.com
  • Nameserver 4: ns2.ourserver.com

The domains would render correctly since the first two would fail and the system would default to the bottom two.

And hopefully, if Google is naive enough and/or they actually look at nameservers as an indication of ownership change, this might be enough to trick the filter.

We tested this on 20 backorder domains and see if it helped the pass rate.

Conclusion: NA

Results weren’t strong enough to make a hard conclusion.  That said, if you look at the trend, it does seem to help marginally…nameserver

While the results here haven’t been totally compelling for PBNs, I have been using this trick for money sites, started from expired or aged domains with good results.

Theory #11: The Whole Boat: What Happens when we Combine all Winning Theories Together?

We seem to be getting marginal, but not complete fixes, when we implement any of these winning theories:

  • Waiting before linking out
  • Consistent content
  • Using domains with a low amount of ownership changes
  • Moving backorder domains out of toxic registrars

But what happens if we combine all these winning theories together.

What effect does this have on expired, backorder, and auction domain pass rates?

Test Structure

Take 20 domains from each category, apply all the winning techniques, and compare against the baseline.

Conclusion: Pass

the whole boat

Final Thoughts

I’ve said before: “I wouldn’t wish PBN management upon my worst enemy.”

This is why.

PBN management doesn’t just involve sourcing, building, maintaining, and renewing thousands of domains.

It also involves testing to be able to stay ahead of the curve.

Ultimately, we found that you’ll find the best results from:

  • Sticking to auction domains
  • Waiting before you link out
  • Keeping at least some content consistent from the old domain to your new rebuild
  • Preferring domains with a low amount of ownership change

But even with these guidelines, be prepared to have some domains that don’t give any result, or even some that produce a negative result.

Although I’m in no way affiliated with Rank Club, I’d still highly recommend going with a vendor that does their own testing for you.

I can’t spill the beans on what they’re up to now, but Rob (R&D Director) has shown what they’re testing and its very, very advanced.

Perhaps for another time…

Share
Tweet
Share