Thread by @GrindstoneSEO

SEO case study involving sites ranking highly for uber-competitive terms they appear unsuited for, based on competition, site age, authority, link equity, on-site optimization, content length, TF-IDF, basically outliers by any metric SEO people love to argue over. 1/

Verticals crowdsourced on twitter via this account and on skype/telegram/slack SEO chats, outliers identified and confirmed by me.
7 broad vertical 'buckets' totaling 117 keywords x analyzed the top 10 results =1174 (some front pages have 8-12 organic results, not my problem). 2/

Tried to use the usual suspects ( @ahrefs @majestic Scrapebox ) for automating said research and while they all helped, ultimately I was left with about 60 sites that I had to analyze individually, both on and off site.

Of these ~60 sites, 13 followed a very similar pattern. 3/

These 13 sites covered a range of verticals, including internet services, high volume Ecom terms, cannabis, pharma and adult. I eliminated local results both because of incongruity in consistent data and also because in some local SERPS, all the ranking national sites matched. 4/

Let me say that again. In the most competitive local niches I analyzed, all the big time national brands that showed up on page 1 city after city were utilizing a strategy similar to this.

When the outliers are no longer outliers...but I digress. 5/

Of the 13 matches, 12 still rank. I'll never expose their URLs. The other one managed to screw up their rankings by deviating from the pattern. I likely won't divulge their URL either. They made a lot of money and could fix this.

If that bothers you, hit the back button. 6/

Going to give you the tl;dr version at this point, as there's value in just copying this:

Relevant links (Only)
Citations
Brand anchors
Limit link loss
Relevant authority links >
Relevant low authority links move the needle but not as much.
Don't rent links, link loss :( 7/

If you hit the back arrow after that last one, your loss. There's a lot more "how" to this than "what".

I realize that I'm going to struggle to explain it well in this format, but I was struggling to create a blog post that did the same. Cest la vie. 8/

But before that, we should probably discuss "why".

Ranking a new site within six months while building links for less than 90 days? With 1/5th to 1/10th the link equity of your competition?

And that site being worth $10k+/day easily. 9/

No, this is not my site. But I know this vertical intimately and tbh, $10k/day is probably an insult. Many days were likely north of six figures.

And they stayed up there for 2 months before things went wrong on them.

I'm pretty sure I've identified how to not go wrong. 10/

Not sure how I got this far without mentioning @serpwoo but they were instrumental in being able to map links to SERP response. Thank you for existing.

ON that same note, I tried to use @ahrefs for all my link mapping but sadly, their "new links" date data is sorely lacking 11/

Not only do their link acquisition dates vary dramatically from what Google shows me via other methods but @ahrefs is literally reporting scraper site links before the high value content, sometimes a week or more before the high value links showed up.

Messy data is bad data. 12/

I still think @ahrefs is a steal for the price and love them. They were instrumental in quickly allowing me to isolate the 60 potential matches that eventually got filtered to 13 matches.

Every tool has value but to effectively map links to SERP changes had to switch gears. 13/

Having to double back and check things via Google advanced queries was actually super valuable because I realized then that our outlier of outliers (the 1/13 not still dominating) had been (likely) renting links and their removal is one of two factors that led to decline. 14/

I used the following query for Google, limiting results to one week chunks.

domain.com -site:domain.com

July 1st-7th, 8th-14th, etc

Screen scraped the results (Link clump and Ginfinity) because of data integrity issues. 15/

Took the screen scraped data, reversed it because when you sort by date it's most recent to oldest and who formats data like that?

Took the now chronological data and used Scrapebox to determine first if link existed and then again if non linked citation existed. 16/

If no link present and no unlinked citation present, feel that it's safe to assume that any result from the query was once a link that no longer exists.

Tried to disprove this theory by doing same for 2 sites still ranking strong, they had very few 'missing' links/citations. 17/

I have to back up. On our chronologic link maps, The Google result from the operator used almost 100% matches the publication date. If no publication date, inferring it's actual indexation date.

Sadly, we're dealing with a black box and not all variables can be eliminated. 18/

Ok, so mapped out links from earliest to latest, and then created 4 columns: +3, +7, +14, +21.

Into each column went the ranks according to @serpwoo 3, 7, 14, 21 days after link showed up in Google.

We don't know link response time but graphing it out should help. 19/

Time to classify links. I went with:

High Authority Domain Relevant - HADR
High Authority URL Relevant - HAUR
High Authority No Relevance - HANR

Low Authority Domain Relevant - LADR
Low Authority URL Relevant - LAUR
Low Authority No Relevance - LANR 20/

Examples: High Authority Domain Relevant would be a link on Avvo article for a lawyer site, High Authority URL Relevant would be in an article about lawyers on the NYT, High Authority No Relevance would be a link in a roundup with no contextual reference to the legal field. 21/

Low authority examples would be exactly the same, except the sites don't have any real authority.

I used @ahrefs again for DR as authority but cross referencing vs @majestic TF was almost a perfect match with regards to bucketing. Moz users can assume whatever they want. 22/

So now we have link drop date (or index date), ranks for 4 time frames following and link classifications.

Surprise, surprise. HADR links are the holy grail of SERP movement. Bet nobody saw that coming. :P

Digging in, we start to see some very interesting patterns play out. 23/

HAUR links are the next biggest mover over these multiple month sample size, followed by LADR.

What's interesting though is that multiple LADR links acquired in a short time frame exceeds the SERP movement of a single HAUR.

Think 2-3 or 3-4. You following? 24/

HANR were surprisingly ineffective at moving the SERP needle while LAUR were similarly ineffective.

And Google's claim that they have figured out how to discount low value links looks to bear some truth, as over hundred of isolated link drops, not much movement either way. 25/

Now the only reason that any of this is relevant to anything is because all 13 sites are following one simple rule: No exact match anchors.

That's right. They aren't building ANY anchors that match the terms they're ranking for. Branded variations, yes. Exact, no. 26/

That's not entirely true... both the no longer ranking outlier and the still ranking outliers have a smattering of exact match anchors...with distinctly negative results EVERY SINGLE TIME they used them...results that slowed their progress for weeks. 27/

And their branded anchoring is very interesting (and systematic) too. It's not brand + keyword (except in rare instances) or keyword + brand.

It's Brand + partial or Brand + phrase, depending on the structure of ranking keyword.

Yeah, I know. Makes no sense. 28/

Ok, we're back on schedule, I'm going to try and push the rest of this out tonight, no promises.

So back to 27/, this was eye opening to me, to say the least. I assumed that anchored links on hyper relevant sites would be fine, and I know other SEO who swear by them. 29/

It seems that the loophole that exists is that combining hyper relevant link sources along with nothing that can be considered over optimization, at least from the link anchoring standpoint, is some powerful mojo kryponite. Prob shouldn't post that but whatever, here we are. 30/

Wanting to make sure I wasn't affirming the consequent, I mapped out all the drops these sites experienced on their way up (note: I only had access to @serpwoo data for 7 of the niches so some extrapolation exists) and every single one can be tied to 2 simple factors. 31/

1. Anchored exact match links provoked an almost immediate loss in rankings, sometimes within 24 hours and had lasting impacts even as the sites continued to acquire links that matched original paradigm.

Feels like a soufflet situation, one ingredient out of ratio and boom. 32/

I won't pretend to understand what combination of algo knobs this is tweaking but it definitely exists. My working theory is relevant sites are obviously relevant and when you add the anchor text, it creates an over optimization score on that link. Who knows? Not me. 33/

2. Link loss is the 2nd factor, and again, there's some uncertainty here. I 'think', and I'm early in verifying this angle, that the loss in ranks matches up with a gain in ranks from when you first acquired the link as opposed to a negative signal 34/

That also fits in well with the "low quality links don't hurt" aspect of them not even being scored best I can determine. If low quality links don't hurt, they can't hurt you when you lose them. But losing high quality links...shit, I just affirmed the consequent. Oops. 35/

Digest that while I look at some notes. I don't want this to be any more difficult to understand than it needs to be. 36/

Ok, our no longer ranking outlier site. After enjoying a two month run of probably fantastic earnings, one of their affiliates dropped 5 anchored links on LADR sites. Think autoblogs with scraped industry news stories. Each anchor was exact match to the big term ranked top 5. 37/

They overnight went to middle of page 2. I remember the day it happened. I didn't know why yet but I knew something had changed. Bounced around page 2 for a bit, then page 3 with some intermittent dips to page 4/5 before recovering to page 2/3 range. 38/

Sometime around the middle of this bouncing around period, they lost a HADR link. Probably their 2nd best link. A pay to play link with a hefty monthly rental, most likely. No longer cost effective paying for it ranking on page 3ish. Boom, page 6 within days. 39/

Shortly after that, they lost another HADR link, also likely a rental. I've inquired into some of these links as I'm in this niche too and the price tag is pretty ridiculous. Anyway, within another ~7 days, they were down to the middle of page 8, where they reside. 40/

What's interesting about this, is that give or take a spot or few, the drops they received when they lost the links pretty much mirrors the gains the links gave them when acquired. Which lends further credence to the theory that it's not a negative signal, you lose the juice. 41/

Also interesting is that the better the links, the quicker the SERP response, in both directions. Whereas lower quality links seemed to take 10-14 days to see movement in those columns, the higher quality links were all evident in the 3 and 7 day columns, both up and down. 42/

Or virtually no movement at all, with regards to low authority non relevant links, even in the 21+ day window, except when obviously influenced gains in the shorter time frames from later acquired links. That reads funny but idc, it is what it is. 43/

Almost forgot, another 'weird' pattern I found that exists. See 28/ for details. Strangely enough, when these sites get off that pattern and use Brand + keyword exactly, it's a mini mirror of the EMD anchor situation, but less extreme and shorter duration. 44/

Right, I'm running out of steam and first chair looks tasty tomorrow so I'm going to admit defeat in getting all this out tonight. Leaves something for the day crew to digest.

Any agencies looking to improve their link systems, I'm available. grindstone@grindstoneseo or DM 45/

Refering back to tweet 7/ we've unpacked most, at least the basics I'm willing to discuss publicly. One topic that I didn't do enough justice too were the citations, raw URL and brand only links. Leveraging these principles prob difficult if leveraging exact match domains. 46/

One of the most common principles amongst these sites are the early focus on unlinked citations, raw URL links and brand only links. All of them built a base over the first 30-60 days on what can best be described as traffic generation, not SEO. Yes, you can manipulate this. 47/

It appears that while niche relevance isn't the focus for this part of the campaign, a higher priority is placed on authority and mostly just establishing a brand. This makes a lot of sense for the link acquisition periods that follow. 48/

Rail riders (and grey hats) know that links without traffic often time lead to quick trip the sandbox. Going for the reverse early makes a lot of sense, from a schematic standpoint. Again, highly manipulatable with a little bit of planning. 49/

Link velocity during this break in period is pretty low, probably by design and fits in when viewed from the wide angle view of avoiding ALL potential over optimizations. Citation/traffic activity is more aggressive, prob a lot to be learned there. 50/

Once this base was laid down over the first 1-2 months, the linkbuilding campaigns begin. It appears the focus is to build a few HAUR and multiple LADR links first. Makes sense from an economic standpoint, why pay the big bucks for the HADR links before you test the waters. 51/

From the limited @serpwoo data I have for this case study, it looks like these sites started springing for the more expensive HADR links once they hit page 4 or so, often combined (by chance?) with multiple drops of LADR links. This recipe appears to be a page 2 coctail. 52/

From there it's just more of the same, bouncing between a couple HADR and more LADR until page 1 rankings were achieved. The sites that are still effective continue to build both, velocity seems to be consistently inconsistent. Link age data specious at best, block of salt. 53/

Just deleted a tweet to fix a typo and twitter dot com broke my tweetstorm. Nice GUI, so thankful we don't have an edit button. 54/

Backing up slightly, and addressing velocity, after our low break in period, overall monthly link velocity ramps up slightly each month but appears to come in bursts. If by design or not, I do not know. Appears sites that were more aggressive bounced more and took longer. 55/

I think that about covers it. There's plenty more but a lot of it I don't want discuss publicly because I've been leveraging these concepts on my sites since October when I started testing and it's going well. Built out the infrastructure to scale each step in the process. 56/

Because if you're not scaling, you have a job. People with jobs will be offended by that, all the entreprenuers are like *fistbump*.
Any agencies who are struggling to get their clients sustainable SEO results should reach out, be happy to consult or use my infrastructure. 57/

Hopefully this was helpful, if you have questions, ask away. If I can answer, I will.