Backlink profile check, worked through on invented numbers
Everything below is a hypothetical example. The shop, the rows and the peers are made up so the method is visible. None of it is real client data.
Anchor Social research team ·
This is an example
The shop below does not exist and the figures were chosen for clean arithmetic. Use the layout, not the values. Pull your own rows with the runbook.
Step one: 600 anchors, bucketed
Hypothetical store selling bakery equipment. Row counts after removing the shop's own footer links.
Bucket
Rows
Share
Brand
312
52%
Bare URL
150
25%
Generic
72
12%
Partial match
30
5%
Exact match
24
4%
Image or empty
12
2%
Reading the table
Brand dominates. That is typical for a small store whose links come from directories, supplier pages and a few local articles. Nothing here looks engineered.
Exact match sits at 4%, 24 rows out of 600. Hold on. How many linking domains produce those 24? If a single trade forum accounts for 20 of them, the real figure is closer to four domains, and four domains is a very different story.
Now count by domain. Suppose the 600 rows come from 410 domains and the brand share by domain is 58%. Close to the row share, so no single site is skewing things.
A last check: sort the exact rows by linking domain and read the pages they sit on. Five minutes of reading beats an hour of percentages.
Step two: a made-up peer set
Three invented competitors, same bucketing rules. Shares of rows.
Site
Brand
Bare URL
Generic
Partial
Exact
Our hypothetical shop
52%
25%
12%
5%
4%
Peer A (bigger, 8 years old)
61%
20%
9%
6%
3%
Peer B (review-heavy)
38%
18%
10%
17%
14%
Peer C (marketplace seller)
47%
29%
14%
6%
3%
What the comparison says
Peer B is the outlier, with 14% exact match. Is it ranking because of that, or despite it? The table cannot say.
Three of four profiles live between 3% and 4% exact match, which makes our shop ordinary. The honest conclusion is dull: nothing to fix in the mix, and a reason to look at content, page speed or internal links if rankings lag.
A different hypothetical would end otherwise. At 22% exact match on 600 rows with no peer above 5%, you would check for paid placements, read the troubleshooting table to rule out a pull error, and only then decide what to disavow or leave alone.
What you can copy from this example
Report row share and domain share side by side
Always include at least three peers
Write the verdict in one sentence per bucket
Name what the data cannot tell you
Save the table as a dated file for next month
Questions about the example
Why 600 anchors and not all of them?
Because 600 is enough to see a pattern and small enough to tag by hand in an afternoon. A real pull can run to 50,000 rows through pagination, which costs 100 credits, or one cent.
Is 4% exact match safe?
There is no official safe number, and nothing in this example claims one. The point is that 4% sits inside the range of three comparable sites, which is a more useful test than any universal figure.
Which Google update matters for this?
Penguin 4.0, September 2016, made link signals real-time and largely devaluing. The December 2022 link spam update added SpamBrain to detect unnatural links. Neither publishes a ratio, so peers are the practical yardstick.