A benchmark is a measurement with a sample and a date. This recipe makes both explicit, uses three API calls, and ends with a sentence you can defend in a meeting.
You need your domain, ten rivals, an API key from the Google sign-in, and a sheet with a date column. That is all.
Choose the rivals before you look at any numbers. If you add or drop one after seeing the data, you have stopped measuring and started decorating.
Replace the domains. The first call is free; compare costs 1 credit, and refdomains 2.
# 1. profile size (free)
curl "https://linkbuildingstat.com/api.php?action=overview&q=mysite.com"
# 2. everyone at once (1 credit)
curl "https://linkbuildingstat.com/api.php?action=compare&q=mysite.com,r1.com,r2.com,r3.com&key=sk_live_..."
# 3. top linking domains for the odd one out (2 credits)
curl -H "Authorization: Bearer sk_live_..." "https://linkbuildingstat.com/api.php?action=refdomains&q=r3.com&per=100"| Site | Referring domains | Notes |
|---|---|---|
| mysite.com | 140 | Ours |
| r1 to r4 | 60, 75, 90, 110 | Low half of the sample |
| r5 to r7 | 150, 180, 210 | Upper middle |
| r8 to r10 | 400, 900, 6,500 | One giant, two large |
Do not use the mean. In the invented set the rival mean is about 870, dragged up by a single giant, while the median of the ten is 165. We sit a little under that median.
So the honest sentence is that we are slightly below the middle rival, not that we are 84 percent behind the average. The second sentence is more dramatic and says less.
Next, run the free domain lookups for the giant, and read the Ahrefs cost page if you wonder what a monthly run costs. If it is a portal with a different business model, remove it from the next month's sample, and write down why.
Next month, rerun the same calls on the same sample. Note any change, and decide in advance how large a change you will treat as real; for referring domains, a shift of a few percent is easily explained by one refresh.
Keep every reading. Three data points on a line say more than a clever chart built from one, and they are the cheapest way to see whether your own link work moved anything.
If you later buy placements, the lost-links routine uses the same sheet.
Ten is not magic. It is the point where one odd rival stops deciding the answer, and where a sheet still fits on one screen.
With a small sample, the median and the range say more than any single figure. In our invented table the range runs from 60 to 6,500, which tells a reader at once that the group is mixed and that the giant should be treated separately.
A rough interquartile view helps too: the lower quartile of the ten is about 90 and the upper about 400, so the middle half of rivals sits in a band several times wide. Being 25 below the median is a small gap inside a band that wide.
Yes. The compare action accepts up to 20 domains for 1 credit. Past ten, scan for outliers first, because one huge domain distorts a chart long before you notice it in a table.
About monthly, because that is how often the index refreshes. A weekly rerun mostly measures noise from partial updates, and it will tempt you into reading changes that are not there.