A site that earns four referring domains a month for two years and then eighty in a fortnight has told Google something. It may have told it something true — a funding round, a launch, a piece of research that landed. It may have told it something else. The point is that it said something, and it said it loudly.
Link velocity is one of the most argued-about concepts in this discipline and one of the least usefully discussed. The argument is usually framed as "is there a safe number of links per month", which has no answer. The useful question is narrower: what makes a given rate defensible?
There is no safe number, and that is not evasion
Any absolute rate is wrong because the same number means different things on different sites. Twenty new referring domains in a month is unremarkable for a company that just closed a Series B and got covered everywhere. On a two-year-old site averaging three, it is an event.
What matters is the shape of the curve relative to the site's own history and the visible reasons for change. A search engine has your entire acquisition history, your publishing history, and the content of the pages linking to you. It is not comparing you to a global threshold; it has a much better comparison available, which is you last quarter.
So the honest formulation is: velocity is a signal in context, and the context is what you control.
What makes a ramp defensible
Four things, and they are the same four that make a ramp genuinely earned rather than merely appearing so.
Something visible caused it. Funding announcements, product launches, original research, a conference talk, an acquisition. If a spike coincides with an event that produced coverage, the coverage explains the spike. If the spike has no event behind it, the only available explanation is procurement.
The sources vary. Eighty domains from eighty different kinds of publication, in different countries, with different link patterns, is what real attention looks like. Eighty domains sharing a hosting range, a template and an outbound profile is one supplier with eighty properties, and the pattern is not subtle at scale. This is the check that catches more bad inventory than any velocity threshold would.
The anchors look like writing. A real spike produces mostly branded and URL anchors, because that is how people cite. A purchased spike produces partial and exact-match anchors, because that is what was ordered.
The page attracting them can carry it. A spike pointed at a research page, a launch page or the homepage reads differently from a spike pointed at a commercial comparison page that has not changed in a year.
The ramp we actually plan to
For a B2B SaaS programme starting from a low base, the shape that has held up across engagements looks like this.
Months one to two: below target. Not because of caution — because of arithmetic. Month one produces a candidate list and pitches, not placements. The median time from first contact to a live URL is around a month, so month one ships nothing and month two ships roughly half a normal month as the first replies land. Any plan that promises full volume in month one is selling inventory it already controls.
Months three to nine: steady at target. The flat part, and the part that does the work. Four to twelve a month depending on how deep the reachable list is.
Beyond month nine: flat or declining, deliberately. In most B2B SaaS categories the reachable publication list is finite. If you are still acquiring at full volume in month fourteen, either the category is unusually deep or the quality bar has quietly dropped to keep the number up. The second is far more common, and it is the point at which programmes start hurting the profile they built.
The failure mode nobody plans for
It is not the spike. It is the cliff.
A programme runs at ten a month for eight months and then stops, because the contract ended or the budget moved. The acquisition curve goes to near zero overnight. That discontinuity is at least as informative as a spike — a site that was earning steadily and abruptly stops was not earning, it was buying.
This is a good argument for tapering rather than stopping, and a better argument for building something during the programme that keeps earning after it. A benchmark, a dataset, a tool: the thing that produces links with nobody pitching. One engagement we ran ended with the retainer cut in half and a single annual benchmark carrying the acquisition instead, and its curve after the programme ended looked healthier than during it.
How to check your own curve
Pull referring domains by month for the last twenty-four months from any backlink tool. Then look for three things.
Any month more than three times the trailing median — and ask what caused it. If you cannot name the cause, find out. Sometimes it is a syndication event nobody knew about, which is fine. Sometimes it is a supplier delivering a batch, which is not.
Any period where the sources cluster: same registrar, same hosting range, same template, similar outbound patterns. This is visible across the batch and invisible link by link, which is exactly why per-link vetting misses it.
Any cliff. Where the curve drops from steady to near zero, and whether anything was built during the steady period that would have carried on.
What to do with the answer
If your curve is smooth and matched to visible activity, velocity is not your problem and you can stop thinking about it. If it has an unexplained spike in the past, the remedy is dilution rather than panic — steady, varied, branded-heavy acquisition afterwards. If it has a cliff, the lesson is for the next programme.
And if you are planning a programme now, set the monthly rate at something you can sustain for the entire period rather than something you can hit in a good month. A rate you have to abandon halfway produces both of the shapes you were trying to avoid: a ramp and a cliff, in the same twelve months, with nothing built to carry the difference.