You optimized the ratio and ignored the velocity

You optimized the ratio and ignored the velocity

A persistent blind spot: teams obsess over the steady-state anchor ratio and never model acquisition velocity. Two profiles can share an identical 4% exact-match lifetime ratio, yet one acquired those exact-match links smoothly over 18 months and the other in a single 10-day burst. Penalty analysis of the second is materially worse.

The mistake treats anchor risk as a static snapshot when the relevant signal is a time series. Spam-detection literature and patent filings around link-spam (e.g. trust/velocity-based dampening) repeatedly emphasize temporal anomaly, not just final proportion.

The fix:
— Plot exact-match anchor acquisition as a weekly time series, not a single percentage
— Flag any week where commercial anchors exceed ~3x the trailing-12-week median
— When recovering from a velocity spike, do not stop link-building entirely — a hard stop is itself an anomalous pattern; taper

Limitation: link-index discovery dates lag real acquisition, smearing your time series; treat the curve as directionally indicative.

Open question: is anchor velocity evaluated in absolute link counts or normalized to the domain's baseline acquisition rate, such that established sites tolerate larger absolute bursts?