How fast do signals propagate through reputation models?

Faster than most senders think on the way down, slower than most senders hope on the way up. That asymmetry is the most important thing to understand about reputation models.

The down direction (hours)

If you light up Spamhaus with a few spamtrap hits, or your Gmail complaint rate jumps over 0.3 percent in a single send, you'll see filtering changes within hours. Sometimes within the same day. Providers prioritize reacting to abuse because the cost of being slow is recipients getting more spam.

Hard bounces and authentication failures register almost in real time inside the provider, even if your ESP dashboard takes a day or two to surface them.

The up direction (days to weeks)

Building positive reputation is a confidence-interval problem. One great campaign doesn't move the needle much. Providers want to see consistency. A clean week of engaged sends starts to register. A clean month meaningfully shifts placement. A clean quarter resets the baseline.

This is why people coming out of a deliverability incident often feel stuck. The bad signals propagated in 24 hours, but the recovery curve runs 30 to 90 days even when they're doing everything right.

Per-provider differences

• Gmail: machine-learning driven, reacts to engagement shifts within days. Rewards consistency hard.
• Outlook / Microsoft 365: slower to recover than Gmail. Their SNDS dashboard updates daily.
• Yahoo / AOL: aggressive on complaints, slow on rehabilitation.

What to do

Treat reputation like a savings account, not a faucet. Don't try to "fix" a bad week with a big push. Slow down sends, tighten the segment, fix the underlying problem (list quality, content, cadence), and let the model see consistency. If you want the picture in real time instead of waiting for your next campaign to find out, SOS is free.