The 15-Day Cliff
Deliverability, measured
Why duration, not depth, decides whether a degraded inbox recovers. The finding behind the Reputation Recovery Playbook, drawn from 71,581 real events.
The cliff, in three numbers
7 to 34 days to recover if you act within two weeks
57 to 74 days to recover once an inbox passes 15 days degraded
~51% recovery probability past the cliff, versus 75 to 80 percent for short events
In this article
- The question: does depth or duration decide recovery?
- The recovery table, and how to read it
- Depth barely matters
- Duration is the cliff
- Why early intervention beats damage control
The question: does depth or duration decide recovery?
When placement drops, two things are true at once: the drop has a depth (how far the inbox fell) and a duration (how long it stays down). Intuition says depth is what should scare you. A 40-point collapse feels like an emergency; a 10-point dip feels survivable. The data says intuition has it backwards.
We walked the day-by-day inbox-rate history of about 117,000 inboxes across 3.7 years of the Warmup Inbox network and detected 71,581 degradation events. An event qualifies when the rate stays at least 5 points below its trailing 14-day baseline for at least three consecutive days, with an initial drop of at least 10 points. Recovery is confirmed when the rate returns within 2 points of baseline for at least seven consecutive days. Then we sorted every event by depth and duration.
The recovery table, and how to read it
| Time below threshold | 10 to 20pp drop | 20 to 40pp drop | 40+ pp drop |
|---|---|---|---|
| 3 days or less | 9 days (81%) | 8 days (76%) | 7 days (80%) |
| 4 to 7 days | 16 days (73%) | 14 days (72%) | 13 days (76%) |
| 8 to 14 days | 34 days (62%) | 27 days (64%) | 21 days (70%) |
| 15+ days | 74 days (52%) | 74 days (51%) | 57 days (51%) |
Source: Warmup Inbox network, 3.7 years, 71,581 detected events. Each cell shows median days to recover and, in parentheses, the share that recovered. 33% of events are censored and reported honestly.
The grid has two axes on purpose. Rows are duration, columns are depth. If depth drove recovery, the numbers would change most as you move left to right. If duration drove it, they would change most as you move top to bottom. Only one of those is true.
Depth barely matters
Read across any row. In the top row, a mild 10-point drop recovers in 9 days and a severe 40-point drop recovers in 7. The deep drop actually comes back slightly faster, because a sharp, short shock is often a discrete event that resolves cleanly. Across every row, moving from the mildest to the most severe column barely moves recovery time. A severe drop caught quickly is not a crisis.
Duration is the cliff
Now read down any column. In the 10-to-20-point column, recovery goes from 9 days (caught in 3 days) to 16, to 34, to 74 days once the event passes 15 days. That is the cliff: recovery time roughly triples between the 8-to-14-day band and the 15+ band, and the recovery probability drops from the 60s and 70s down to about 51 percent.
Fifty-one percent is the number to remember. Once an inbox has been degraded for 15 or more days, whether it ever recovers is close to a coin flip, and if it does, it takes roughly two months. This holds no matter how deep the original drop was. The 40-point, 15-day event and the 10-point, 15-day event both land at about 51 percent and both take around 57 to 74 days.
For balance, most events never reach the cliff. Across all severities, 67 percent of detected events recovered, with a median recovery time of 20 days. The cliff is not the common case, it is the expensive one, and it is entirely a function of how long a problem is allowed to persist.
Why early intervention beats damage control
This single finding rewrites the priority order for deliverability. The instinct is to triage by severity: panic about the deep drops, shrug at the shallow ones. The data says triage by age instead. A shallow drop that has been sitting for twelve days is more dangerous than a deep drop that happened yesterday, because the shallow one is about to cross the cliff and the deep one is still in the fast-recovery zone.
- Detect fast. You cannot beat a 15-day clock you did not know was running. Gmail is your earliest signal, spiking first in 85 percent of events.
- Prioritize by duration, not depth. Sort open issues by how long they have been degraded. The oldest ones, not the deepest, are the emergencies.
- Act inside the first two weeks. Recovery in the top two rows runs 7 to 34 days at good odds. That is the window where effort pays off.
- Know before you commit. The recovery predictor tool models your exact position on this table from your drop depth and duration.
The cliff is the reason the Reputation Recovery Playbook is organized around a clock rather than a severity scale, and the reason Gmail early-warning monitoring matters so much. Catch the drop before day 15 and the odds are firmly on your side.
Related reading
- The Reputation Recovery Playbook
- Try the recovery predictor tool
- Gmail is the canary AND the harshest filter
Frequently asked questions
What is the 15-day cliff?
It is the point where a placement drop becomes hard to reverse. Events resolved within two weeks recover in 7 to 34 days at 60 to 81 percent odds. Once an inbox stays degraded for 15 or more days, recovery time jumps to 57 to 74 days and the odds fall to about 51 percent, regardless of how deep the drop was.
Does the depth of the drop not matter at all?
It matters far less than people expect. Across the data, a 40-point drop caught in three days recovers about as fast as a 10-point drop of the same duration, roughly a week. Duration, not depth, is the main driver of recovery time and probability.
How many drops actually reach the cliff?
Most do not. Across all severities, 67 percent of detected events recovered, with a median recovery time of 20 days. The 15+ day cliff is the minority case, but it is the expensive one, and it is caused by letting a problem persist.
How do I use this in practice?
Prioritize open placement issues by how long they have been degraded, not how deep they are. An old shallow drop is closer to the cliff than a fresh deep one. Detect early using Gmail, act within two weeks, and use the recovery predictor to see where you sit on the curve.
From the Warmup Inbox network. Recovery analysis based on 71,581 detected degradation events over 3.7 years. 33 percent of events are censored and reported rather than excluded.
