IP reputation can be rented and replaced. Domain reputation cannot. How providers weight the two, what survives a migration, and where reputation spending is misallocated.
Senders talk about reputation as one thing and providers track at least two. The IP address that opened the connection has a history, the domain in the From header has a separate history, and the two can point in opposite directions for the same message. Which one a given provider weights explains outcomes that otherwise look arbitrary: a brand new IP delivering fine because the domain is trusted, or a pristine dedicated IP delivering badly because the domain it carries has been burning goodwill for a year. Our note on shared and dedicated pools covered the IP side of that choice. This covers what the choice does not buy you.
Two identifiers, two histories
An IP address is infrastructure, and it is rented. It can be replaced in an afternoon, shared with strangers, or handed back to a provider and reassigned to somebody else next month. Receivers know this, which is why IP reputation builds fast, decays fast, and carries the least weight for any sender with domain history to draw on. A domain is identity. It appears in the From header the recipient reads, it survives every infrastructure change underneath it, and it cannot be swapped without abandoning the brand equity attached to it. Google's own guidance has been explicit for years that domain reputation dominates its evaluation, and the practical test is simple: move to a new IP and observe how little changes.
Where each one is measured
The asymmetry shows up plainly in the tooling. Microsoft's SNDS is organised entirely around IP addresses, reports per-IP complaint rates and trap hits, and offers no domain view at all, which reflects a filtering stack still weighting connection-level history heavily. Gmail's Postmaster Tools reports both, and the domain reputation graph is the one that moves placement. That difference alone explains why the same sending change produces a visible Microsoft response and a muted Gmail response, or the reverse, and it is a large part of why a single averaged strategy across providers underperforms a per-provider one.
What the weighting changes operationally
Decisions that follow from it
- 1
Warm the domain, not only the IP
A plan that ramps volume on a new IP while the From domain has no history is warming one identifier and neglecting the more important one. Ramp both together and treat domain history as the slower of the two clocks.
- 2
Do not rotate domains to escape a problem
Domain hopping works briefly, is trivially detectable as a pattern, and trains receivers to distrust the whole family including the parent. It converts a recoverable reputation problem into a structural one.
- 3
Separate streams by subdomain before separating them by IP
Subdomain separation costs a DNS record and delivers most of the isolation benefit. Dedicated IPs cost volume commitments and warming time, and deliver less than most senders expect.
- 4
Keep the From domain stable across ESP migrations
Providers and MTAs change; the domain should not. A migration that preserves the domain carries most accumulated trust across, which is the single biggest lever for a smooth cutover.
- 5
Authenticate with your own domain everywhere
A provider-signed message earns reputation for the provider. Signing and aligning with your own domain is what converts sending volume into history you keep.
“An IP address is rented infrastructure. A domain is the only sending asset you actually own.”
The practical summary is that infrastructure is replaceable and identity is not, and that most reputation spending is misallocated toward the replaceable half. Teams buy dedicated IPs and warming tools while sending from a domain with no authentication history and no stream separation, then attribute the disappointing result to the provider. Spend on the domain first: authenticate it, segment it, and protect its history as the asset it is. The IP is a lease.
Frequently Asked Questions
Does a dedicated IP improve domain reputation?
How long does domain reputation take to recover?
Should each stream get its own domain or its own subdomain?
Do providers share reputation data with each other?
Key Takeaways
- Providers track the connecting IP and the From domain separately, and the two can disagree about one message
- IP reputation is fast, rented, and replaceable; domain reputation is slow, owned, and survives every migration
- Microsoft's tooling is IP-centric and Gmail's weighting is domain-centric, producing divergent responses to the same change
- Subdomains inherit standing downward and contain damage upward, making them the cheapest isolation available
- Domain hopping to escape a reputation problem converts a recoverable issue into a structural one
Related articles
Half-Year Review: Email in the AI Inbox Era
Six months that rearranged the reading layer: Gemini in Gmail, Microsoft rejecting outright, DMARC finally a standard. What the first half of 2026 means for senders.
Per-Provider Playbooks: Operating Gmail, Microsoft, and Yahoo Differently
The big three score senders differently, expose different data, and fail in different ways. One operating playbook per provider beats one averaged strategy.
Subaddressing: What Plus Addresses Tell Senders
user+anything@gmail.com is one mailbox wearing labels. How subaddressing works, why blocking it backfires, and what plus addresses reveal about your data flows.