A perfectly authenticated message can still end up in spam. Authentication proves who sent an email; reputation tells filters how far that sender can be trusted. Mailbox providers build this assessment by watching, over time, the behaviour of two main identities: the IP address messages come from and the domain that signs them and appears in From.
Two identities, two histories
IP reputation
This is the oldest form. An IP address that sends unwanted mail ends up on blacklists and is penalized by filters; one that sends welcome messages builds up credit. The limitation is obvious: an IP can be changed. Spammers have always had an incentive to move to fresh addresses, and legitimate senders may find themselves sharing an address with others.
Domain reputation
With the spread of DKIM and DMARC, providers became able to attribute messages to a domain reliably. A domain is a much more stable identity than an IP: it carries its history with it even when you change provider or server. That is why domain reputation carries growing weight today, and in many cases more than IP reputation.
The practical consequence matters: switching sending platforms does not reset your problems. If a domain has a bad reputation from sending to poor-quality lists, it will carry it to the new platform too.
What reputation depends on
Providers don't publish their algorithms, but the main signals are well known and consistent:
- spam complaints from users, the strongest negative signal;
- non-existent addresses: a high bounce rate indicates old or poorly collected lists;
- spam traps: addresses nobody uses, which exist only to catch senders mailing without consent;
- engagement: messages opened, read, moved from spam to the inbox, replied to;
- consistency: steady volumes are more trustworthy than sudden spikes;
- correct authentication and coherent infrastructure, with sensible PTR records and server names.
Engagement deserves a clarification. It is less about the open rates measured by your platform, which are increasingly unreliable, and more about the behaviour the provider observes directly in its own mailboxes.
Shared or dedicated IP
Sending platforms usually offer two options.
| Aspect | Shared IP | Dedicated IP |
|---|---|---|
| IP reputation | shared with other customers | yours alone |
| Volume needed | suited to low and medium volumes | needs steady, substantial volume |
| Warm-up | not needed | essential |
| Risk | other people's behaviour affects you | your mistakes weigh only on you |
| Control | limited | complete |
A dedicated IP is not automatically better. With few sends a month, providers don't have enough data to form an opinion and the IP stays «unknown», which doesn't help. In those cases a shared pool, well managed by the platform, often performs better. A dedicated IP makes sense when volumes are regular and high enough to build a history of their own.
Separate to protect
A good practice, regardless of the IP choice, is to separate mail streams with different reputations. Transactional emails, such as order confirmations and password resets, have high engagement and almost no complaints. Promotional newsletters naturally have weaker metrics. If both use the same domain and the same IP, a campaign problem can delay receipts.
Separation is done with different subdomains for From, or at least for the DKIM signature and Return-Path, for example news.example.com for campaigns and account.example.com for notifications, and where possible with different IPs or pools. The main domain's reputation remains linked, so separation reduces risk rather than eliminating it.
How to observe your own reputation
The main providers offer tools for senders: Gmail Postmaster Tools shows the spam rate and the domain's compliance status, Microsoft offers the SNDS service for IP addresses. Add to these checks against public blacklists and, above all, the trend in your own metrics: a sudden drop in opens at a single provider is often the first sign that messages are landing in spam right there.
In my work, when a client reports a drop in delivery, the first question is not about technical configuration but about recent changes: a newly imported list, a jump in volume, a new type of content. The cause is almost always there.
In short
- Authentication identifies the sender; reputation determines how much filters trust it.
- Domain reputation carries ever more weight and follows the sender even across platforms.
- Spam complaints, bounces and spam traps damage it; engagement and consistency build it.
- A dedicated IP pays off only with regular, substantial volumes, and it requires warm-up.
- Separating transactional and promotional mail with different subdomains limits the spread of problems.
