Business Email Address Hosting: Options, DNS and Fixes

November 15, 2026 ARPHost Uncategorized

The popular advice about business email address hosting is to compare mailbox size, storage and an uptime percentage. That is the wrong checklist. A mailbox on your own domain, sales@yourcompany.com, is a provider running the SMTP servers that send, the IMAP servers that store and sync, and a webmail front end. What decides whether it works is the part no plan comparison mentions: your SPF, DKIM and DMARC records and the reputation of the IP address your mail leaves from. A generous mailbox on a broken domain bounces invoices; a small mailbox on a clean, authenticated domain delivers. Check authentication first, then pick the hosting model that matches who will maintain those records.

Table of Contents

Business Email Hosting Options Compared

Five delivery models cover almost every business, and they differ in one thing that matters: who patches the mail stack and who answers when a message bounces. Everything else, mailbox size, spam filtering, and webmail branding, is secondary.

OptionWho runs the mail stackBest fitMain risk
Shared hosting mailProviderSmall teams with low mail volumeSending IP reputation shared with other tenants
Self-managed VPS with Postfix and DovecotYouTeams with mail administration skillsPatching, PTR records and blacklist upkeep
Dedicated server or colocationYouHigh-volume or compliance-bound mailHardware and staffing cost
Managed mail on an administered hostProviderSmall teams that want admin offloadedLess direct control over the mail stack
Cloud suite such as Microsoft 365 or Google WorkspaceVendorTeams that want calendar and collaboration toolsDNS and authentication remain your responsibility

The responsibility split that decides your risk

Shared hosting mail is the cheapest path and the one most small businesses start on. The provider runs Postfix and Dovecot, you get mailboxes and aliases on your own domain, and a control panel handles the rest. The risk is that your outbound mail leaves from an IP address shared with every other tenant on that server. One neighbour running a compromised WordPress plugin and sending pharmacy spam gets the whole range throttled, and your quotes start landing in spam with no configuration change on your side.

Self-managed mail on a VPS flips the split entirely. You run Postfix and Dovecot, you own the patching, the reverse DNS record (set at the hosting provider, not the registrar), the port 25 unblock request and the blacklist delisting when it goes wrong. Teams that need it get full control and a dedicated address. Teams that don't need it underestimate the maintenance by a wide margin. If you're going this route, the server build and Postfix configuration steps are covered in our guide to setting up an email server.

Dedicated hardware or colocation earns its place when mail volume is high, when retention rules require hardware you control, or when a clean, exclusively yours sending IP is worth the invoice. A dedicated server with a properly delegated reverse DNS zone and no other tenants on the address is still the most reliable way to send at scale from your own infrastructure. Current configurations are listed on the bare metal servers page, and the same logic applies to colocation if you already own the hardware.

Managed mail keeps the operational burden on the provider while you retain registrar and DNS control. Cloud suites move the mailboxes to a vendor, but the SPF, DKIM and DMARC records stay yours, which is why migrations to a suite break deliverability just as often as migrations between small hosts.

What this looks like in production: on multi-tenant mail infrastructure, delivery failures almost never trace back to storage. They trace to a missing DKIM selector, an SPF record that expired when a vendor changed its ranges, or a reverse DNS entry that points at the previous owner of the IP. Choose the model that matches who will watch those three things at 2 a.m.

Mail Protocols, Ports and Features to Require

Protocols decide which applications work. Get the ports wrong and Outlook connects while your billing system's SMTP relay times out.

ProtocolPortEncryptionUse
SMTP submission587STARTTLSSending from mail clients and applications
SMTPS465Implicit TLSSending where clients require TLS from connection start
IMAP993Implicit TLSMailbox sync across phones and desktops
IMAP143STARTTLSLegacy clients only, avoid where possible
POP3S995Implicit TLSSingle-device download and removal
SMTP relay25STARTTLS optionalServer-to-server delivery only

Point every mail client and every application at port 587 with STARTTLS. It's the submission port, it requires authentication, and it's the one that survives network filtering. Port 465 is still correct for older clients that demand TLS from the first byte. Port 993 is the only IMAP port worth publishing; 143 exists for legacy clients and should be disabled if nothing legitimate uses it. POP3S on 995 downloads and removes mail from the server, so it's for a single-device archive, not for a mailbox three people share. Port 25 is server-to-server relay and nothing else. If a support engineer tells you to put a desktop client on port 25, find another support engineer.

Features worth paying for

  • DKIM signing on outbound mail: essential. Without a signature on the way out, DMARC alignment has only one leg to stand on.
  • IMAP sync with server-side folders: one mailbox on phone, laptop and webmail, with the same sent items. POP3 breaks this by design.
  • Per-user quotas and an admin console: so one person receiving 40 MB of scanned PDFs doesn't take the whole domain down.
  • Catch-all and forwarding rules: useful during a migration, dangerous in permanent use because catch-alls collect backscatter.
  • Custom domains and aliases: sales@, billing@ and accounts@ all landing in one mailbox is normal for a small team.
  • Shared calendars and contacts: if staff book each other's time, CalDAV support matters more than storage.

Verify the certificate before you blame the client

Three commands settle most "the mail app won't connect" tickets.

$ openssl s_client -starttls smtp -connect mail.example.com:587
...
Verify return code: 0 (ok)
$ openssl s_client -connect mail.example.com:993
...
Verify return code: 0 (ok)

If either returns Verify return code: 21 (unable to verify the first certificate) or a hostname mismatch, the problem is on the server: a missing intermediate certificate, or a certificate issued for example.com while clients connect to mail.example.com. Fix it once on the host. Chasing it through each user's client settings wastes a day and fixes nothing.

Publish SPF, DKIM and DMARC Records Correctly

The records are what make a mailbox deliverable, and they live in DNS, not in the hosting panel. Five entries do the work.

A professional man works on a laptop to secure business email with SPF, DKIM and DMARC settings.

Record typeHost nameExample valuePurpose
MX@10 mail.example.comRoutes inbound mail to the mailbox host
SPF TXT@v=spf1 ip4:203.0.113.10 include:spf.provider.example ~allLists the hosts authorized to send for the domain
DKIM TXTselector1._domainkeyv=DKIM1; k=rsa; p=(public key from host)Publishes the key that verifies message signatures
DMARC TXT_dmarcv=DMARC1; p=none; rua=mailto:dmarc@example.comSets the policy for failed messages and the reporting address
Autodiscover CNAMEautodiscovermail.example.comLets mail clients find server settings automatically

The publishing order that avoids blocking your own invoices

Most domains break because someone publishes a strict policy before inventorying senders. The NIST guidance on trustworthy email is explicit that SPF, DKIM and DMARC are complementary: SPF authorizes sending hosts, DKIM proves the message wasn't altered, and DMARC tells the receiver what to do when a message fails and where to send reports. A DMARC pass needs at least one of SPF or DKIM to pass and the authenticated domain to align with the visible From domain. Work in this order.

  1. Inventory every sender. The mailbox host, website contact forms, the CRM, invoicing, the newsletter platform, the applicant tracking system. Export the last 30 days of DMARC reports if you already have reporting.
  2. Publish one SPF record that covers all of them. One TXT record at the apex, no duplicates, no second v=spf1 string left over from a previous host.
  3. Turn on DKIM signing at the host and publish the selector's public key. Use a selector name that means something, such as selector1 or 2024mail, so rotation is easy later.
  4. Publish DMARC with p=none and a reporting address you read. This changes nothing about delivery, it just tells you who is sending as you.
  5. Move to quarantine after two weeks of clean reports, when every legitimate source shows alignment.

Lower the TTL on MX, SPF and DMARC records to 300 seconds a day before any cutover. If you have to back out at 11 p.m., a five minute TTL means the rollback is live before you finish reading the bounce.

$ dig +short MX example.com
10 mail.example.com.

$ dig +short TXT example.com
"v=spf1 ip4:203.0.113.10 include:spf.provider.example ~all"

$ dig +short TXT selector1._domainkey.example.com
"v=DKIM1; k=rsa; p=MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8A..."

$ dig +short TXT _dmarc.example.com
"v=DMARC1; p=none; rua=mailto:dmarc@example.com"

Four answers, four records confirmed from a public resolver. If dig returns nothing where you expect a record, you published it at the wrong name or the change hasn't left the provider's nameservers yet. The record structure and the staged rollout are worth getting right once, which is why our email security best practices walk through the same sequence for new domains.

Diagnose Spam Folder and Bounce Problems

Someone reports that mail to a customer went to spam, or a bounce arrived with a wall of text. There are four causes, and they account for nearly everything. Work down the list in this order.

Cause 1. SPF permerror from too many DNS lookups

Symptom: the receiving server reports permerror, or authentication shows spf=permerror. Every SPF record is correct on its own. Count the mechanisms that cost a DNS query: include, a, mx, ptr, exists and redirect. RFC 7208 caps the total at 10, and exceeding it requires the receiver to return a permanent error. ip4, ip6 and all are free.

$ dig +short TXT example.com
"v=spf1 include:spf.protection.outlook.com include:_spf.provider.example include:servers.mcsv.net include:_spf.crm.example ~all"

Four includes here, but each one nests further includes, and the count is what the receiver tallies, not the number of entries you see. Follow each chain with dig +short TXT spf.protection.outlook.com and keep a tally. The fix is to consolidate: replace nested includes with the explicit ip4: ranges the vendor publishes, and delete senders you stopped using two years ago.

Cause 2. DMARC fails while SPF passes

This is the migration classic. RFC 7489 defines DMARC as an alignment check, not a pass or fail of SPF alone. The receiver validates SPF against the envelope sender and DKIM against the signing domain, then compares whichever passed against the domain the user sees in the From: header. A provider sending as bounce.provider.example with your address in From: can return spf=pass and still fail DMARC.

Authentication-Results: mx.google.com;
       spf=pass smtp.mailfrom=bounce.provider.example;
       dkim=pass header.d=example.com header.s=selector1;
       dmarc=fail header.from=example.com

Read smtp.mailfrom and header.d in a received message. If the envelope domain isn't yours, configure a return path on your own domain or enable DKIM signing with your domain as d=. Anything that passes SPF or DKIM while aligned with your From: domain satisfies DMARC.

Cause 3. DKIM breaks at a forwarder or a disclaimer gateway

A signature that validated on the way out fails after a ticketing system appends a footer, a mailing list rewrites the subject, or an antivirus appliance re-encodes the body. RFC 6376 allows simple and relaxed canonicalization for headers and body, and simple tolerates almost no modification at all. Compare the DKIM result before and after each hop, confirm the selector in the DKIM-Signature header (s=) and signing domain (d=) match a live record with dig +short TXT selector1._domainkey.example.com, then configure the intermediary to re-sign. Sign as late in the chain as possible.

Cause 4. A 5.7.1 rejection that has nothing to do with content

When Gmail or Outlook answers with 5.7.1, check reverse DNS first.

$ dig -x 203.0.113.10 +short
mail.example.com.

$ dig +short mail.example.com
203.0.113.10

The forward and reverse lookups must agree. After that, Google's sender requirements apply to every sender, not just bulk ones: SPF or DKIM is mandatory, and bulk senders need both plus DMARC with alignment. A small business sending 200 messages a day still faces the first rule.

A pattern worth recognising: a domain that accumulates Microsoft 365, a CRM, a newsletter tool and a ticketing system tends to hit the SPF lookup ceiling within weeks of the fourth integration. Consolidate senders onto one relay you control rather than appending another include. If bounce handling and reputation monitoring is the part you keep revisiting, our notes on email deliverability best practices cover the ongoing side.

Size Mailboxes and Plan Backups

Sizing is arithmetic, and anyone can repeat it. Take 12 mailboxes averaging 4 GB today, assume 30 percent yearly growth from attachments, and plan three years out. Each mailbox reaches about 4 GB multiplied by 1.3 cubed, so roughly 8.8 GB. Twelve of those is about 106 GB, and 20 percent headroom takes you to roughly 127 GB. Repeat per group: a shared support queue with screenshots grows faster than an executive mailbox, and a sales team sending 20 MB decks grows faster still.

Quotas and shared mailboxes

Set per-user quotas around 1.5 times the expected three-year size, not the current average. That gives a user a real ceiling without a weekly "mailbox full" ticket. Move functions, not people, into shared mailboxes: support@, billing@ and info@ should be shared mailboxes with delegated access so a departing employee doesn't take the history with them.

Backup tests that catch real failures

Daily backups usually succeed. Restores fail, because nobody tested the restore path or the export missed a folder.

  1. Confirm the backup exists and how many restore points the provider keeps. "We back up daily" means nothing if retention is 24 hours.
  2. Ask how long a single mailbox restore takes. A four-hour restore of a mailbox that receives customer orders is an outage, not a recovery.
  3. Restore into a scratch account monthly, not to the original mailbox.
  4. Compare message counts after the restore with doveadm mailbox status -u user@example.com messages INBOX.
$ doveadm mailbox status -u user@example.com messages INBOX
INBOX messages=48213

If the count is short by a few thousand, the restore skipped a folder or stopped at a quota limit. Keep the pre-migration export for 30 days so a failed cutover can be reversed by restoring the old mailboxes and reverting the MX record.

Read SLA and Support Terms Before You Sign

An uptime percentage is a promise about the mail server process being reachable, and it converts to time you can plan around.

Uptime targetDowntime per yearDowntime per month (30.4-day average)
99.9 percentabout 8 hours 46 minutesabout 44 minutes
99.95 percentabout 4 hours 23 minutesabout 22 minutes

What the percentage actually buys

An SLA measures availability, not deliverability. If your SPF record breaks and every invoice lands in quarantine, the mail server was up the whole time and you get no credit. That distinction matters more than the difference between 99.9 and 99.95, because authentication failures are far more common than server outages.

Read the covered components: mail server, webmail, network path, DNS. Then the credit calculation, the measurement window, maintenance exclusions, and the process for filing a claim. A credit you can't claim because the window closed at 30 days is marketing.

Clauses worth reading in full

  • Maintenance windows: whether planned work is excluded from the uptime calculation, and when it happens.
  • Support hours and channels: ticket, chat or phone, and whether anyone answers at 3 a.m. on a Sunday.
  • DNS handling: whether the provider will publish and correct SPF, DKIM and DMARC records on your behalf, or whether that stays with you.
  • Restore commitments: the stated time to restore a mailbox, not just the fact that backups exist.

Ask for the SLA document itself, not the marketing page. Then ask which monitoring proves the number, and whether the answer includes a public status page you can check during an incident.

Confirm Your Mail Flow Before Migrating

Run the cutover in a fixed order, and keep the rollback in your pocket.

  1. Inventory every sender that uses the domain, from the mailbox host to the website form.
  2. Lower MX and SPF TTLs to 300 seconds at least a day before cutover.
  3. Create the mailboxes on the new host and migrate one test user first. Send and receive from that account before touching anyone else.
  4. Send test messages from every application and confirm spf=pass, dkim=pass and dmarc=pass with aligned domains using the header checks from the diagnosis steps.
  5. Cut over the MX record, then watch the queue for the first hour.
  6. Keep the old host reachable for seven days. Old mail is still there, and someone will need it.

Rollback means restoring the previous MX value (and the TTL, if you raised it), then confirming inbound mail returns to the old host. Practise that sentence before you need it.

The decision that follows is who maintains this. Teams with mail administration skills and a reason to control the stack should run the self-managed model and own the records. Teams without a mail administrator should buy the managed option and hold on to domain and DNS control. ARPHost, LLC provides managed IT services and support, including email hosting, website hosting, virtual private servers, bare metal servers, cloud solutions, virtual PBX services and dedicated internet access, so the mail stack and its DNS work can sit with the same team through managed services.


Business email problems are almost always authentication problems wearing a storage costume, and they're fixable with four dig queries and a header read. ARPHost, LLC runs email hosting alongside managed IT services and support, VPS, bare metal servers and colocation, so the people answering your ticket can read a DNS zone and a mail log in the same session. Send us a message header you can't explain and we'll trace it end to end. Start at ARPHost, LLC.

Tags: , , , ,

Leave a Reply