# Where Dealer Cold Email Dies: 43,302 Sends by Mail Gateway

*Email Deliverability · Updated 2026-09-15 · 8 min read*

**Across 43,302 cold emails to US dealerships sent from Google-hosted inboxes, the receiving mail gateway explained more of the outcome than the copy did. Google Workspace recipients replied at 1.09%; Microsoft 365 recipients, 45% of the audience, replied at 0.07%. Mimecast rejected 52 to 72% of sends outright while a paid verifier called the same addresses deliverable. Classify every dealer list by MX gateway before sending, route Microsoft recipients to Microsoft-hosted senders, and stop paying to verify bounces that are policy.**

Most cold email advice treats deliverability as one number. This is what the number hides. In June 2026 Provena audited 43,302 contacted dealership leads across three sending workspaces, all sending from Google-hosted inboxes with clean authentication and warm-up scores of 99 to 100. We classified every recipient domain by its MX record into the mail gateway that receives it, then looked at bounces and replies by gateway. The gateway, not the message, decided whether a dealer ever saw the email.

## Who actually receives dealership email?

Dealerships do not run their own mail. They sit behind a small number of gateways, and the mix is not what most vendors assume. In a 10,760-contact sample from one workspace, Microsoft 365 received 45% of the audience, Google Workspace 20%, and a long tail of small hosts 26%. Security gateways that sit in front of a mailbox, Barracuda, Sophos and Mimecast, covered the remaining 9%. The market-wide picture, by brand and state, is in the [dealership email gateway census](/blog/dealership-email-gateway-census).

![Dealership contacts by mail gateway: Microsoft 365 45%, other 26%, Google Workspace 20%, Barracuda and Sophos 5%, Mimecast 4%](/images/blog/dealer-email-gateway-share.svg)

*Share of dealership contacts by receiving mail gateway, June 2026. Microsoft 365 is the largest single destination.*

That mix matters because the four gateways behave completely differently toward the same sender. Treating them as one audience produces one blended bounce rate and one blended reply rate, and both are lies.

## How did reply rate change by gateway?

Same copy, same senders, same sequence. Google Workspace recipients replied at 1.09%. Microsoft 365 recipients replied at 0.07%. That is a 14.8x gap on the largest part of the audience, and it is not a difference in interest.

![Reply rate by gateway: Google Workspace 1.09%, Barracuda 2.1 to 2.8%, Sophos 2.4 to 3%, Microsoft 365 0.07%, Mimecast 0%](/images/blog/dealer-email-reply-rate-by-gateway.svg)

*Reply rate by receiving gateway across 43,302 contacted dealership leads. Microsoft 365 is 45% of the audience and effectively silent.*

We know it is delivery rather than audience because of what happened to the negative replies. On Google Workspace, positive and negative replies arrived in roughly equal numbers: 75 positives, 75 polite declines. On Microsoft 365 the declines collapsed seventeen-fold. Dealers cannot say no to an email they never saw. Microsoft was accepting the mail at the door, bounce rate only 2.6 to 4.4%, then filing it where nobody reads.

| Receiving gateway | Share of audience | Bounce rate | Reply rate | Positive : negative replies |
| --- | --- | --- | --- | --- |
| Google Workspace | 20% | about 1% | 1.09% | 75 : 75 |
| Microsoft 365 | 45% | 2.6% to 4.4% | 0.07% | 24 : 5 |
| Barracuda | about 4% | 7% to 12% | 2.1% to 2.8% of delivered | 4 : 1 |
| Sophos | about 1% | low | 2.4% to 3.0% | small sample |
| Proofpoint | under 1% | 0% to 0.9% | no replies observed | 0 : 0 |
| Mimecast | 4% | 52% to 72% | 0% | 0 : 0 |

*Outcomes by receiving gateway for Google-hosted senders, June 2026. Barracuda's reply rate is measured on delivered mail.*

## Which bounces are policy rather than bad addresses?

Mimecast bounced 52% of sends in one workspace and 72% in another, and produced zero replies. Those addresses were real. Before the audit, the same list had been run through a paid verification service that marked them deliverable, and they bounced anyway. Mimecast was rejecting the sender class, cold mail from young Google-hosted domains, not the mailbox. No amount of verification spend fixes a gateway policy.

![Bounce rate by gateway: Mimecast 52 to 72%, Barracuda 7 to 12%, Microsoft 365 2.6 to 4.4%, Proofpoint and Google about 1%](/images/blog/dealer-email-bounce-by-gateway.svg)

*Bounce rate by receiving gateway. Mimecast bounces are a rejection of the sender class; Barracuda bounces some mail and converts the rest.*

Barracuda looks like a problem on the bounce chart and is the opposite. It bounced 7 to 12% of sends, then delivered the rest into inboxes that replied at 2.1 to 2.8%, the best delivered reply rate of any gateway. Sophos behaved the same way with a lower bounce. A team suppressing every high-bounce domain would have thrown away its best-converting recipients.

## What did the sending infrastructure look like?

The obvious objection is that the senders were burned. They were not. Every sending domain had SPF, DKIM and DMARC aligned; warm-up scores sat at 99 to 100; no inbox was banned; and bounce was spread evenly across sending domains rather than concentrated on a few. The one infrastructure fault the audit found was density, about three inboxes per domain where two is the safe ceiling, and it did not explain the gateway pattern. The pattern was on the receiving side.

- Sending inboxes: 100 per workspace, all Google Workspace, warm-up average 99 to 100, none banned.
- Authentication: SPF, DKIM and DMARC aligned on every sending domain.
- Bounce spread: uniform across sending domains, so no single domain was the cause.
- Sender to receiver path: Google-hosted sender into Microsoft-hosted recipient was the failing path on every workspace.

## What else did the replies reveal?

Of 221 negative or stale replies we read by hand, 44% were about the contact rather than the offer: retired, left the store, sold the dealership, moved roles. The list had passed every structural check, no duplicates, personal addresses, correct titles, populated variables, and still failed on freshness. That failure is invisible to a verifier and to a scorecard. Only a reply exposes it, which is another reason a reply rate below 1% on dealers is a data problem before it is a copy problem.

## What does this change about how dealer outbound should run?

1. Classify every dealer list by MX gateway before the first send. It takes minutes and it is the single most predictive field on the record.
2. Send Google to Google. Route Google Workspace recipients to Google-hosted senders, where the 1.09% reply rate lives.
3. Route Microsoft 365 recipients to Microsoft-hosted senders. Do not send Google-to-Microsoft cold mail and read the silence as disinterest; 45% of the dealer market is behind that gateway.
4. Keep Barracuda and Sophos. Accept the bounce, keep the recipients, measure reply on delivered mail.
5. Suppress Mimecast rather than verify it. The bounce is policy; the money spent verifying is wasted.
6. Watch bounce and reply by gateway, not overall. The blended numbers hid a dead channel that was nearly half the audience.
7. Treat negative-reply collapse as a placement alarm. When declines vanish on one gateway and not another, the mail is not arriving.

Provena now runs every dealer campaign this way, as part of its [automotive outbound service](/solutions/automotive), and the [dealer outbound playbook](/blog/dealership-outbound-playbook) and the [cold calling guide](/blog/cold-calling-car-dealerships) both assume the list has been gateway-classified first. The SellMyRide and VettX [case studies](/case-studies) were run on gateway-routed lists after this audit. For the authentication side, see the [SPF, DKIM and DMARC deep dive](/blog/spf-dkim-dmarc-deep-dive) and the rest of the [email deliverability category](/blog/category/email-deliverability).

## How was this measured?

Population: 43,302 contacted dealership leads across three Provena-operated sending workspaces, exported in full through the sending platform's API on 9 June 2026. Gateway classification: the MX records of every recipient domain were resolved and mapped to Google Workspace, Microsoft 365, Barracuda, Sophos, Proofpoint, Mimecast or other, with the mapping cached and spot-checked by hand. Bounce and reply rates are per contacted lead. Positive and negative reply labels were assigned by reading each reply. Ranges are given where the three workspaces differed. This is Provena's own operating data, not a survey, and the sample is US franchise and independent dealerships contacted for two automotive software clients and one agency offer.

## Sources

[Provena outbound audit, June 2026, three sending workspaces, 43,302 contacted leads](https://www.provena-ai.com/editorial-standards): exported through the sending platform API and classified by recipient MX record. Gateway behaviour was cross-checked against [Google's sender guidelines](https://support.google.com/a/answer/81126) and [Microsoft's anti-spam protection documentation](https://learn.microsoft.com/en-us/defender-office-365/anti-spam-protection-about). Reply labels were assigned by hand. The underlying campaign data is client-confidential; aggregate figures are published under Provena's [editorial standards](/editorial-standards).

## Frequently asked questions

### Why does cold email to Microsoft 365 dealerships get so few replies?

Because most of it is not being read. In Provena's 43,302-send audit, Microsoft 365 accepted the mail with a 2.6 to 4.4% bounce rate and then replied at 0.07%, against 1.09% for Google Workspace on identical sends. Negative replies collapsed seventeen-fold on Microsoft at the same time, which only happens when the mail is filed out of sight. The failing path was Google-hosted sender to Microsoft-hosted recipient.

### Should you suppress high-bounce domains in dealer cold email?

Only the ones that reject everything. Mimecast bounced 52 to 72% of sends with zero replies and should be suppressed. Barracuda bounced 7 to 12% and then produced the best delivered reply rate of any gateway, 2.1 to 2.8%. Suppressing by bounce rate alone removes the best recipients along with the worst.

### Does email verification fix dealer bounce rates?

Not for gateway bounces. The addresses that Mimecast rejected had been marked deliverable by a paid verification service before the audit, and they bounced anyway, because the gateway was rejecting the sender class rather than the mailbox. Verification still catches dead addresses; it cannot see a policy.

### How do you find out which mail gateway a dealership uses?

Resolve the MX record for the recipient domain. Microsoft 365 domains point at outlook.com hosts, Google Workspace at google.com hosts, and security gateways such as Mimecast, Barracuda, Proofpoint and Sophos at their own hostnames. Provena classifies every list this way before sending and routes by the result.

### What reply rate should a dealer cold email campaign expect?

On Google Workspace recipients from Google-hosted senders with a fresh list, Provena measured 1.09% across 43,302 leads, with individual campaigns reaching 1.27 to 1.66%. Blended across all gateways without routing, the same campaigns showed 0.26 to 0.35%, which is why the blended number is not a useful benchmark.

## Sources

- [Provena outbound audit, June 2026, three sending workspaces, 43,302 contacted leads](https://www.provena-ai.com/editorial-standards)
- [Google Workspace email sender guidelines](https://support.google.com/a/answer/81126)
- [Microsoft 365 outbound and inbound spam protection overview](https://learn.microsoft.com/en-us/defender-office-365/anti-spam-protection-about)

---
Source: https://www.provena-ai.com/blog/dealer-cold-email-deliverability-by-mail-gateway