Why Your Cold Email Dies at Enterprise Inboxes, and Why Sending More Makes It Worse

Cold email dies at enterprise accounts because an enterprise inbox is a stack of independent filters, and every additional send gives that stack more evidence against you. A personal Gmail inbox is 1 system with a published sender contract. A Microsoft 365 mailbox sitting behind Proofpoint or Mimecast is 3 or 4 systems, each returning its own verdict, and at least 1 of them keeps re-judging the message for 48 hours after it lands. Understory Agency runs cold and signal-based outbound into exactly these accounts. This guide sets out what each layer does, from the vendors' own documentation, and gives you a diagnostic that tells you in a week whether your problem is copy or placement.

The question behind most of the searches that land here is a simple one. Your program gets replies from mid-market companies on Google Workspace and almost nothing from the large enterprises you actually want. Is it the message, or is Outlook filtering you out? Usually it is the filtering, and usually the reply data already shows it, once you split it by the recipient's mail provider.

Key takeaways

• Enterprise inboxes are a stack, and each layer decides separately. Microsoft documents that when a Microsoft 365 domain routes mail through a non-Microsoft service first, "the message adopts the source IP of the last hop that sits in front of Microsoft 365" (Microsoft Learn, Enhanced Filtering for Connectors, 2026-08-11). A gateway, then Defender, then the mailbox, then a 48-hour re-check.
• Microsoft 365 junks bulk mail on a complaint score you never see. Bulk complaint level (BCL) runs 0 to 9. The default anti-spam policy sends anything at 7 or above to Junk Email, the Standard preset junks at 6, and Strict quarantines at 5 (Microsoft Learn, Bulk email detection, 2026-08-25).
• Google publishes the ceiling. Senders of 5,000 or more messages a day to Gmail must authenticate with SPF, DKIM and DMARC and keep the Postmaster Tools spam rate below 0.10%, never reaching 0.30% (Google, Email sender guidelines). Outlook.com adopted the same 5,000-a-day authentication bar on May 5, 2025 (Microsoft Tech Community, 2025-04-02).
• Reply rate by mail provider is the diagnostic almost nobody runs. On a medical-aesthetics software client's program Understory Agency ran, campaigns segmented to Google-hosted recipients replied at 2.7% and 3.4% against 1% to 1.5% for everyone else, on the same list build and the same copy.
• Volume is not the lever at enterprise. Understory Agency sent 242,750 emails for the same client at a 0.8% lifetime bounce rate on 210 branded mailboxes across 2 domain pools, and the reply rate came from segmentation and list hygiene, not send count.
• Understory Agency engineers deliverability as infrastructure: dedicated sending domains separate from your primary domain, premium AI-powered warmup, monthly inbox rotation and regular inbox-placement tests against major providers, on custom flat retainers, never a percentage of spend (Understory Agency, GTM engineering).

Why do enterprise inboxes behave differently from everyone else's?

Enterprise inboxes behave differently because the message passes through more decision points, and the decisions are made before delivery, at delivery, at click, and again after delivery. A small company on Google Workspace runs Gmail's scanning, which Google says cannot be turned off, plus whatever policy the admin adds. A large enterprise on Microsoft 365 typically runs a third-party gateway at the MX record, then Exchange Online Protection or Defender for Office 365 policies, then per-mailbox signals. After delivery, zero-hour auto purge watches the mail for 48 hours (Microsoft Learn, Zero-hour auto purge, 2026-06-01).

The practical consequence is that a message can pass the gateway and still die in the tenant policy, or pass both and be pulled out of the inbox 20 hours later. Open and click dashboards cannot see any of this. Reply data, split by provider, can.

What does Microsoft 365 actually do to a cold email?

Microsoft 365 assigns every inbound message from a bulk sender a bulk complaint level, and the tenant's anti-spam policy decides what happens at its threshold. Microsoft's own description of the scale: BCL 1 to 3 is a bulk sender "that generates few complaints," 4 to 7 "a mixed number of complaints," and 8 or 9 "a high number of complaints." The default policy junks mail at BCL 7 or above. The Standard preset junks at 6 and the Strict preset quarantines at 5 (Microsoft Learn, Bulk email detection, 2026-08-25). Microsoft says it "uses both internal and external sources to identify bulk mail," which means a sending domain's complaint history follows it into every tenant.

3 more behaviors decide whether a human ever reads the message:

1. First contact safety tip. Available to all cloud mailboxes, it shows "You don't often get email from" above a message the first time a sender writes, and it has no dependency on spoof or impersonation settings (Microsoft Learn, Anti-phishing policies, 2026-08-20). Every cold email is, by definition, first contact.
2. Unauthenticated sender indicators. A "?" appears on the sender's photo when the message fails SPF or DKIM and also fails DMARC or composite authentication. A "via" tag appears when the From domain differs from the DKIM signing domain or the MAIL FROM address. Both are on the same Microsoft page. A sending setup that signs with the sequencer's domain instead of your own earns the tag.
3. Safe Links and zero-hour auto purge. Safe Links rewrites URLs to the safelinks.protection.outlook.com prefix, scans them before delivery, detonates URLs "that don't have a valid reputation" in the background, and checks again at click (Microsoft Learn, Safe Links, 2026-05-22). Zero-hour auto purge then keeps watching: for unread messages identified as spam after delivery, it moves them to Junk Email or quarantine, searching the last 48 hours of delivered mail (Microsoft Learn, Zero-hour auto purge).

1 detail worth knowing for anyone reading headers: the spam confidence level (SCL) no longer determines the verdict in cloud organizations. Microsoft says the same SCL value "can appear on messages with different verdicts," and that the CAT field in the X-Forefront-Antispam-Report header identifies what filtered the message (Microsoft Learn, Spam confidence level, 2026-08-03). If a friendly recipient forwards you a header, read CAT and BCL.

What does Google Workspace do differently?

Google publishes the sender contract, and Workspace admins add policy on top of it. The guidelines are written for personal Gmail accounts, and they are explicit. Senders of 5,000 or more messages a day must set up SPF, DKIM and DMARC and align the From domain with SPF or DKIM. They must keep the spam rate reported in Postmaster Tools below 0.10% and never at 0.30% or above, and support one-click unsubscribe on marketing mail. Google's stated consequence: "your email might not be delivered as expected, or might be marked as spam" (Google, Email sender guidelines).

Inside a Google Workspace tenant the admin layer is visible in the console. Google's admin documentation says "By default, Gmail scans all email messages for spam" and "You can't turn off Gmail's spam scanning." Admins can "Turn on more aggressive spam filtering" and send messages to a quarantine for review instead of the spam folder. Filters are bypassed only for approved senders and internal mail (Google Workspace Admin Help, custom spam filters). A separate setting, enhanced pre-delivery message scanning, means "message delivery is slightly delayed so that Gmail can do additional security checks on the message," after which a warning is displayed and the message might be moved to spam (Google Workspace Admin Help, pre-delivery scanning).

The measurable difference shows up in reply rates. On the outbound program Understory Agency built for a vertical SaaS company selling into medical aesthetics, campaigns were segmented by whether the recipient was on Google. In March 2026 the program sent roughly 47,000 emails and produced 69 positive replies and a closed deal. The Gmail-segmented campaigns replied at 2.7% and 3.4%, against 1% to 1.5% for non-Google recipients, on the same list build. The only cost was an extra column in the list. That gap is the enterprise problem in miniature, because the non-Google side of any B2B list is where the gateways and the Defender policies live.

What do Proofpoint and Mimecast add on top?

Proofpoint and Mimecast add a gateway ahead of the mail platform, and the first thing to know about a target account is whether one is there. The MX record tells you. Microsoft 365 tenants point their MX at a mail.protection.outlook.com hostname (Microsoft Learn, External DNS records for Microsoft 365, 2026-06-26); Google Workspace uses smtp.google.com, with older aspmx records still supported (Google Workspace Admin Help, MX records). An MX record that names a security vendor instead means every message you send is judged by that vendor before Microsoft or Google ever sees it.

What the gateways document about their own behavior, on their public pages:

• Proofpoint describes Core Email Protection as AI-driven threat identification and classification with "Sandboxing for malicious URLs and attachments," and states that after delivery "Emails are checked again for new indicators of compromise, unusual behavior, or delayed-detonation payloads" (Proofpoint, Core Email Protection). Its URL Defense data sheet describes protection "on any device, at any time, from any location, by following the email and checking for the URL destination's safety in real-time" when the recipient clicks (Proofpoint, URL Defense data sheet).
• Mimecast states that its URL protection "rewrites all links in inbound email and scans the destination website in real-time when clicked by the user," defends against typo-squatted domains and alternate character sets, and performs "real-time inspection and sandboxing of downloaded files" (Mimecast, URL Protection).

2 consequences for a cold email program follow directly. First, any link in your message is rewritten, and a machine will request it before or instead of a person, so open and click metrics on gateway-protected accounts are unreliable evidence of anything. Second, when a gateway sits in front of Microsoft 365, Microsoft's own filtering still runs afterwards, and Microsoft documents Enhanced Filtering for Connectors precisely so its spoof and phishing models can recover the true source IP behind the gateway (Microsoft Learn, Enhanced Filtering for Connectors, 2026-08-11). Passing the gateway is 1 verdict of 3.

Which domain and sending-infrastructure mistakes kill cold email at enterprise accounts?

The infrastructure mistakes that kill enterprise cold email are all visible in DNS and in the mailbox count, which is why they are the first things Understory Agency audits on any program. In order of damage:

1. Sending from the primary company domain. A cold program's complaint history attaches to the domain that sends it. Understory Agency registers and configures dedicated sending domains separate from your primary domain "so a deliverability issue never threatens your real email" (Understory Agency, GTM engineering FAQ).
2. Missing or misaligned authentication. No SPF or DKIM pass earns Microsoft's "?" indicator; a DKIM signature on a domain other than the From domain earns the "via" tag (Microsoft Learn, Anti-phishing policies). Above 5,000 messages a day, both Google and Outlook.com require SPF, DKIM and DMARC with alignment, and Outlook.com began routing non-compliant high-volume mail to Junk on May 5, 2025 (Microsoft Tech Community).
3. Too few mailboxes carrying too much volume. The programs that hold up spread sends wide. For the medical-aesthetics software client, Understory Agency built 210 branded mailboxes on the client's own domains, ran 4 campaign angles across 2 domain pools refreshed on a cycle, and took daily volume from 600 to over 2,000. For WareSpace, sending ran from 4 locations at 34 inboxes each while daily volume climbed from 1,200 to over 3,000.
4. No rotation and no placement testing. Understory Agency rotates inboxes monthly and builds automations that manage the ratio of cold sends to warmup emails. It runs regular inbox-placement tests against major providers and does not trust in-platform spam scores alone (Understory Agency, GTM engineering).
5. Unvalidated lists. Bounces are a reputation signal and they drift upward as lists age. Every address on an Understory Agency program is double-validated through a waterfall before a single send. On the medical-aesthetics program, a purge to valid-only addresses took the bounce rate from 3.6% in June to 1.3% by late July, on a program that finished 242,750 sends at 0.8% lifetime.

None of these is a copy problem, and none of them shows up in a sequencer's dashboard as anything other than a falling reply rate.

Why do reply rates collapse when you scale volume?

Reply rates collapse at volume because the 2 numbers the filters weigh most, complaints and bounces, both grow faster than replies do. Microsoft's bulk complaint level is a complaint score, and it rises with "unsolicited messages that closely resemble spam and generate many complaints from recipients" (Microsoft Learn, Bulk email detection). Google's ceiling is a spam rate of 0.30% of delivered mail, with 0.10% as the target (Google, Email sender guidelines). Doubling sends to people who did not ask to hear from you doubles the raw complaint count while the reply count stays flat, so the ratio that defines your domain gets worse with every batch.

The second mechanism is the list. The first 5,000 contacts in any build are the best-fit ones. Scaling means sending deeper into contacts that fit worse, and older records that bounce more, which is why a bounce rate is a maintenance number. The third is fatigue inside the target tenant: the same domain hitting the same organization's mailboxes repeatedly is exactly the pattern Microsoft's bulk detection is built to recognize.

Here is the failure signature, in the shape it usually takes. A program sends into an audience that sits almost entirely on Microsoft 365. The dashboard shows a reply rate that looks alive. Split the replies and most are out-of-office notices and bounce receipts, with almost no human replies and no positive ones. Sending more from the same setup produces the same result, with a worse domain reputation attached.

The counter-example is what a well-built program looks like when it does scale. For WareSpace, Understory Agency sourced the list segment by segment from Google Maps and ecommerce data sources, and verified every record for a working website and phone number in Clay before it entered a sequence. Daily volume then scaled from 1,200 to over 3,000 across 4 locations at 34 inboxes each. The first 2 active months produced roughly 300 positive replies, including 50 in 10 days from a single market on a 7,200-contact send. Both programs spread volume across many inboxes on purpose. The safe ceiling for any program is the one its bounce rate and its placement tests tolerate, checked monthly, and it is always lower than the sequencer allows.

Does copy and personalization still matter at enterprise accounts?

Copy still decides the reply once the message is in front of a person, and the filter stack changes what good copy looks like. 3 rules follow from the documentation above.

Fewer links, and none you cannot afford to have rewritten. Safe Links and Mimecast rewrite URLs, and all 3 gateways check the destination at click. A link to a tracking redirect, a calendar tool or a shortener gets wrapped and scanned, and a destination without a valid reputation can meet a warning page instead of your booking form. Understory Agency writes first-touch enterprise emails to earn a reply, with the link arriving after the reply.

Nothing that pattern-matches to impersonation. Defender's impersonation protection looks for display names that resemble protected users and for domains that are "even slightly similar" to the recipient's, and Mimecast checks for typo-squatted domains and alternate character sets. A sending domain that echoes the target company's name, or a signature that mimics an internal role, is a phishing signature to the filter, whatever the intent.

Personalization on real variables, tested. Understory Agency personalizes every message on real, relevant variables built in Clay. It tests most emails with 2 to 3 variants, with AI determining the winner in the first 40% of the campaign, and weights the winning variant up automatically (Understory Agency, GTM engineering). The first-contact safety tip is going to sit above your message on every Microsoft 365 account; the copy underneath it has to read like a person who knows why they wrote.

Which warmup beliefs are myths?

Warmup builds a sending reputation for a new domain, and 4 things people expect it to do are things it cannot do.

1. Warmup does not create a relationship with the target tenant. Microsoft's first contact safety tip fires "The first time they get a message from a sender," and mailbox intelligence is built from each recipient's own contact history (Microsoft Learn, Anti-phishing policies). Warmup traffic between your own mailboxes creates no history inside the account you want to reach.
2. Warmup does not fix authentication. SPF, DKIM, DMARC and alignment are DNS records. Google and Outlook.com both require them above 5,000 messages a day, and Microsoft's "?" and "via" indicators are triggered by the records, not by sending history.
3. An in-platform spam score is not a placement result. Understory Agency runs regular inbox-placement tests against major providers and never trusts in-platform spam scores alone (Understory Agency, GTM engineering FAQ). A score computed by the sending tool cannot see a tenant's BCL threshold or a gateway's verdict.
4. Warm is not permanent. Microsoft's bulk complaint level is computed from ongoing complaint behavior, and zero-hour auto purge re-judges delivered mail for 48 hours. Reputation is maintained, which is why Understory Agency rotates inboxes monthly and manages the ratio of cold sends to warmup emails as an automation, on premium AI-powered warmup that mimics human sending patterns.

Understory Agency's own FAQ states the reason month 1 of a cold email program is setup: "Sending too fast on cold domains is how you land in spam and burn the channel." Campaigns typically go live in roughly 3 to 4 weeks for that reason.

Which diagnostic can you run on your own program this week?

The diagnostic that separates a copy problem from a placement problem takes 1 export and 1 DNS lookup per target domain, and Understory Agency runs a version of it on every program it audits.

1. Tag every target domain by its MX record. A mail.protection.outlook.com host is Microsoft 365. smtp.google.com or an aspmx host is Google Workspace. Anything else is a gateway ahead of the platform. This is the single column that explains most enterprise reply data.
2. Split reply rate by that tag. On the medical-aesthetics program, Google-hosted recipients replied at roughly 2 to 3 times the rate of everyone else. If your Google segment replies and your Microsoft and gateway segments do not, the copy is fine.
3. Split human replies from automatic ones. Out-of-office notices and bounce receipts are replies to a sequencer. A reply rate made mostly of auto-responses is a placement problem wearing a reply-rate costume.
4. Read bounce rate by domain pool and by list source. A rising bounce rate is list decay, and a pool that bounces above the others is the one to pause. Purging to valid-only addresses took the medical-aesthetics program from 3.6% to 1.3% in about 6 weeks.
5. Get 1 header from a friendly recipient at a Microsoft 365 account. Read the CAT and BCL values in X-Forefront-Antispam-Report, and check whether your links came back wrapped as safelinks.protection.outlook.com. Both tell you which layer acted.
6. Check your own DNS. SPF, DKIM signing on the From domain, and a DMARC record at p=none or stronger, on every sending domain as well as the primary.
7. Run a placement test against the major providers, then put it on a schedule. 1 test is a snapshot; monthly tests are the trend line.

If the diagnostic shows the Microsoft and gateway segments dark while Google replies, the next section is the fix.

When should an enterprise account move to LinkedIn instead?

An enterprise account moves to LinkedIn when its MX record names a gateway and its reply rate by provider sits at a fraction of your Google segment after the infrastructure is clean. A LinkedIn message never passes through Proofpoint, Mimecast or Defender, which makes it the channel of record for banks, large corporates and any account whose mail stack is built to keep strangers out.

Understory Agency runs LinkedIn outreach through HeyReach as a HeyReach Expert partner, coordinated with email sequences so a prospect is not contacted on both surfaces in the same hour. Connection requests, follow-ups and DM sequences are scoped to the same ICP and copy framework as email, and replies route into 1 unified inbox (Understory Agency, GTM engineering FAQ). LinkedIn outreach typically launches in 2 to 3 weeks, against 3 to 4 for cold email. The decision is made per segment, from the reply data, not per program.

How does Understory Agency engineer deliverability for enterprise outbound?

Understory Agency treats deliverability as infrastructure that is built before any message is written, and every item below is stated on its GTM engineering page:

• Infrastructure setup and intelligent warmup. Optimized sending domains, or in-depth health checks on existing ones, with Instantly's AI-powered premium warmup mimicking human sending patterns to build sender reputation and maximize inbox placement.
• Inbox health management. Monthly inbox rotation, automations that optimize the ratio of cold sends to warmup emails, and diversified email infrastructure.
• Dedicated sending domains separate from your primary domain, so a deliverability issue never threatens your real email, and regular inbox-placement tests against major providers, never relying on in-platform spam scores alone.
• Double-validated lists. Prospect lists built from custom databases, custom scrapers built in Claude Code and Clay enrichment, with every email address double-validated through waterfall enrichment before sending.
• Reply handling that thinks. A unified master inbox tracks replies from all sending domains, reply classification catches signals a naive model misses, and conversations hand off to your sales team.
• Attribution in your CRM. Every send, open, reply and meeting lands on the contact record in HubSpot or Salesforce through OutboundSync, where Understory Agency is a Gold Agency Partner, with replies routed into Slack instantly.
• A live dashboard and benchmarks. A client-facing dashboard tracking opens, replies, positive replies, meetings booked, opportunities created and pipeline sourced, with every recommendation checked against what is working across 100+ engagements.

Understory Agency is an Enterprise Clay Partner and one of the first five certified Clay Experts. It runs cold and signal-based outbound across email and LinkedIn, built in Clay and the wider GTM stack, and prices each service as a custom flat retainer, never a percentage of spend, so nobody is paid more for sending more. Cold email campaigns typically go live in 3 to 4 weeks, with month 1 spent on domain registration, DNS, warmup, ICP intake, list building, copy, sequencing and inbox configuration.

What should an outbound agency be able to show you?

An outbound agency that engineers deliverability can show you 8 things before you sign, and each one is a document or a screen, never a promise:

1. The sending domains it will use, and that none of them is your primary domain.
2. The authentication records it will publish for those domains, SPF, DKIM signing on the From domain, and DMARC, before the first send.
3. Its warmup method by name, and how the ratio of cold sends to warmup emails is managed once real sending starts.
4. Its rotation schedule, meaning how often inboxes are cycled, and how volume spreads across them.
5. Its placement-testing cadence and the last result, against the major providers, separate from any in-platform score.
6. How addresses are validated before they enter a sequence, and the bounce rate on a comparable program.
7. A reply report split by recipient mail provider and by human versus automatic reply, because that is the number that tells you whether enterprise accounts are being reached.
8. Where replies and meetings land, on contact records in your CRM with source attribution, and the dashboard you will read them on.

Understory Agency's answers to all 8 are on its GTM engineering page, and its client-facing dashboard, bi-weekly working sessions and quarterly growth audit are how the answers stay current through the engagement.

What do Understory Agency's outbound clients say?

Understory Agency publishes 18 named written testimonials and 13 client video case studies on its homepage, re-counted live on 2026-09-03. 2 of them, verbatim, from clients whose engagement was outbound:

FAQ

Our cold email gets no replies at large enterprises. Is it the copy or is Outlook filtering us out?

Usually the filtering, and your own reply data can settle it in a week. Split reply rate by the recipient's mail provider, using the MX record of each target domain, and split human replies from auto-responses. If Google Workspace recipients reply and Microsoft 365 and gateway-protected recipients do not, the copy is not the problem. Microsoft 365 junks bulk mail at a complaint score of 7 by default (6 on the Standard preset, quarantine at 5 on Strict). It shows a first-contact safety tip on every new sender, and re-judges delivered mail for 48 hours through zero-hour auto purge, all documented on Microsoft Learn. Understory Agency runs this split on every program it audits, and on the medical-aesthetics program its Google-segmented campaigns replied at 2.7% and 3.4% against 1% to 1.5% for everyone else.

Why does cold email not land in the inbox on Microsoft 365?

Because a Microsoft 365 tenant applies a stack of decisions a personal inbox does not. Exchange Online Protection stamps a bulk complaint level and junks or quarantines at the tenant's threshold. Anti-phishing policy adds a "?" for senders failing SPF or DKIM and a "via" tag when the DKIM domain does not match the From domain. Safe Links rewrites and scans every URL, and zero-hour auto purge removes unread mail identified as spam within 48 hours of delivery. A third-party gateway such as Proofpoint or Mimecast at the MX record adds its own verdict before any of that. Understory Agency's response is infrastructure: dedicated sending domains separate from your primary domain, authenticated and warmed, with inboxes rotated monthly and placement tests against the major providers instead of in-platform spam scores.

How many cold emails per day is safe to send from 1 inbox?

The safe number is the one your bounce rate and your placement tests tolerate, checked monthly, and it is lower than the sequencer allows. The programs Understory Agency runs spread volume wide. For the medical-aesthetics software client, over 2,000 daily sends ran across 210 branded mailboxes on 2 domain pools. For WareSpace, over 3,000 daily sends ran across 4 locations at 34 inboxes each. Above 5,000 messages a day per domain, Google and Outlook.com both require SPF, DKIM and DMARC with alignment, and Google's spam-rate ceiling is 0.30% of delivered mail.

Does email warmup fix enterprise deliverability?

No. Warmup builds a sending reputation for a new domain and cannot do 4 things enterprise deliverability depends on. It creates no contact history inside the target tenant, so Microsoft's first-contact safety tip still appears. It does not publish the SPF, DKIM and DMARC records Google and Outlook.com require above 5,000 messages a day. An in-platform spam score is not a placement result, and reputation has to be maintained, because Microsoft's bulk complaint level tracks ongoing complaints. Understory Agency uses Instantly's AI-powered premium warmup, then keeps the program healthy with monthly inbox rotation, managed cold-send-to-warmup ratios and regular inbox-placement tests against major providers.

When should you switch from cold email to LinkedIn for enterprise accounts?

Switch a segment to LinkedIn when its MX record names a security gateway and its reply rate by provider stays at a fraction of your Google segment after your domains, authentication and warmup are clean. A LinkedIn message does not pass through Proofpoint, Mimecast or Defender for Office 365. Understory Agency runs LinkedIn outreach through HeyReach as a HeyReach Expert partner, coordinated with email so a prospect is not contacted on both surfaces in the same hour. LinkedIn outreach typically launches in 2 to 3 weeks against 3 to 4 for cold email.

Related reading

• Understory Agency GTM engineering, the service line behind everything in this guide.
• Best B2B SaaS cold email agencies, the buyer's guide to the category, graded on deliverability engineering.
• Best outbound marketing agencies for B2B SaaS, for the wider outbound program.