Cold Email Deliverability: How to Measure Spam Placement

Most teams measure cold email deliverability with the one number that cannot measure it. We run outbound for 50+ B2B companies and send over 8 million personalized emails a year, and the delivered rate on a campaign dashboard has never once told us where a single message landed. Below, the 3 numbers that actually describe placement, the sourced 2026 benchmarks to judge them against, and the thresholds that should stop a campaign.

What Does Cold Email Deliverability Actually Measure?

Deliverability and delivery are 2 different measurements, and almost every argument about cold email performance comes from treating them as 1. Validity makes the distinction in plain language in its 2026 Email Deliverability Benchmark Report: delivery tracks whether the receiving mail server accepted the message, while deliverability is whether it reached the inbox instead of the spam folder.

A spam-foldered email was accepted. It counts as delivered. That is the whole problem in one sentence.

Once you split the 2, the measurement gets simple. Every message you send ends up in exactly 1 of 3 places, and those 3 places are the only deliverability numbers worth tracking.

• Inbox. The message reached the primary inbox or a tab the recipient actually looks at. This is inbox placement rate.
• Spam. The message was accepted and filtered to junk. This is spam placement rate, and it is the number this page exists to help you find.
• Missing. The message was deferred, blocked at the gateway, or silently dropped. It never reached either folder.

Missing is the category most senders have never heard of, and it is the one that hides the worst outcomes. A blocked message produces no bounce you recognize and no complaint, so the campaign looks quiet rather than broken. We cover the full measurement stack and how it fits together in what email deliverability is and why it makes or breaks outbound.

Inbox Placement Rate (IPR)

The number of emails delivered to the inbox out of total emails sent, expressed as a percentage. It is the most direct measurement of deliverability because it counts the outcome you care about rather than the handoff between servers. IPR cannot be read from a sending platform. It is measured with seed accounts, provider dashboards, or panel data.

Why Does Your Delivered Rate Say 98% When Half Your Mail Is in Spam?

Because the delivered rate is sent minus bounced, and nothing more. Your sending tool hands the message to the receiving server, the server says yes, and the tool writes down a success. What the server does with the message after that happens on the other side of a wall your platform cannot see through.

This is not a flaw in any particular tool. No sending platform on the market can report inbox placement for mail it handed off, because the receiving provider never tells it. Any dashboard claiming otherwise is modelling, not measuring.

Here is what each number in a typical campaign view actually counts, and what it quietly leaves out.

Read that table as a stack rather than a menu. Delivered rate is the floor, reply rate is the ceiling, and the gap between them is where the deliverability question actually lives. A campaign at 98% delivered and 0.4% reply has a problem somewhere in the middle, and you cannot tell whether it is placement or copy until you measure placement separately.

Bounce rate is the one number in that list your platform does report honestly, which makes it the cheapest early warning you own. We break down the codes and what each one implies in cold email bounce rates, causes, codes, and fixes, and the upstream fix in why email verification matters before you hit send.

What Are the Benchmark Numbers You Should Measure Against?

A placement number means nothing without a comparison. Validity's 2026 report tracks placement with seed accounts across hundreds of mailbox providers, and the provider-level split is wide enough that a single global average will mislead you.

Source for every row above: Validity's 2026 benchmark report, covering 2025 placement data.

Three things in that table change how you read your own numbers.

Microsoft is the hardest inbox in B2B, by a wide margin. At 77.4% inbox and 15.1% spam, a list heavy with Outlook and Hotmail addresses will measure worse than the same copy sent to Gmail, with nothing wrong in the campaign. Segment your placement reporting by provider or you will chase a copy problem that is really a list composition problem.

B2B gets filtered twice. Validity found 83.0% inbox placement at hosting platforms and 79.0% behind corporate filtering companies, and most businesses run both. Office 365 and Google Apps hold over 95% of the hosting market between them, while Proofpoint alone covers slightly over 50% of corporate filtering. Your mail clears the gateway, then clears the host, then gets sorted.

The global average flatters you. If you are selling to businesses, 87.2% is not your bar. The 79% to 83% band is. A cold campaign measuring 84% inbox placement into corporate domains is doing better than the global number suggests, and one measuring 84% into consumer Gmail is underperforming.

Which Numbers Do the Mailbox Providers Measure You On?

You get graded whether or not you are measuring. The providers publish their thresholds, and those published numbers are the only hard lines in deliverability.

Google's email sender guidelines define a bulk sender as a domain sending more than 5,000 messages a day to personal Gmail accounts, require SPF, DKIM, and DMARC at that volume, require a TLS connection and valid forward and reverse DNS, and tell senders to keep the spam rate below 0.30% with a target under 0.10%. Yahoo's sender best practices publish the same 0.30% line, require both SPF and DKIM for bulk senders, and require a valid DMARC policy of at least p=none that passes.

Microsoft joined them in 2025. Its announcement of new requirements for high volume senders applies the same 5,000 a day threshold to Outlook.com, Hotmail.com, and Live.com, and non-compliant mail is refused with 550 5.7.515 Access denied, sending domain does not meet the required authentication level. A rejection at that stage never reaches an inbox or a junk folder, so it shows up in your numbers as a bounce, not as spam placement.

Spam Complaint Rate

The percentage of delivered messages that a recipient marks as junk. Google measures it only on mail that reached an engaged recipient's inbox, which means the denominator shrinks as your placement worsens. A campaign sliding into the spam folder will often show a complaint rate that looks healthier each week while its actual placement collapses.

That last point is the single most useful thing on this page, so it is worth stating twice. Complaint rate is measured on inbox-delivered mail. If placement drops, fewer people ever see the message, fewer people report it, and the complaint number falls. A falling complaint rate is a good sign only when placement is holding. Read the 2 together or you will read a collapse as an improvement.

All 3 providers now treat authentication as the entry fee rather than a best practice. If your records are not clean, nothing else you measure matters. Start at SPF, DKIM, and DMARC explained for cold email senders and email deliverability DNS records explained, and size your sending against the 5,000 a day line using how many cold emails you should send per day.

One authentication detail bites B2B senders specifically. When mail is forwarded through a corporate gateway, the sending IP changes and SPF breaks, which is why Validity recommends implementing Authenticated Received Chain, defined in RFC 8617, so intermediate servers can preserve the original authentication result.

Why Is Open Rate the Worst Deliverability Metric You Own?

Because it stopped measuring humans. Open tracking works by embedding an invisible image and counting the load. Apple's Mail Privacy Protection downloads remote content in the background when the message arrives, not when it is read, and routes it through relays that hide the recipient's IP address.

The scale of that is the part people miss. Litmus Email Analytics measured Apple at 76.56% of all tracked email opens in September 2026, against 16.10% for Gmail and 3.48% for Outlook. When 3 out of 4 recorded opens come from a client that prefetches by design, your open rate is largely a measurement of Apple's infrastructure.

Open rate is still worth logging, because a sharp drop against your own baseline on the same list and the same client mix is a real signal. What it cannot do is carry a decision. Nobody should pause a domain, rewrite a sequence, or rotate infrastructure on an open rate alone.

Two replacements do the job better. Reply rate counts a human who read the message and chose to respond, and Validity predicts it will become a primary sender KPI in 2026 precisely because mailbox providers read replies as an engagement signal. Microsoft's own guidance tells senders to make sure the From or Reply-To address is valid and can receive replies.

Use positive reply rate as the quality layer on top of raw replies, and judge both against the segment medians in cold email reply rate benchmarks. Our own book sits at 4.6% reply against the 3.43% median published in Instantly's 2026 cold email benchmark report, and reply rate is the number we watch daily. Open rate is not on the dashboard.

How Do You Measure Placement Without Guessing?

No single tool sees the whole picture, which is why teams that rely on one of them stay confidently wrong for weeks. Here is what each one actually sees.

1. Seed placement tests. You send to a set of accounts you control across Gmail, Outlook, Yahoo, and corporate domains, then record which folder each message landed in. Free tools like Mail Tester give you a quick content and authentication read, and paid seed panels give you folder placement by provider. This is the only surface that reports spam placement directly. It is also the easiest to fool, which we cover in inbox placement tests and what they miss.
2. Google Postmaster Tools. Postmaster Tools reports domain reputation on a 4 step scale of Bad, Low, Medium, and High, plus user-reported spam rate, authentication pass rates, and delivery errors for your Gmail traffic. It is the closest thing to a provider telling you what it thinks of you, and it is free. It sees Gmail only, and it needs meaningful volume to a domain before it reports anything at all.
3. Blocklist lookups. A listing on a major blocklist suppresses delivery at thousands of receivers at once and produces symptoms that look like a content problem. Check the sending domain and IP against the public lists maintained by Spamhaus before you touch anything else, and run it on a schedule rather than only during an incident. Process and tooling in cold email blocklist monitoring.
4. DMARC aggregate reports. The rua tag on your DMARC record sends you XML reports showing which sending sources passed and failed SPF and DKIM alignment. This is the only surface that catches a forwarding or third-party sending problem before placement drops. It tells you nothing about which folder anything landed in.

Run those 4 on a fixed cadence and the picture resolves. Placement tests before every new campaign and weekly on active ones. Postmaster Tools weekly. Blocklist lookups weekly. DMARC reports weekly, or on an alert if your tooling supports it. The operational version of that loop, with alert thresholds and a Monday checklist, lives in how to monitor cold email deliverability day to day.

It is also worth reading the standards the providers themselves point at. All 4 major providers encourage senders to follow the sender best common practices published by M3AAWG, and the one-click unsubscribe header Google and Yahoo now require is specified in RFC 8058.

What Thresholds Should Force a Decision?

Measurement without a threshold is a hobby. Every number below has a band where you keep sending, a band where you investigate, and a band where more volume makes the problem permanently worse.

The complaint and bounce thresholds are the providers' published lines. The placement bands come from our own sending, calibrated against the B2B segment numbers in Validity's report rather than the global average, because a cold campaign into corporate domains is being judged by a stricter filter than a newsletter into Gmail.

Two rules make the table usable. First, when a measurement lands in the stop band, pause that domain rather than the campaign. A pool of sending domains exists exactly so one bad reading does not take the operation offline, which is the argument in multi domain sending, 5 domains vs 50. Second, never fix 2 things at once. Change one variable, let it run a week, measure again. Fixing 3 things simultaneously means you learn nothing about which one mattered.

If you are standing up infrastructure rather than repairing it, the order of operations in cold email infrastructure setup, how to set up email domains for outbound, and what a secondary domain is will keep you out of the stop band in the first place.

What Does the Measurement Loop Look Like on a Live Campaign?

Ours runs on invites rather than pitches, which changes the numbers but not the method. Every lead goes through 2 verifiers before it is ever loaded, because a bounce rate is the one deliverability metric you can fix before sending instead of after. Domains warm for weeks before they carry real volume, and each one sends a small fraction of the daily total so no single reading can take the operation down.

Then the loop. Placement tests before a campaign opens and weekly after. Postmaster Tools and blocklist checks every Monday. Bounce rate watched daily, because it moves first. Reply rate watched daily, because it is the only number that proves a human read the message.

Across that book we hold 4.6% reply against the 3.43% median in Instantly's 2026 benchmark report. The infrastructure half of that gap is unglamorous: clean authentication, verified lists, warmed domains, conservative per-mailbox volume. The other half is the ask, which is a separate problem that no amount of measurement will solve.

For the invite-specific version of all of this, including why an invite trips fewer content filters than a pitch, see podcast invites and the spam folder and domains and warmup for podcast invite campaigns. The warmup mechanics themselves are in email warmup explained and how to warm up a new email domain.

What Will Measurement Never Tell You?

Whether the email was worth sending.

Placement is a floor, not a ceiling. A campaign at 95% inbox placement with a bad list and a worse ask will produce nothing, and every number on the dashboard will look healthy while it does. We have watched teams spend 2 months tuning DNS records on a campaign whose real problem was that it was selling to the wrong 3,000 people.

Deliverability work is also downstream of targeting in a way most teams invert. A list full of role-based addresses and wrong-fit companies produces bounces, complaints, and silence, and all 3 of those degrade the reputation you then try to repair with warmup. The list is the deliverability lever. Building the list properly and personalizing at a depth that is actually worth the cost move placement more than any DNS change.

And the ask moves it further than either. Asking a stranger for 30 minutes is a large request from someone they have never heard of. Asking them to be a guest on a show is a small one that reads as a compliment, which is the mechanism behind invite versus pitch in B2B outbound and the reply gap in podcast invite reply rate benchmarks. Follow-up timing compounds it, which is in cold email follow up sequence best practices.

That is the whole engine we run for clients. We invite their ideal buyers onto their own podcast by email, book the recordings onto their calendar, and edit and publish every episode. They host their own show and own every recording. The commitment is 30 recorded conversations with their ideal buyers in 90 days, or their money back.

The Practitioner Takeaway

Stop reading the delivered rate. It is the number your platform can see, which is exactly why it is the number that cannot answer the question you are asking.

Measure 3 things instead: where mail lands, what the providers think of your domain, and whether humans reply. Seed tests and Postmaster Tools for the first, blocklists and DMARC reports for the second, reply rate for the third. Put a threshold on each one, write down what happens when a number crosses it, and hold the discipline of changing a single variable at a time.

The teams with consistent placement are not the ones with the best tooling. They are the ones who decided in advance what each number means and what they would do about it.