What is the real state of email marketing going into 2027?

What changed in the inbox?

Two companies decide whether your mail arrives, and both changed the rules. Google published requirements in 2024 for anyone sending more than 5,000 messages a day to personal Gmail accounts. Microsoft followed on May 5, 2025 for domains sending at that volume to Outlook.com, where failures are now rejected outright rather than filed in junk.

Read the 0.3 percent as a budget. Sending 100,000 messages gives you 300 complaints to spend, and a campaign that annoys one recipient in two hundred has already overspent. Nothing in that rule cares how good your copy is, which is what makes it structural rather than a passing tightening of filters.

Volume now carries a penalty that lands on the sender. A list of 50,000 strangers used to cost you the sending fee. In 2027 it costs the reputation of the domain you sent from, and that reputation sits between your company and every future message, including the ones people asked for.

Why do the agency numbers sound impressive?

Look at how outbound agencies describe results. The headline is usually volume: four million emails sent last quarter, 60,000 mailboxes under management, 12,000 replies generated. Each number describes activity rather than outcome, and two of the three describe expense.

Run the arithmetic on the reply claim. One million messages at a 0.5 percent reply rate produce 5,000 replies, which sounds like a sales team's year made. Then separate them. Some say stop, some are out of office, some are a polite no, and a small remainder ask a question. A 0.5 percent reply rate holding one positive reply in ten is a 0.05 percent rate of interest, meaning 500 conversations from a million attempts at the people you most wanted to impress.

The cost side is easier to count. Sixty sending mailboxes on Google Workspace Business Starter cost $420 a month at the published list price of $7 per user per month,* and that excludes domains, warmup services, enrichment data, sequencing software and the person who watches the apparatus. The machine is affordable. What it spends on your behalf, in domain reputation and in the attention of people who will buy in two years, never appears on the invoice.

Reply rates have fallen for everyone, which the industry explains as a filtering problem and we would explain differently: recipients became expert at recognizing the format. A message built from a template with three variables swapped in reads as a template, and the reader has seen four hundred of them.

*\* Google Workspace Business Starter list price as published in late 2026, excluding tax.*

What does compliance actually cost?

The sender rules push serious senders into an infrastructure pattern that clients are rarely told about. Volume gets split across many secondary domains, each with a handful of mailboxes, each warmed for weeks by software that exchanges mail with other warmup accounts to simulate engagement. A domain whose reputation degrades gets retired, and another takes its place.

Three things follow. Warmup means a program takes weeks to reach capacity, which puts the spend ahead of the sending. Look-alike domains mean recipients sometimes see a name that nearly matches your brand, and security teams at large companies increasingly flag it. The treadmill also never stops, because reputation decays whenever messages underperform, which makes the real cost maintenance rather than setup.

A serious sender can work inside that pattern and still do well. The costs simply argue for sizing a program to what your market can absorb rather than to what the software can send.

Where does the law sit now?

Three regimes matter for most of our clients, and they have converged on the same principle.

In the European Union, the ePrivacy Directive requires prior consent for unsolicited commercial email to individuals, with a narrow exemption for existing customers, and member states treat business addresses differently. The GDPR sits underneath it, which means a lawful basis for holding the contact data is needed as well, and legitimate interest requires a documented balancing test.

In the United States, CAN-SPAM remains permissive about the first message. California's privacy law, as amended by the California Privacy Rights Act, gives residents rights over the personal information a company holds about them, including the right to know where it came from and to have it deleted. A purchased list of California contacts therefore creates obligations the purchaser usually has not planned for.

In India, the Digital Personal Data Protection Act 2023 is consent-first, with no general legitimate-interest basis of the kind European marketers lean on.

The reading is the same in all three. Buying a list and sending to it is the part of the old playbook that has aged worst, and the defensible position is contact data you can account for, a reason the person is relevant, and a record of how you got it.

Is personalization the answer?

Personalization became the industry's answer to falling reply rates, and most of what gets sold under that name is decoration. A sentence about a prospect's university, their recent promotion or their company's new office tells the reader that software looked them up. It carries no argument about why they should care.

A simple test separates the two. Remove the personalized fact from the message and read what remains. When the rest of the message still makes sense, the fact was decoration, and the reader will recognize it as such. When the message collapses without it, the fact was the reason for writing.

That distinction matters more as AI makes the decoration free. Any team can now generate a bespoke paragraph for every contact on a list of 50,000, which devalues the signal personalization used to carry, meaning somebody took the trouble. We made the wider argument in why AI outbound fails: pointing a faster machine at a weak reason to write produces more of the same message, delivered sooner.

What makes email work going into 2027?

Email works when it carries a reason that existed before the email did. That reason comes from one of three places.

A signal that something changed. A new executive in a relevant seat, a funding round, a hiring run in a stack you support, several people from one company reading your pricing page. Signals decide timing, which is the variable most outbound ignores, and we covered which ones deserve attention separately.

A relationship that already exists. A past client, a former colleague, a conversation from a conference, an introduction from a partner. These produce reply rates that make list-based sending look like what it is.

Familiarity you built on purpose. Someone who has read your writing, seen the founder's posts or attended a session arrives at your email already knowing who you are. This is the part most teams skip, and it is why email performs best for companies who also publish. The LinkedIn B2B Institute's work on the 95-5 rule explains the arithmetic: most of your market is not buying this quarter, and the work of being remembered is what makes the message land when they are.

Underneath all three sits the unglamorous part, meaning accurate data, authenticated domains, a sending volume your reputation can carry, a real unsubscribe path and a person who reads every reply. Email is now the follow-through of a go-to-market system rather than a substitute for one.

What we sell, and when we tell clients not to buy it

We run email programs for clients, and that commercial interest belongs on the table while you read this section. Email remains one of the highest-return channels we operate when the conditions above hold. The programs that work share a shape, starting with a few hundred accounts chosen deliberately and messages written from a real reason. Sequences stop when somebody says they are not interested, and a person owns every reply.

We say no, or not yet, in three situations. A company that cannot describe which customers have bought and why will aim the program at a guess. A company with nothing published has no way for a curious recipient to check whether it is credible, and the first thing an interested reader does is look. A company whose answer to weak results is more volume will spend its domain reputation faster than we can rebuild it.

Judge any email program, ours included, on four numbers rather than on sends: positive replies as a share of people contacted, meetings booked per hundred accounts touched, the spam rate in Google Postmaster Tools, and whether the domain still passes authentication everywhere that matters. A program that improves the first two while holding the second two is working. A program that cannot show you all four is reporting activity.