Key takeaway
Email marketing is a person choosing who gets a message and when. Marketing automation is a standing rule choosing both, off events about one contact. That's a difference in who pulls the trigger, not in software categories, and the split now lives mostly on pricing pages. The upgrade gets sold on 70.22% cart abandonment, but what you're buying is the event feed. I'd fix whether events reach your email system at all, and the question answers itself.
What each term actually names.
Email marketing is you deciding who gets a message and when, then sending it. Marketing automation is a standing rule deciding both for you, off data the system already holds about one contact. That's a difference in who pulls the trigger, not a difference in software categories. The two names describe how these tools were sold when an email service provider and a marketing automation platform were separate purchases, and the label on the box now tells a buyer very little about what's inside it. I've been building email platforms for over a decade. The team I keep meeting bought the bigger of the two categories, paid for it every month for a year, and shipped the same Tuesday broadcast out of it the whole time.
The mechanical difference is two sentences long and every page on this topic carries it. A campaign has a date, and a person types that date in. A flow has a condition, and the software watches for it.
Here's the sentence that usually goes missing. A condition needs an event to test, and an event has to reach the email system from somewhere else: a checkout, a product, or a billing job saying something happened to one person. That's what an email automation workflow is built on, and in most accounts it's the part that doesn't exist yet.
The labels stopped tracking the capabilities years ago. Tools that sell themselves as email marketing ship triggers, branching, and behavioural conditions as standard, so a buyer choosing between the two phrases isn't choosing between two feature sets. The choice is between two ways of working, and the price difference gets charged for the second whether or not you work that way yet.
Automation is just sending messages based on user actions. We built a custom system for Bluethumb to track when buyers follow an artist. The platform automatically sends new emails to collectors whenever an artist uploads new work, an architecture I detailed on eCommerce Australia. As of November 2024, that single automated alert "drives a chunk of our kind of retention sales" without any manual effort.
The five differences underneath the split.
Take the marketing off both categories and five real differences survive, each one showing up in what a team has to do rather than in what a vendor puts on a pricing page.
| Dimension | Email marketing | Marketing automation |
|---|---|---|
| Timing | A person picks the send date. | A condition about one contact picks it. |
| Data | Needs a list and some segmentation. | Needs events, and the events are the part most teams don't have yet. |
| Reach | One channel by definition. | Usually sold as several channels sharing one record of the person. |
| Upkeep | Finished the moment it sends. | Keeps running with nobody watching it until somebody turns it off. |
| Reporting | Measures a send; needs no system of record. | Measures a person over time, which needs a system of record. |
Four of those five follow from the second. Events decide whether a condition has anything to test, so they decide the timing. They're what a second channel has to write into if the shared record is to mean anything, so they decide the reach. A flow that survives unattended is one whose inputs keep arriving, and a report about a person over time is assembled out of that person's events. Data isn't one dimension of five. It's the one the other four are made of.
Which is why the upgrade lands badly so often. A team with no events buys the second category, runs the first one inside it at the second one's price, and concludes automation was oversold. It wasn't. It was bought a step early.
All five are observed from the outside, which is the limit on them. They tell a buyer what changes in their own work, and not one of them says which product to buy. Only one is testable before the money moves: whether the events already arrive.
The other thing "marketing automation" means.
The phrase covers two different things, and almost nothing written about it says which one is on the table. The first is the flow layer inside an email tool: triggers, branches, waits, and a delay somebody tuned once. It's email end to end, it sits beside the campaign screen, and on most plans now it's included rather than sold.
The second is the B2B lead-management category, and it's a genuinely different product. Lead scoring, an MQL definition somebody argued about, a handoff into a sales process, a CRM holding the record of the person, and email as one output of all that rather than the point of it. It's priced for a sales organisation because that's who it's for, and what it sells is the handoff rather than the sending.
The blur is expensive in a specific way. The second gets bought on the strength of the first, because the demo everybody remembers is a welcome sequence, and then it sends newsletters at a price set for a sales organisation rather than for a newsletter. The mistake runs the other way too, and I've watched that one cost more: a sales team scoring leads inside a broadcast tool with no record of a person to score, so the score gets recomputed from opens, and opens are the one signal a mailbox provider will happily fake for you.
If there's a sales team downstream consuming those scores, this page isn't your answer and the second category is a real purchase. Nitrosend specializes in email, so a multi-channel system of record for a sales organisation is that category's job rather than ours.
What a mailbox provider makes of the distinction.
A receiving mail server has no idea which category you bought. It sees a domain, an authentication result, a complaint rate, and a pattern of sending, and it decides where the message lands out of those four. The invoice you're paying isn't an input.
The rules are published, and they apply identically to a campaign and to a flow. Past 5,000 messages a day to Gmail accounts, Google's requirements for bulk senders ask for SPF, DKIM, and DMARC on the sending domain, one-click unsubscribe on marketing and subscribed mail, and a spam rate in Postmaster Tools kept below 0.3%. One-click unsubscribe isn't a vendor convention either: RFC 8058 is the standard describing how the one-click function is signalled in the List-Unsubscribe header. Suppression runs on the same clock for both, and that is the half automation gets wrong: a campaign is checked once, a flow has the next three sends already queued when the unsubscribe arrives. A trigger sent it rather than a person, and the reader draws no distinction.
So the boundary that changes results runs at right angles to the one this comparison argues about. It's authentication and complaint rate, and it splits transactional email from marketing mail rather than splitting campaigns from flows. Automation makes that boundary harder to hold, not easier, because a standing rule keeps mailing on its own schedule while the person who'd have noticed a rising complaint rate is looking at something else. A bad campaign is a bad afternoon. A bad flow is a bad quarter, discovered late.
Where this stops being true: none of it decides which tool to buy. It decides what either tool has to do before the buying question matters at all, and a team failing on authentication will fail on it in both categories at the same rate.
Three tests for whether the upgrade is worth paying for.
Three questions settle it, and they run in this order because each one only matters if the one before it came back yes.
- Do the events already reach the email system? A purchase, a trial start, or a usage threshold arriving as an event, rather than as a CSV somebody exports on Fridays.
- Is there a second channel with an owner? A named person accountable for it now, rather than a plan to hire for it next year.
- Is anyone downstream consuming a score? A sales team that acts on it, or nobody, in which case scoring is a report that nobody reads.
Cart abandonment is the flow every vendor demonstrates, and it's the fair example of the first test. Baymard puts the average documented rate at 70.22% across 50 studies, which is exactly why recovery is the first flow anybody is shown and the first one anybody wants. The flow is worth building. It's also worth nothing whatsoever until the abandonment event reaches the email system, and that's a job for whoever owns the checkout rather than for whoever owns the email.
Three noes isn't a verdict against automation. It's a verdict about sequencing. Fix the events first, and the tool question tends to answer itself six months later, usually in favour of the tool you already have.
Most visitors leave your site without buying anything. A top tier store might see a 1% conversion rate while the other 99 out of 100 people bounce. Capturing emails changes the math entirely. A decent popup will "capture 6% of that traffic into your email funnel" as I outlined on QA Selling Online. You can effectively quadruple your conversion rate with a simple automated sequence.
The option this comparison hides.
Both sides of this argument are dashboards. One is a dashboard where a person picks a send date and the other is a dashboard where a person drags a flow onto a canvas, and every page ranking for this question is asking which of those two dashboards you should log into. AI shows up on those pages as an optimizer, picking a send time or a next best action, and never as the thing operating the stack.
There's a third answer, and it starts from noticing that a flow is just rules. Rules are text. Text can be written, read, reviewed, diffed against last month, and changed by something that isn't a person with a mouse. A flow you can read a year later beats a flow you drew, and a flow an agent can write beats both, because the reason flows go unbuilt is almost never that the canvas was too hard. It's that building one was a two-hour job somebody kept not starting.
Every existing email platform was designed for humans clicking buttons, and retrofitting an agent onto a dashboard-first product is bolting a motor onto a bicycle. We built the other way round. Nitrosend is MCP-first: every capability is an API endpoint and an MCP tool before it's a screen, so contacts, segments, templates, campaigns, and flows are things an agent creates and changes by being asked. BYO sending keys are available on Pro and above, covering Amazon SES, Resend, Postmark, Mailgun, and SendGrid, so the sending can stay in an account you already own, and a team with the dev resources to run it that way sees 70 to 80 percent lower sending costs at scale doing it. Contacts are unlimited on every plan, including Free, which matters here because the other category's pricing is what turns a list into a cost centre.
An agent can write the rules for the mail. It can't own the lead record a sales team works from, and that stays a different purchase.
Building the software to run these campaigns is rarely fast. Phil and I originally wanted a simple tool to let stores use their data for automated email marketing. We assumed the project would take three or four months. The reality on QA Selling Online was that "17 months later, we have an MVP" ready for users. We finally launched the Shopify plugin back in 2016.
What to do with the answer.
The decision was never which of two phrases to buy. It's whether the events exist, whether anybody owns a second channel, and whether anybody downstream acts on a score. Answer those three, and then the one this comparison hides: whether the rules are something a person or an agent can read. By then the category question has dissolved, because for most teams the flows they can already build have never been built properly, and doing that costs a couple of weeks rather than a migration and a year of licence fees.
So start there. If the answer is that you already have the flows and just need them run properly, you need somewhere sane to run them. Nitrosend's free tier is 8,000 emails to start and then 500 a month, with unlimited contacts and the MCP server, API, and CLI switched on, and no card. Write one flow as rules an agent can read, wire one real event into it, and send it. You'll know inside a fortnight whether you ever needed the bigger category at all.
Sources
- Google, Email sender guidelines: the 5,000-a-day threshold for Gmail bulk senders, SPF, DKIM, and DMARC on the sending domain, one-click unsubscribe on marketing and subscribed mail, and the requirement to keep the Postmaster Tools spam rate below 0.3%.
- RFC 8058: the standard describing how a one-click unsubscribe function is signalled in the
List-Unsubscribeheader field. - Baymard Institute, cart abandonment rate: 70.22% as the average documented online shopping cart abandonment rate across 50 studies.
Common questions
<p>Inside an email tool, that's close to the truth. The flow layer is triggers, branches, and waits, and the mail it sends is the same mail a campaign sends. The phrase also names a second thing, though: the B2B lead-management category built around lead scoring, an MQL handoff, and a CRM holding the record of the person. Email is one output of that rather than the point of it, and the price you're quoted depends entirely on which of the two you're being sold.</p>
<p>Email automation is flows inside an email tool, end to end: a trigger fires, a branch decides, a message goes out. Marketing automation usually means those same flows plus other channels, one shared record of each person, and often lead scoring feeding a sales process. The distinction that costs money isn't the name on either of them. It's whether events about your customers reach the system at all, because a flow with nothing to test is a broadcast on a delay.</p>
<p>Almost certainly not. Triggers, branching, waits, and behavioural conditions ship as standard in tools that call themselves email marketing, so a single-channel sender is usually paying the second category's price to do the first category's job. The platform earns that price when a second channel has a named owner and a sales team is genuinely acting on scores. Neither of those is true in most of the accounts that buy one.</p>
<p>Both, and the argument is about vocabulary rather than about software. A drip is a fixed sequence of messages sent on a delay after somebody joins it, so a person designed the timing once and the system runs it from then on. That puts it on the automation side by mechanism and on the email marketing side by everything else. The sequencing itself, the delays, and where a drip goes wrong are questions about drip sequencing rather than this one.</p>
<p>Events that arrive on their own. A purchase, a trial start, a plan change, or a usage threshold, reaching the email system as it happens rather than as a CSV somebody exports on Fridays. On top of that you want a stable identifier for the person, so events from different systems land on one contact, and consent recorded against that contact. With those three, a flow has something to test. Without them, automation is a broadcast tool on a bigger plan.</p>
<p>Yes, and nothing about the trigger changes it. Google's sender guidelines ask marketing and subscribed messages to support one-click unsubscribe and to carry a clearly visible unsubscribe link in the body, and they don't ask who pressed send. The statutory clock runs the same way for both, and Yahoo asks bulk senders for two days. A campaign is checked once. A flow has the next three sends queued when the unsubscribe arrives, so the check belongs immediately before each send.</p>
<p>No, and it can quietly make things worse. A mailbox provider scores a sending domain on authentication, complaint rate, and sending pattern, and it doesn't know whether a person or a rule chose the recipient. What automation adds is a standing rule that keeps sending while nobody is watching the complaint rate. A campaign stops the moment it's sent. A flow keeps going until somebody turns it off, which is usually a week after the numbers turned.</p>
<p>Yes, if the platform was built that way round. A visual builder is one way of editing a flow rather than the flow itself, and a tool that only exposes the builder leaves an agent with nothing to hold. Nitrosend is MCP-first: every capability is an API endpoint and an MCP tool before it's a screen, so contacts, segments, campaigns, and flows are all things an agent can create and change. Describe the flow you want and the agent builds it, and what you keep afterwards is a definition anybody on the team can open and review.</p>