Nobody remembers who set up the email. It was probably a backend engineer during the first sprint, maybe the second. A few lines of code, an SMTP relay, a shared IP address, and it worked. It kept working through the beta, through the first thousand users, through the launch. Then somewhere around 50,000 active accounts, a support ticket came in that changed the morning: “None of our customers are getting password resets.”
That is roughly the point where most SaaS companies discover that the email infrastructure they never thought about has quietly become the most fragile part of their product.
What breaks first
The pattern is remarkably consistent. Most email providers route early-stage companies through shared servers alongside hundreds of other senders. At low volume, this works fine. But as a company grows, it becomes more exposed to what everyone else on those servers is doing. If another sender triggers a wave of spam complaints, the damage spills over. Your password resets start landing in junk folders because of someone else’s bad campaign.
The second problem is subtler. Most start-ups send every type of email, from welcome messages to promotional blasts, through the same setup. It is faster to build that way, so nobody questions it. But mailbox providers do not see “transactional” and “marketing” as separate categories. They see one sender. When a promotional email gets ignored or flagged, that poor engagement drags down deliverability for everything the company sends, including the emails users are actually waiting for.
Q1 2025 data from deliverability researchers tells a stark story: senders pushing more than one million messages a month saw average inbox placement rates fall to 27.63 percent. Scale, without the right architecture underneath it, actively works against you.
The decisions that compound
There are a handful of infrastructure choices that barely matter at 10,000 sends a month and matter enormously at 10 million. Most of them are boring. None of them are optional at scale.
The most basic is domain authentication. SPF, DKIM, and DMARC are protocols that prove to mailbox providers your emails are actually coming from who they claim to be coming from. Google, Yahoo, and Microsoft made them mandatory for bulk senders in 2024 and 2025. Yet according to the Mailjet Road to Inbox 2025 report, only about 54 percent of senders have a DMARC policy in place, and more than a quarter are not sure whether their domains are properly authenticated at all. Without this foundation, nothing else you build on top of it will hold.
After that comes the question of how your different types of email are delivered. Transactional messages, bulk sends, and promotional campaigns should run on separate infrastructure so that problems in one category stay contained. If a marketing blast generates complaints, password resets and billing alerts should be completely unaffected. Very few companies set this up before they need it, and by the time they need it, the damage is usually already compounding.
Then there is the move from shared IP addresses on servers to dedicated ones. A company that has outgrown its shared infrastructure cannot simply flip a switch. A new dedicated IP has no sending history, so mailbox providers treat it with suspicion. It needs a structured warm-up, a gradual ramp over weeks, while someone watches the metrics closely. Skip that process and the IP gets flagged before it ever builds a reputation.
And finally, per-provider monitoring. What lands in Gmail’s primary tab may be hitting Outlook’s junk folder. Without that visibility, problems can persist for weeks before anyone connects the dots.
The cost of getting it wrong
The revenue math is not abstract. Research on high-volume sending programs found that improving inbox placement from 85 percent to 97 percent on a $10 million email-driven program translates to roughly $1.4 million in recovered annual revenue.
And the losses compound on themselves. When fewer emails reach the inbox, engagement drops. Mailbox providers interpret that drop as a sign that recipients do not want the messages, so they filter even more aggressively. Placement falls further, engagement falls with it, and the cycle accelerates.
Meanwhile, engineers are spending hours chasing symptoms they cannot diagnose, support is resending emails one at a time, and product team is launching experiments aimed at the wrong problem entirely. The root cause almost always turns out to be an infrastructure decision made months or years earlier that nobody thought to revisit.
How providers handle the scaling problem
Not all email platforms are built with the same scaling trajectory in mind, and the differences become clear at volume.
Amazon SES is the cheapest option at about $0.10 per thousand messages, but new accounts start with a limit of 200 emails per day. There is no built-in stream separation, no automated warm-up, and no deliverability consulting. Teams choosing SES are agreeing to build the infrastructure layer themselves.
Twilio SendGrid handles billions of messages monthly and remains the default for many enterprise teams. But its shared IP pools have drawn persistent complaints, and accessing dedicated IPs or marketing tools requires separate, more expensive plans.
Mailgun offers a flexible API, though the Sinch acquisition and a Flex plan price increase in late 2025 have introduced uncertainty about direction.
Mailtrap, a product of Ukrainian software studio Railsware, has targeted the scaling segment specifically. The company started in 2011 as an internal tool and later expanded into an email delivery platform. Railsware, whose early clients included Calendly, made the 2024 Inc. 5000 list and reached $17 million in revenue that year. The platform claims over 150,000 active customers, according to company figures.
The broader market remains fragmented, and no single provider has locked up the scaling segment. Choosing the wrong platform at 50,000 users means migrating at 500,000, which is roughly the worst possible time to be rethinking email infrastructure.
The architecture you wish you had built
At 10,000 sends a month, email infrastructure is a checkbox. At 10 million, it is architecture, as critical to product reliability as the database or the authentication layer. The companies that build it right early on rarely think about it again. The ones that do not spend months untangling decisions they barely remember making, usually while customers are asking why their emails stopped arriving.
Leave a Comment