5. Sending Safely: Domains, Inboxes, and Limits

How much you can send, and how to set up sending without damaging your company's email.

Start here: which situation are you in?

You are emailing people who already know you

Lost deals, past customers, people who signed up, champions who moved, event contacts, your own product users. There is nothing for you to do on this page. Send from your connected work email. Lightfield already paces it at 20 a day with human-like gaps between sends. Those defaults are correct for this. Leave them alone. Go build a sequence. Come back when you want to email strangers.

You are emailing people who have never heard of you

This is cold outbound, and it is the only situation that carries real risk. You can absolutely run it. Lightfield already caps your daily volume, spaces the sends, and checks every email, so a modest amount of cold from your connected work inbox is safe by default. Keep the numbers conservative and you are fine.

Separate sending domains are only for higher volume. Once you want to send more cold than your connected inbox should carry, or you want your primary domain fully out of the blast radius, move to dedicated domains. Two ways to get them:

• Bring your own. Upload domains and mailboxes you already have, or that you set up and warm yourself, and use them here.

• Have Lightfield manage it. If you would rather not run the setup, reach out to your Lightfield rep and we will buy the domains, warm them, and set SPF, DKIM and DMARC for you.

Warming a new domain takes four to six weeks, and the clock runs whether or not you have started. If you know you will need domains, start now and run warm plays from your work email in the meantime.

If you have never done this, here is what these terms mean

The section above assumes words you may not have met yet. Here is each one in plain terms.

A sending domain is a web address you send cold email from that is not your real company domain. If you are acme.com, you send cold from try-acme.com or getacme.com. The point is isolation. If a cold domain gets burned, your real one is untouched and your invoices and customer replies keep landing.

A mailbox is one account on one of those domains, like jane@try-acme.com. Each domain carries two or three. Mailboxes are what actually send, and each has its own daily limit, so adding mailboxes is how you send more without pushing any single one too hard.

Warming is a slow, growing trickle of low-stakes email from a new mailbox over a few weeks, so providers see steady human-looking activity before it ever sends anything cold. A new mailbox that immediately fires 50 cold emails looks exactly like spam. Warming runs in the background and you do not do it by hand.

SPF, DKIM and DMARC are three settings on the domain that prove the mail is really from you. Without them Microsoft rejects your mail outright and Google routes it to spam. They are set once per domain and then left alone.

Who sets this up. Two options. Bring your own: upload domains and mailboxes you already have and use them here. Or have Lightfield manage it: reach out to your rep and we buy the domains, stand up the mailboxes, warm them, and set SPF, DKIM and DMARC for you. Managed is opt-in, so ask for it if you want it. Either way, the part you own is starting early, because the four-to-six-week warm-up runs the same whether you kick it off in week one or week three.

What it costs. If you bring your own, budget roughly $12 a year per domain and about $7 a month per Google Workspace mailbox, so low hundreds a month while you run cold, not thousands. If Lightfield manages it, the infrastructure is handled as part of your setup with us.

Why any of this matters

Gmail and Outlook decide whether your email reaches the inbox or the spam folder. They make that decision about your whole domain, not about one campaign. So if you send too much cold email too fast, or send to addresses that do not exist, they conclude your company sends spam. After that everything from that domain starts going to spam. Your campaigns, your replies to customers, your invoices. That is why cold outbound at any real volume runs on separate domains. If a cold domain gets flagged, the damage stays there.

The limits, and why each one is what it is

Lightfield ships with these already set, in Settings > Sequences. You can raise them to the ceiling. You cannot go past it.

Connected email is 20 because that address is not replaceable. It is your real work email on your real company domain. Twenty a day is inside what a person could plausibly send by hand. Never raise this to push cold volume - the mailbox is irreplaceable and the campaign is not.

Warmed mailboxes default to 50 each, and "each" is the important word. Five mailboxes at default is 250 a day. To send more, add mailboxes rather than raising the number.

The 50 assumes the mailbox is actually warmed. On a two-week-old mailbox, 50 is too much no matter what the default says. Independent guidance puts sustained safe volume nearer 20 to 30 a day, so treat 50 as a number you grow into.

LinkedIn invitations cap at 20 because that is LinkedIn's own limit divided by five working days. LinkedIn allows roughly 100 per rolling seven days, on every tier. Paying for Sales Navigator does not change it.

The invitation default is 10 because your real limit is often lower than 100. New accounts, low connection counts and poor acceptance rates all get throttled without warning. Starting at 10 means you find out you have room before you find out you do not.

LinkedIn messages are 40 because the person already accepted you. A message inside an existing connection is nothing like a request from a stranger.

The gap between sends matters as much as the daily number. Twenty emails in four minutes and twenty across a working day look completely different to a spam filter, and the burst is what gets flagged. Lightfield randomizes within the range rather than sending on a fixed interval, because exactly one email every four minutes is a pattern no human produces.

Should you raise anything?

No, for almost everyone. The defaults are the recommendation, not a starting position. Reasonable: a mailbox that has run clean for months with bounces under 1%, raised gradually while you watch the numbers. Not reasonable: you have 2,000 people and want to finish this week. That is how domains get burned. Add mailboxes, or cut the list - which will probably perform better anyway.

Things Lightfield does for you automatically

Worth knowing these exist, so you understand what happened when one stops your campaign.

• Send timing is randomized within the gap range so it looks human.

• Every email is quality-checked before it sends. If a draft cannot pass, the step stops rather than sending something bad, and the enrollment pauses. Em dashes are blocked outright, because they are the clearest sign a machine wrote it.

• A paused enrollment does not restart by itself. Ask the agent to resume it. If your sending stops and nothing looks broken, check for a paused enrollment first.

• Emails are plain text. No formatting, no hyperlinked text, no attachments. Paste the raw URL. Plain text also gets better reply rates, so this costs you nothing.

• Sending stays inside working hours on weekdays.

• LinkedIn stops when you hit the weekly limit. It does not resume on its own, and today there is no notification. Check your sequence if LinkedIn steps go quiet.

The two numbers to watch

Bounces: keep under 1%

A bounce is an email that could not be delivered.

Above 2% sustained, providers suppress your mail across their whole network, not just the campaign that caused it. Sitting at 1.9% normally means one bad list tips you over. Hard bounces are the damaging ones - the address does not exist, which is exactly what spammers produce. Suppress them permanently and never retry. Soft bounces like a full mailbox barely matter. Check per inbox, not per campaign. A campaign averaging 1.4% can hide one inbox at 6%. Verify every address before the first send. Verified lists get roughly twice the reply rate and close to three times the meetings. Watch for domains that accept mail to any address - they pass verification and bounce anyway.

Spam complaints: keep under 0.1%

Someone pressing "report spam." Google publishes 0.3% as the point where they act, and recommends staying under 0.1%. Aim for 0.1%. It is measured daily rather than averaged, so one bad send can trigger filtering.

LinkedIn, in full

• Roughly 100 invitations per rolling seven days. Not a calendar week.

• Same limit on Free, Premium and Sales Navigator. The most expensive misunderstanding in LinkedIn outbound. Sales Navigator buys better search and alerts. It does not buy invitations.

• New accounts are throttled well below 100, sometimes to 20 a week.

• Free accounts can only add a note to a handful of invitations a month. If the note matters, you need a paid seat.

• Keep acceptance above 30%. Below that, LinkedIn quietly shrinks your limit, so one bad list makes every future week smaller.

Do the arithmetic before planning around it. One person at 100 a week reaches 400 to 500 people a month. At one-third acceptance that is about 150 new conversations per person per month, before you have said anything useful. That is the whole envelope, and it does not grow with budget or headcount. So spend invitations only on people you cannot email. Someone who already knows you will open an email from your work address. An invitation spent on them is one you cannot spend on someone unreachable.

What it costs to get this wrong

Email: everything from a burned domain lands in spam, including your replies to real customers, and a badly burned domain sometimes never fully recovers. LinkedIn: exceeding the cap restricts the account, and repeating it escalates to a suspension that needs ID verification to undo. Both take far longer to earn back than to lose.

If something goes wrong mid-campaign

If the problem is the copy or the setup - a wrong phrase, the wrong sender, a bad merge field - pause the sequence, fix it with the agent, and resume. You do not touch inboxes for this.

If the problem is deliverability, isolate rather than pausing everything. Pause only the affected inbox, find where the bounces cluster (usually one data source or one recipient provider), re-verify that part of the list, and resume from a healthy inbox on the clean part. Leave the affected inbox alone for two to three weeks before returning it to cold. Pausing for a day without fixing the cause just delays the same outcome.

For reading campaign performance rather than sending health, see the Keeping It Running article.

The case for sending less

This page is full of volume arithmetic, so it is worth ending on the opposite point. Campaigns under 50 recipients average around a 5.8% reply rate. Large sends average roughly 2.1%. Nearly triple, for going small. Two hundred well-researched people with a real reason for each message will usually beat two thousand who matched a filter, and it does it without 48 mailboxes, without a monthly infrastructure bill, and without risking a domain. If you are choosing between adding inboxes and cleaning your list, clean the list.

Checklist

If you are only emailing people who know you:

• Sending from your connected work email

• Defaults left alone

• That is it

If you are sending cold:

• Volume checked: low volume can run from your connected inbox, real volume needs separate domains

• Domains sorted: your own uploaded, or Lightfield managing them for you

• Cold sending kept off the domain your customer mail runs on

• Two weeks of warmup done, then a four-to-six-week ramp

• Every address verified

• Bounce alert set at 1.8% per inbox

• Defaults left alone unless you have a specific reason

If you are using LinkedIn:

• Daily pace at the default of 10

• Acceptance rate above 30% and monitored

• Invitations reserved for people you cannot email

• No links in a first message

Where these numbers come from

Google and Microsoft publish the authentication requirements and spam thresholds, so those are precise. Per-inbox sending limits are not published by anyone, so the figures here come from operators and sit deliberately at the conservative end - sending 20 a day when 40 was safe costs you some volume, while sending 40 when 20 was safe costs you a domain. Your own bounce and complaint data always overrides the table.