Key takeaway
Affiliate email marketing is promoting someone else's product to a list you own, through a tracked link you get paid on. The usual advice is to find a platform that tolerates affiliate links, revisited whenever their shared pool has a bad month. The two things you can hold are the From domain and the link domain. Authenticate one, control the other, keep complaints under 0.30%, and nobody's policy is load-bearing. In my experience the operations that survive are the dull ones.
What affiliate email marketing actually asks of you.
Affiliate email marketing is promoting somebody else's product to a list you own, through a tracked link, and being paid on what that link produces. The product is theirs. The audience is yours, and so is the sending. That split is the whole subject, and every affiliate problem worth solving comes out of it: the revenue belongs to a relationship you don't control, and the reputation belongs to infrastructure you do.
The first few campaigns hide that, because nothing much is at stake while the volume is small. What breaks the arrangement is the offer that converts. Volume goes up, complaints go up with it, and the first thing anybody notices isn't a deliverability number on a chart, it's an account review in an inbox. In a decade of building email platforms I've never once seen an affiliate account closed over its first campaign. It's always the fourth one, the one that looked worth rushing.
The two you can actually hold are the domain you send from and the domain your links resolve through, and most of what gets written for affiliate senders covers the first thinly and the second not at all. That's how somebody who followed every guide still ends up wondering why one offer stopped landing while the rest kept going.
Affiliate campaigns often fail when they rely on isolated outbound blasts. A better approach requires coordination across multiple products. We built a campaign like this around a Black Friday hub. The model relies on assembling complementary tools on a single landing page. We grouped our app with others in the space to run discounts and collect emails. I framed the workload clearly on QA Selling Online: we put a lot of effort into "external collabs and webinars" to drive the initial traffic.
Why platforms ban affiliate senders.
Most email platforms send their smaller customers through shared IP pools. A shared pool means one sender's complaint rate is charged against everybody else's mail, so the platform isn't selling you sending so much as a share of a reputation other people are spending too.
Affiliate lists complain more than average, and the reason is structural rather than moral. The consent that produced the list was usually consent to something else: a lead magnet, a comparison post, a giveaway. The offer in the message isn't the thing the reader signed up for, so a proportion of them treat it as mail they never asked for, and the spam button is the fastest way to say so. I'm blunt about the specific version of this that keeps recurring: an old, unengaged list that gets mailed again anyway, hoping the affiliate offer lands where the original one didn't. It won't. It'll get marked as spam and start cooking the sending domain's reputation for everything else on it.
A platform that can't price that risk customer by customer has to manage it some other way, and the cheapest way is a content policy. That's why so much of the advice written for affiliates is a list of which vendors tolerate you and which don't. It isn't an editorial choice about affiliate marketing, it's a business model showing through.
The numbers make the pool concrete. Google's sender guidelines ask bulk senders to keep the spam rate reported in Postmaster Tools below 0.30%, with 0.10% recommended. The same page tells merchants to monitor their affiliates and remove the ones that send spam, which is the other half of the same distrust: your network is being told to watch you while your platform is being told to watch its pool.
A tolerant platform is a real thing and it's worth having. It just isn't a position you hold, because it's somebody else's decision and they get to revisit it whenever their pool has a bad month.
The three reputations you're actually managing.
A filter doesn't judge you. It judges identifiers attached to the message, and three of them decide most of what happens to an affiliate send.
- Sending domain: the domain in the From header, and the SPF, DKIM, and DMARC records published under it. Yours, if you set it up.
- Sending IP: the address the message leaves from, shared with strangers on most plans, and the one you don't control.
- Link domain: every hostname in the body, including the redirect an affiliate link travels through. Also yours, and the one nobody manages.
Two of the three are already yours, and the advice written for affiliates spends nearly all of its attention on the one that isn't. That's backwards. Authenticate the domain and a platform's tolerance stops being load-bearing, because reputation follows a domain you own rather than an account you rent.
Yahoo's sender best practices set the floor plainly: SPF and DKIM both implemented, a DMARC policy published at p=none or stronger with the From domain aligned to the SPF or DKIM domain, a spam rate under 0.3%, and unsubscribes honored within 2 days. The same page says to send only to people who asked, and warns off purchased lists and pre-checked opt-in boxes, which is a fair description of where complaint rates come from.
BYO sending keys on Pro and above put the sending account in your name too. The mail leaves through Amazon SES, Resend, Postmark, Mailgun, or SendGrid on credentials you hold, so the tool you compose the campaign in and the reputation you're building are no longer the same decision.
Authentication decides whether mail is accepted, never whether it's wanted. A perfectly aligned DMARC record on a message nobody opens teaches the filter nothing good.
The link is a reputation too.
Filters inspect the body, not only the envelope, and domains inside a message get looked up much as a sending domain does. Spamhaus's Domain Blocklist is exactly that: a list of domain names with poor reputation, consulted during content inspection by looking up domains appearing in the mail headers and body, URLs included. It carries a return code specifically for an abused spammed redirector domain, and it expects URL shortening services to check every outbound domain against the same list.
Every guide says to put a redirect in front of the raw network URL, and none says what that does. It doesn't remove the risk. It moves it onto whichever domain does the redirecting, which is only an improvement because that domain is one you can keep clean. I'd run one hostname per offer wherever the volume justifies it, so a bad advertiser burns one name rather than all of them.
Point that at a public shortener and you're standing behind everybody else's history, including the people using shorteners for precisely what you'd expect. Point it at the merchant's own tracking domain and you inherit theirs, which can be listed while yours is clean. It happens the other way round too, which is why an offer that suddenly stops delivering often has nothing to do with your list.
The practical version is dull and it works. Redirect through a hostname you control, in DNS you administer. And look up the destination before you mail it, rather than learning about it from the send.
A clean link domain doesn't rescue a message nobody wanted. It buys the benefit of the doubt on a send that was going to land anyway.
Whose message this is.
The consequences land on you, not on the merchant whose product is in the message. You pressed send, from your domain, to a list you assembled. The merchant's reputation is untouched by a complaint your send earned, and no affiliate agreement moves that back to them. An affiliate who treats the offer as somebody else's mail is running somebody else's risk on their own domain.
Three things every affiliate email owes, whoever's product is inside it.
- A subject line that reads as what it is: the reader can tell it's a promotion before opening it, not thirty seconds after.
- A real sender identity: a name, a company, and a way to reach you that isn't the unsubscribe link.
- An opt-out that works: one click, no confirmation page, actioned within the two days Yahoo asks bulk senders for.
The fake reply and the fake receipt run straight into the first of those. Both buy the open and spend the trust, and the reader who feels tricked reaches for the spam button rather than the unsubscribe link, which is the expensive way to lose them.
Disclosure is a separate job with a separate test. If you're paid when the reader buys, say so where they're reading, in words that let them weigh the recommendation while it's in front of them. A commission line does not have to be elaborate. It has to be somewhere other than the footer of a message that is one long recommendation, which is the shape that fails.
That's also the version that survives a reader forwarding the email to somebody sceptical, which is the audit that actually happens.
Every item here is a floor, and clearing it does nothing for the complaint rate. A message can name its sender, read honestly, disclose the commission, and still be the send that ends the account, because none of that makes the offer wanted.
A run book that survives the offer that works.
Five steps, run the same way every time, because the campaign that ends an account is always the one that got rushed out for being good.
- Authenticate first: publish SPF, DKIM, and DMARC on the sending domain before the first campaign, not after the first bounce report.
- Send the new offer narrow: segment to the contacts who opened something recently, and mail them before anybody else sees it.
- Redirect through your own hostname: check the destination isn't already carrying somebody else's history.
- Suppress on the first signal: complaints, hard bounces, and anything imported you can't account for.
- Watch the rate as volume rises: hold the send against the 0.30% ceiling, and slow the ramp rather than the sequence when it moves.
Make leaving cheap while you're in there, because one-click unsubscribe is a header pair, not a footer link. RFC 8058 asks for a List-Unsubscribe header carrying an HTTPS URI plus a List-Unsubscribe-Post header carrying List-Unsubscribe=One-Click, with at least one valid DKIM signature covering both. A reader who can leave in one tap doesn't reach for the spam button instead.
The failure mode matters, because the ceiling is a rate. The same number of complaints on a bigger send is a smaller problem, and the same number on a narrower one is a bigger problem. Which is why the fix for a bad week is almost never a bigger send. It's a smaller one, aimed at the people who were going to open it anyway, held there until the rate comes back down.
The dashboard is the part that doesn't scale.
Look at what's in that run book. Five operations, same order, a different offer each time. That's a script, and it's been one since the second offer.
Most email platforms are built for somebody clicking through a campaign builder. For one offer a month that's fine, and the builder is faster than writing anything. For somebody running twelve across four networks, the clicking is the job, and it's the first thing skipped when a campaign is worth rushing. Retrofitting an agent onto a dashboard-first product is bolting a motor onto a bicycle.
My rule of thumb is that the operations worth automating are the dull ones, not the clever ones: verifying a sending domain, creating a segment, importing a suppression list, and sending a campaign are each one call with an argument. Nitrosend is MCP-first, so each is a tool an agent can call before it's a screen somebody opens, and contacts are unlimited on every plan, so a list growing faster than the revenue on it doesn't change the bill.
None of it judges an offer, and none of it runs your link domain: the redirect hostname stays DNS you administer, wherever you compose the campaign. Whether a product is worth recommending to people who trust you enough to open the message stays a human decision, and it's the one that sets the complaint rate everything above is measured against.
If you're mailing offers for a living, the thing between you and an audience you built shouldn't be an account you rent. Start on the free tier: 8,000 emails to begin with, then 500 a month, unlimited contacts, and full MCP, API, and CLI access with no card. Authenticate one domain, send one campaign through it, and watch what the rate does. Nitrosend specializes in email, and this is the job it's for.
Go deeper
Sources
- Google, Email sender guidelines: the 0.30% spam-rate ceiling reported in Postmaster Tools, the 0.10% recommendation, and the instruction to merchants to monitor affiliates and remove ones that send spam.
- Yahoo Sender Hub, best practices: SPF and DKIM, a DMARC policy at
p=noneor stronger with From alignment, a spam rate under 0.3%, unsubscribes honored within 2 days, and the advice against purchased lists and pre-checked opt-in boxes. - Spamhaus, Domain Blocklist: domains in the headers and body looked up during content inspection, the return code for an abused spammed redirector domain, and the expectation that shorteners check their own outbound domains.
- RFC 8058: the
List-UnsubscribeandList-Unsubscribe-Postheader pair and the DKIM signature that has to cover both.
Common questions
Yes, and no receiving mail server treats a link as forbidden because it pays a commission. What filters do is look up the domains inside the message, so what matters is the reputation of the hostname the link resolves through rather than the fact that it's an affiliate link. A raw network URL, a public shortener, and a hostname you control all carry different histories. Whether a given platform's acceptable-use policy allows the send is a separate question from what happens once the message leaves, and the two get confused constantly.
Only if you take it. The unsubscribes, complaints, and hard bounces you've collected are your own records, so export them before you close an account and import them on day one of the new one, because nothing carries them across for you. Mailing somebody who already opted out is the fastest way to open a new platform with a complaint against it, and a migration is where that happens. Reputation is the other half of the answer: built on a domain you own it follows you, and anything accumulated against a shared IP pool stays with the pool.
If there's a material connection, yes. The FTC's endorsement guides treat monetary payment and free or discounted product as material connections that have to be disclosed clearly and conspicuously. The standard isn't the complete details of the arrangement. It's enough for a reader to evaluate the significance of the connection, and it has to be somewhere they'll see it while they're reading the recommendation. A line in the footer of a message that's one long recommendation is the shape that fails.
There's no count worth copying, because the variables are how long the message is, how many distinct offers it carries, and how many separate hostnames those links resolve through. The last one is what matters mechanically: five links to one destination you control looks different to a filter than five links to five networks. Write for the single action you want the reader to take, and let the number of links fall out of that.
Redirecting through a hostname you control is worth doing, but hiding is the wrong word for what it achieves. It doesn't remove the reputation risk in the link, it moves that risk onto the domain doing the redirect, which is an improvement only because that domain is one you can keep clean. It also means a merchant changing their tracking URL doesn't break every email you've already sent. Disguising where a link actually goes is a different act, and the reason not to is that a reader who works it out reaches for the spam button.
Yes, and it's the highest-value hour in the whole setup. Gmail and Yahoo both expect bulk senders to publish SPF and DKIM and a DMARC policy on the sending domain, with the From address aligned to one of them. Doing it means reputation accumulates against a domain you own rather than an account you rent, so changing platform later doesn't reset it. Of the three reputations in play, it's the only one you can fix in an afternoon.
It's a rule of thumb: roughly four messages that are useful on their own for every one that asks for something. The ratio isn't the useful part of it. What a reader usually means by the question is how much promotion a list will absorb, and the list answers that better than any number does. Watch the unsubscribe and complaint rates on the sends that ask, and the ratio yours tolerates will be obvious inside a month.
Often enough that the list remembers who you are, rarely enough that the offers don't become the relationship. The signals to steer by are the complaint rate against the 0.30% ceiling Google publishes for bulk senders, and what happens to opens on the sends that aren't promotional. If those slide, the frequency is wrong whatever the revenue says this week. Sending only when an offer converts well trains a list to expect nothing else from you.