A step by step order for cleaning an email list

Start with rules that need no network. The SMTP specification sets the maximum length of a local-part at 64 octets and of a domain name at 255 octets ¹. The Internet Message Format defines the local-part as a dot-atom or a quoted string, and says the dot-atom form should be used and the quoted-string form should not ². A dot-atom is a run of allowed characters separated by single dots, so a local part with a doubled dot, a leading dot or a space is not a normal address.

Our own input check is stricter than the grammar in two ways worth knowing: it rejects an address longer than 254 characters and any local part over 64, and it does not accept quoted local parts or domain literals such as a@[192.0.2.1]. Those rows come back as invalid_input, and a row that fails this way is not charged.

Deduplicate only after normalizing, or near-duplicates survive. Trim whitespace and lowercase the domain. The local part needs a decision: the SMTP specification says a local-part must be treated as case sensitive, and also says that exploiting that is discouraged ¹. In a marketing list, lowercasing the local part is the practical choice, because two addresses that differ only by case almost certainly reach the same person. Our check lowercases both parts before it looks anything up.

• Keep the first occurrence of each normalized address and record where the others came from.
• Do not strip plus tags or dots from the local part: that is a rule of some providers, not of the standard.
• Keep the original value in a separate column so you can trace a result back.

The domain step asks whether the part after the @ can receive mail at all. A domain that publishes a null MX is announcing that it accepts no mail ³. The email-domain-check product reads the domain's MX, SPF and DMARC records and open disposable-domain and free-provider lists, and sets deliverability_signal to good, risky or undeliverable. It does not contact the mail server, so it cannot confirm that a mailbox exists.

Role accounts such as info@ or support@ belong to a team rather than a person. The same check sets role_account when the local part matches its role list, which lets you route those rows to a separate segment instead of deleting them.

The dashboard runs a CSV through a product in the background. The limits below are read from the same configuration the service enforces. See the bulk CSV jobs guide for the full list.

1. Prepare a CSV with one email column that has already passed steps 1 and 2.
2. Split the file if it is over the row or size limit.
3. Open jobs in the dashboard, upload the file, choose email-domain-check and the email column.
4. Start the job and download the results when it shows completed.
5. Filter on deliverability_signal and role_account, then send to the rest in small batches.

1const seen = new Set();
2const keep = [];
3for (const raw of rows) {
4  const email = raw.trim().toLowerCase();
5  const at = email.lastIndexOf('@');
6  if (at < 1 || at === email.length - 1) continue; // no local part or no domain
7  if (email.slice(0, at).length > 64) continue;    // local-part limit
8  if (email.slice(at + 1).length > 255) continue;  // domain limit
9  if (seen.has(email)) continue;                   // duplicate after normalizing
10  seen.add(email);
11  keep.push(email);
12}

Before you spend a job on a file, try the check on a single address in the free email domain checker to see the fields you will filter on. When the list is clean, set up the same call for new signups through the email domain check product page.