How to create an email course in Kit, and what to check

The line between a broadcast and a sequence

A newsletter-native tool holds two quite different things side by side: a broadcast, written once and sent to everybody now or at an hour you pick, and a sequence, written once and sent to each person in turn from whenever that person entered it. A course belongs entirely to the second, so the first thing to establish in a Kit account is which of the two screens you are composing in.

It matters because both produce five emails on five mornings and only one keeps producing them. Five broadcasts a day apart serve the people on the list that week; everybody arriving on Wednesday starts at day three of something whose beginning they never saw. If the two screens are hard to tell apart on the way in, ten minutes spent telling them apart is cheaper than a month of readers starting in the middle.

The second thing to establish is what one person is. A course wants one record per human with one run of the sequence attached, so a reader who writes in on day three gives you one place to look. Add an address that is already on the list through a second form, and see whether one record or two comes back and whether the sequence starts again.

Tags, forms, and the doors they quietly open

A tool of this kind usually starts a sequence from something that happened to a record: a form was submitted, a tag was applied, a link was clicked, a purchase was recorded. That is flexible, and it is also the part that surprises people, because a tag is applied by more things than the form the reader filled in.

So the question to ask in Kit is not only what starts the course but what else is able to start it. If the course begins when a tag appears, an import with that tag in a column begins the course, and so does a bulk edit somebody made while tidying up on a Friday. Ask whether the confirmation step is bound to the form or to the record: an address imported with the starting tag already on it may reach day one having confirmed nothing.

The same mechanism gives the unsubscribe question a second half. Ask what an unsubscribe removes: the tag, the sequence, the list, or the person. A reader taken out of a sequence and left on the list keeps receiving broadcasts after asking you to stop, and that second message is the one that gets reported rather than unsubscribed from.

Where to look, and what to look for there

Labels move between releases and screenshots go stale, so what follows names areas rather than buttons. If a name has changed since the date above, the area it describes is still the area to open.

The forms and landing pages area answers the confirmation question for one door only. Find the setting that decides whether a new address is asked to confirm before anything else, then go and find out what the other doors do: an address added by hand, an address arriving through the API, a spreadsheet import, and whatever an integration with a shop or a webinar pushes in.

The sequence area answers the clock. Look for what the first step waits on, and what each later delay is counted from: the reader entering, the previous message actually leaving, or a date. Look for a send window or a fixed hour, and for whether a time zone is read from the reader or from the account. Then look for what happens when the same person enters the sequence twice.

The account settings answer identity: the From name, the address a reply reaches, and the domain the mail is authenticated against. Settle the last one first. Ask what has to be added to DNS, because that answer is the one a later upgrade rarely changes.

If you already use Kit, test these four things first

A list that is already there is a real reason to stay, and the fastest way to find out whether staying works is to behave like two readers for two days. Use addresses you own at a mailbox provider you do not otherwise use, and put a real lesson behind the form rather than a message saying test.

Four tests, in the order that finds problems earliest:

• Join today with one address and tomorrow with a second. The second should receive day one, and that single result settles the clock.
• Apply the thing that starts the course to an address by hand, then apply it again, and count what arrives.
• Import a one-row spreadsheet already carrying it, and see whether anything asks that address to confirm.
• Unsubscribe on day three with the mail client’s own button, then send yourself a broadcast and see whether it arrives.

What no settings screen will show you

Some of this is invisible from inside any account, and a page that pretends otherwise is selling something. A send is a request to a mail provider, and the failure that decides whether a tool is trustworthy is the request that returns nothing after the message was already accepted. Sent late is an inconvenience; sent twice is a complaint.

Two questions about it are worth asking in writing rather than hunting for in a menu: is a duplicate send prevented by the database or by the sending code, and what happened the last time the mail provider timed out mid-send. This page has no way to answer either, and neither has any other page written from outside the building.