Klaviyo flow filters vs trigger filters: which one decides who gets the email

Checked once at the door, or checked at every step

The trigger filter is the bouncer at the door. The flow filter is the steward who walks the room. One looks at somebody the moment the trigger fires, decides whether they come in, and then stops caring. The other checks everybody again before each email goes out, and shows people out when they no longer belong.

That's the whole idea, and everything else follows from it. Ask of any condition: does this describe the person at the moment they entered, or does it describe somebody who should still be getting this tomorrow?

What a trigger filter is for

The trigger is the event: a checkout started, a product viewed, an order placed, a list joined. The trigger filter narrows that event at the point it happens, and it's the right place for anything about the event itself.

Every row in that card names its trigger under the flow name, and they're all different events. That's the useful discipline: one flow, one moment, one job. If two flows share a trigger and differ only in their filters, they usually want to be one flow with a split.

What a flow filter is for

Flow filters answer a different question. Should this person still be here? They're checked before every message, so they're the right home for anything that can change while the sequence is running.

• Has not placed an order since starting the flow, which is the one every cart flow needs
• Is not already in another sequence you wouldn't want to overlap with
• Still consents to email marketing, which is checked for you but worth understanding
• Has not become a customer, for flows written to people who have never bought
• Is not in your suppression segment, saved once and used everywhere

Somebody who fails a flow filter leaves quietly. They don't get the rest of the emails and they don't show up as a problem anywhere, and that's exactly why nobody notices a filter that's too tight.

The reminder that arrives after the order

Somebody starts a checkout, gets the first reminder, buys an hour later, and then gets two more emails asking whether they forgot something. The condition was written as a trigger filter, so it was true when they entered and never looked at again.

It costs more than the embarrassment. Those sends go to people who were about to be your best customers, they train the reader to ignore the sequence, and complaints from irritated buyers are the expensive kind. The abandoned cart flow is the one to check first, because it's the one that leaks.

A conditional split isn't a filter

Splits get used as filters, and that's the second commonest fault we find. A split routes people down one path or another and keeps everybody in the flow. A filter removes them.

The split is also the honest way to handle a discount. Route the people who need one down a branch that offers it, instead of filtering out the people who don't and losing them altogether.

One more thing about delays. A filter isn't watching while somebody waits, so the useful mental model is a queue with a check at the front of it. That's why a long delay with no filter after it is the most dangerous shape a flow can take.

The order to read a flow in

Open the flow and read it in this order, writing each answer down. It takes about ten minutes per flow and it finds most faults.

1. Name the trigger, and say out loud what moment it describes
2. Read the trigger filters: are any of them about the person, not the event?
3. Read the flow filters: is the since starting the flow condition actually there?
4. Walk each split and check both branches end somewhere sensible
5. Check the delays, then confirm the filters run again after each one
6. Send yourself through it, buy something in the middle, and see what arrives

That last step is the one people skip, and the only one that proves anything. Every flow we rebuild gets walked this way before it goes live, and the flow library each publish the trigger and the filters we use.