Understanding time delays near splits
You will learn
Learn when to place time delays before a split, and when to place them after.
When placing time delays near a split in a flow, it's important to be mindful about where you place the delay, as it can dramatically affect how a flow behaves.
Time delays and splits
A split lets you send recipients down different paths based on a definition you establish. A path's conditions can use profile properties and activity, and — in metric-triggered flows — data from the trigger event.
After you drag in and set the condition(s) for your split, it functions as a decision point — when a recipient reaches the split, Klaviyo evaluates the paths in order and sends the recipient down the first one they match, or down "Everyone else."
The time delay before a split matters because a split evaluates each profile once, at the moment it reaches the split — a point-in-time evaluation.
For example, let's say you have a split that is defined as: has Placed Order zero times over all time. A non-purchaser matches this path; existing customers go down "Everyone else." It's possible, at any point, that someone in this flow may purchase from you and become a customer, so the timing of when this split is evaluated in your series is important.
Place time delays before a split
A split divides recipients based on their actions or profile properties. If you split recipients by their actions, you need to allow time for them to take those actions. Further, if you intend to place messages on more than one path of a split, we usually recommend setting these emails or SMS to send immediately after the split itself by not adding time delays directly after it. This way, right after your split is evaluated, recipients will receive a relevant message. If you put an additional time delay after a split, before a message, it's possible that, after this additional wait period, the recipient will no longer qualify for that path.
Conditions based on the trigger event don't need a wait — that data is available as soon as the profile enters the flow. (Trigger-event conditions are available in metric-triggered flows only.)
In the below example of an abandoned cart flow, the goal is to send shoppers an email four hours after they initiated a checkout if they didn't end up completing the order. Here, we want to create different paths based on whether or not this cart-abandoner has ever purchased in the past. To achieve this, set the time delay component to wait four hours after the trigger. Before sending the first abandoned cart reminder email, add a split to evaluate whether or not someone moving through this flow has bought before. After four hours, someone will either get email #1a or #1b depending on whether or not they bought at least once in the past.
Placing time delays after a split
While typically it's better to add the time delay before the split, there are some cases where you may want to have a delay after. The most prevalent example is when you're testing the timing of a message.
For example, say you have an abandoned cart flow and you want to determine whether sending the first message after 3 or 4 hours is better. In this case, create a split with a 50% random sample condition on the first path. Then, add a time delay to each of the paths, one set to 3 hours and the other to 4. Note that the first message should be exactly the same; when A/B testing, you should only test one variable at a time. With this setup, you'll be able to see how the two timings compare side-by-side.
Additional resources