Understanding flow branching

You will learn

Learn how flow branching helps you personalize flows so that you can send individuals down different paths depending on what you already know about them.

Flows allow you to automate timely touchpoints with your contacts without losing any of the relevance and personalization required to build an impactful marketing strategy. For example, an abandoned cart flow doesn't just send a generic follow-up email to all cart abandoners — using Klaviyo's default abandoned cart flow, all shoppers will receive an email personalized with each item they left behind.

Flow branching helps you take this personalization to the next level. Klaviyo supports multi-branch splits, which route profiles down different paths — up to 20 in a single split. Learn more in the multi-branch split help article.

The multi-branch split

With a multi-branch split, you define conditions for each path using the same logic you see in our segment builder. In this way, splits look similar to flow filters. Unlike flow filters and segments, however, splits don't keep certain people in your flow and filter the rest out — everyone continues through the flow, just down different paths.

In the visual flow builder, splits are a configurable component. A split creates multiple paths — up to 20 (19 you configure, plus an "Everyone else" path). On the canvas you'll see a path card for each. Klaviyo evaluates the paths in the order you arrange them — Check first, Check second, and so on, ending with Otherwise — and sends each profile down the first path whose conditions it matches. A profile that doesn't match any configured path goes down the "Everyone else" path.

There's a single split action. Each path is defined by conditions you build in the filter builder. You can branch on profile data — properties, activity, location, list membership, and more — and, in metric-triggered flows, on data from the flow's trigger event.

Best practices

Abandoned cart flow

Let's say we'd like to create unique paths for those that have over a certain value left behind in their cart.

For those who left behind high-value carts, you may want to offer a greater incentive for recovery — for example, free shipping, or a percentage off their order if they complete their purchase. For shoppers that left less valuable carts behind, you certainly want to follow up, but you may not want to offer a discount given the expected ROI.

By adding a multi-branch split into an abandoned cart flow, you can immediately create 2 paths for cart abandoners based on their cart. This data point (cart value) is captured in the Started Checkout event that serves as the flow's trigger.

Post-purchase flow

Let's say we'd like to create unique paths for first-time buyers and repeat purchasers in our post-purchase flow.

When someone buys from you, this is a great time to reach out and strengthen your relationship.

When a customer buys from you for the first time, it is important to recognize they've just gone through a process of discovering your business, evaluating your products, and making a decision to purchase from you. Follow-up that validates this choice and fosters a positive customer experience is essential.

Those who place an order but have already ordered before in the past are farther along in their relationship with you. In the customer lifecycle, you're working to retain them and inspire them to become not only loyal customers but also loyal promoters and advocates. Your content in a post-purchase series should be different for these repeat purchasers than for your first-time buyers.

To treat first-time buyers, repeat buyers, and everyone else differently, use one split with three paths — Check first: placed 0 orders; Check second: placed 2 or more orders; Otherwise: Everyone else. You no longer chain multiple splits to make more than two groups.

Tip: to create more groups, add more paths to the same split (up to 20) — don't place another split under the "Everyone else" path to get extra branches. And because profiles take the first path they match, put your most specific paths first.

Browse abandonment

In your browse abandonment flow, you may want to create unique paths for those who viewed under 3 products in the last several days and those who viewed more than 4 different items.

There are 2 key types of visitors to your website: those who know exactly what they want when they get there, and those who are there to browse. Targeting your site visitors with the right messaging based on their browsing behavior can yield high returns.

Those who only visited a single product page may have arrived through a direct search or otherwise knew how to land right where they wanted — your follow-up should focus mainly on that specific item, with the goal to get that shopper to come back and buy it (or a similar recommended item).

Those who viewed over 4 items, on the other hand, may be window shopping. Don't focus as much on the most recently viewed individual item, but use this as a chance to promote your most popular items, show more of a specific collection, or provide social proof for why this shopper should come back and take a second look at your store. The stakes are higher to spark interest in what you have to offer more broadly, as the visitor hasn't indicated they know exactly what they want.

If you create a browse abandonment series and add a multi-branch split to target these 2 primary types of visitors, you can put your data to work early to drive deeper engagement starting with the very first page view.

Troubleshooting

If a flow branch is not behaving as expected, such as if profiles are going down the wrong path, check the following:

Make sure that the split has a condition set up

An unconfigured path shows an "Add path criteria to {path name}" alert and is skipped at send time. If no path is configured, every profile goes down "Everyone else." Configure each path's conditions before setting the flow live.

Click on the split to edit its condition in the settings sidebar.

Make sure there is enough time for the split to check the profile

If you're encountering problems where your split fails to evaluate actions like profile clicks or data additions such as custom properties, it's often because there isn't sufficient time between the action processing or data being added to the customer profile by the time the customer meets the split.

In order to make sure there's enough time before your customers are evaluated by the split, you will need to add a time delay right before the split. For example, if you want to add a split to filter customers who clicked a link in your email, there needs to be a long enough time delay to allow the customer a chance to click. If the customer clicks after they were already evaluated by the split, they will go down the path for customers not clicking the link.

If you are seeing profiles going down the wrong path, we recommend reviewing when the customer took the action or when the data was added to their profile relative to when they went through the multi-branch split. To view when the customer went through the split, click the multi-branch split and select View details in the top right. Next, open the profile you want to review and compare the timestamp of when they took the action you are evaluating. If you are reviewing when data was added to the customer profile, such as a custom property, you will need to compare when the action took place that added the property to their profile.

Make sure that the paths have the correct messages

Drag and drop individual messages into the correct path. If a split has exactly two paths and all messages are in the wrong path, you can swap them by clicking the action button on the right of the split and selecting Flip split.

Make sure that the condition of the split says exactly what you intended

Click on the split to view the settings sidebar. Review the split conditions to confirm that it is set up to check for exactly what you intended. For example, if you want to check if someone has placed an order at any point in the past, make sure the condition is set to over all time and not in the last 30 days, which is the default setup.

Additional resources