What does the new graph-based identity stitching mean for CJA Webinar Re-Cap

Back in March Stephanie’s Adobe Summit session flew under my radar. But when I learned about it, I knew we had to host a webinar to get the word out and get ready for this new feature. By the way, Graph-based stitching for CJA goes to general availability by the end of June!

This much awaited feature from Adobe will help solve one of the most difficult challenges in achieving a truly unified customer journey analysis. If you were not sure that CJA was worth it yet as a marketer, the new graph-based stitching unlocks a whole new realm of insights that was previously reserved for data science / data engineering resources. You now have access not just to people-centric analyses with a consolidated customer profile across devices, but now also access to experience-centric analyses across customer channels.

It was incredibly exciting to host this webinar with Stephanie and Matt and cannot wait to put the learnings into action. Below is a recap of what we discussed, but please don’t miss out on the recording itself.

Why is unified experience important?

*84% of customers say the experience a company provides is just as important as the product or service itself. When customers experience friction, *96% of those customers are going to stop purchasing with you. The only way to fully understand where friction happens is by analyzing a fully stitched journey. Companies that routinely and consistently provide a seamless experience can charge a *16% premium on their goods or services.

*Market research performed by Adobe – https://business.adobe.com/summit/2024/sessions/cjas-secret-sauce-infusing-the-identity-graph-for-s113.html

Why Stitching?

The various places a customer interacts with you comes across as different data points on different data sets, causing an identity crisis. You need find a way to overcome this with an identity resolution approach. Identity resolution is when you take these different identities and group them together to consolidate them as one identity, or a single person, across all the sources. Identity stitching is when we join all those identities together, which coupled with CJA’s ability to sequence the overall events across data sets in the order that they all happened will help your company have more accurate and complete information about customer.

There are two ways of stitching identities.

Field Based Stitching

This has been in practice for several years with CJA. In this approach you rely on a “golden person identity” that should exist in every data set you are trying to connect. The problem is not every data set has the same ID available. Many times, the biggest culprit is digital data. In digital you may have a Person ID but most often you have a cookie ID which is a large source of traffic because people are not always telling you who they are right away while they are browsing. We try to accommodate that with field-based stitching.

If you were to use a Cookie ID to connect with other data in our CJA connection, wyoue will miss out because other data sources don’t have this cookie ID. Conversely, if you use the Person ID as our connection, we will miss out on unauthenticated traffic. We need to see all the data to understand and see what our customer is experiencing. Therefore, field-based stitching is how we get the best of both worlds.

Jump to 9:38 to hear the explanation on how this works.

Pros and Cons of Field-Based Stitching

Pros:

• Allows for both authenticated and unauthenticated data to exist.
• Able to connect customers to more offline or outside channels of engagement.
• Supports shared devices.
• Attribution is carried across devices.
• Restatement of historical data allows for greater accuracy in analyzing the customer journey.

Cons:

• Stitching does not transform the field used for stitching in any manner. Stitching uses the value in the specified field as it exists in the unstitched dataset within data lake.
• The stitching process is case-sensitive.
• Replay window limited to 7 days.
• Only solves for missing identification on events in one single dataset.
• Reliant on authentication event to occur within a single dataset.
• Still requires every dataset to have that same golden ID present.

Go to 12:33 in YouTube video to hear this explained.

Graph Based Stitching

Similar to how field-based stitching works, Graph-based stitching also generates a copy of the dataset and it will append new columns in the data for the stitched ID and the Stitched ID Namespace. You also have the ability to replay data that will restate the data tied to a stitched identifier. The difference here is even though the original digital data set may not have email, CJA is making a reference outside of the data set to the identity graph, the identity resolution, to be able to identify this person. You don’t need to wait for the authentication to happen online, you can see in the first few rows you can use the cookie ID to go to the graph and find the association.

To see an example of how this works, jump to 18:30 in the YouTube video.

How do we use the graph?

When using the graph, you will first decide what data you want to connect, and how do you want to connect it. Think of a “Golden ID”, or the most durable first party ID you have, that you will have across as many sources of data that you can. The second question you will ask, is what data do you want to bring in?

To see how to consider these decisions and what it will look like, jump to 23:15 in the YouTube Video.

How does CJA compare with Cross Device Analytics?

Hear Stephanie give the full explanation and breakdown of the similarities and differences between utilizing CDA versus customer journey analytics at 27:44.

Pros & Cons of Graph Based Stitching

Pros:

• Connect and view the holistic journey of customers across channels.
• Improve identification rates using a cluster of identities known in a graph.
• Join more data sources together without requiring additional ETL processes to append missing identity fields.
• Have the same consistent understanding of a “person” across all AEP app services.

Cons:

• GBS won’t support shared devices on first release.
• No identity map support; ECID/any namespace will need to be in other fields.
• UIS limitatons of 50 IDs per graph.
• Only one ID can be chosen as the first golden ID for stitching upon first release.

Hear the full list of pros & cons at 32:25.

How does GA4 do Identity?

In short, GA4 doesn’t. The level of identity stitching available in GA4 is limited to field-based stitching at best. Its most similar to Cross Device Analytics. There is a possibility of uploading offline data, but it’s limited to user attributes rather than entire data sets.

To stitch full data sets similar to how CJA can, you can do it in BigQuery as long as you have the resources and data engineering help to support. Where CJA differentiates itself is by giving the power back to the marketers in doing the sticthing and reducing the overall time-to-insight.

Hear Jean-Paul’s full synopsis at 36:10.

Identification Rates

One of the biggest things to note with graph-based stitching is the impact it has on reporting and identification. Using this method, you can more easily identify customers and improve your reporting authentication percentage. The second callout is the value of being able to bring that data together. Before graph-based stitching you would only be able to pull sources together if you had an email attached to it. Our main goal in general is to create a better customer experience, with graph-based stitching we can pull together touchpoints and integrate experiences to more accurately understand our customers’ journey and in turn improve their overall experience.

Hear all the callouts at 41:13.

Question & Answer Section begins at 46:15