Graph-based stitching overview

Transcript

Hi everyone, this is Matt Thomas from the Adobe Analytics Product Management team. I wanted to introduce you to graph-based stitching for CJA. This feature harnesses the identity-rich graph that is stored in AEP and used by several solutions including RT-CDP and AJL. To get things started, I’m assuming that you are familiar with the current offering of shield-based stitching. With that assumption, let’s pick up where we left off on Corey’s journey. This journey is very common. Corey recently had a baby and had some paternity leave. He has some home projects that need to be done but realizes he needs a few tools first. Since he holds the baby a lot, he can do some initial research from his phone. As he narrows the products down, he goes to his laptop to do some final comparisons. He purchases most of the tools online but realizes he needs one right away and decides to head into the store. After purchasing, he gets home and unboxes it and realizes that a piece is missing. While he can use the tool and get the job done, he calls in to the support center and requests a replacement part be sent. After his interaction on the phone, he has sent a survey which he completed. When bringing this data into AEP, typically it’s in a dataset and it’s centered around a single or primary identity. And it most likely is not common with all of the other ones in the sandbox. Even with field-based stitching today, we could elevate a single dataset with identities that are found into a higher identity. This still may not be enough. As you can see here, this web and mobile dataset were elevated to email which allows us to connect it to the voice of customer data. But that’s where it stops. When you go to make that connection in CJA, it is required that every row have a person ID and to get the full benefit of CJA, it really needs to be in common across all the datasets. To make this possible, you start to share identities across datasets. As we know, authentication or the ability to know a person doesn’t always happen just in a single channel. So by enabling identity services on a wide range of datasets, the graph starts to build these identities and the relationships they have into a graph as seen here in this visual. In our example, we take five datasets and ensure that all identities that are found in them are marked on the appropriate schema. Whether it’s an email or an ECID or credit card, we need to mark them as identities on the schema. Once they are marked and the dataset is enabled for identity services, then they start to contribute to the graph. As you can see here, each of those different identities start to flow into the graph. Now the stitching process comes into play so we can help to elevate these datasets to a preferred or golden identity. In this case, we are using email and as you can see, the CRM voice of customer datasets are already aligned to email, so no stitching is needed. The web call center and point of cell datasets, however, do need some elevation. We take the predominant identity found in each of these referred to as the persistent ID and basically do a look up against the graph. For instance, we have a credit card and point of cell dataset. We go and look up from the graph and retrieve an email. We take this email and plug it in as the stitched ID column. Let me walk you through an example. As you can see, we have a web dataset and a very well built out graph. The graph may not start out this way, but over time it can resemble something like this. As the first value comes in, we look up that device ID against the graph and resolve to the preferred namespace, in this case, which is CRM ID and we plug Cori into there. We repeat this process for additional rows. On this third row, it becomes a little interesting because we have an email address on the row itself, but we again take the device ID we have and look it up against the graph and we know that the CRM ID is the preferred identity, so we plug that in. So now we’ve added an identity to this dataset that it didn’t have any reference to before. We repeat this process through all of the different rows in the dataset. And as you can see, we’ve appropriately assigned multiple rows of data to the right person. Now the last step is really making the connection to these datasets. You simply make a connection specifying the correct person ID, which in this case is stitched ID for the ones that we stitched and email for the other two. Now all of the datasets are aligned together around the right identity and richer analysis can be performed. Hopefully you have a better understanding of graph-based stitching and how it can help you in your customer journey analysis and activation. Thank you.