In part 1 of this series, I explained the reason why we need a new tool. To summarise, there was no solution in the market that could be used in digital marketing with a true 360-degree view of the customer. In this post, I will explain the main core components of the Adobe Experience Platform.
Everything in AEP revolves around profiles. Remember, AEP is for digital marketing, where we have to be maniacally focused on the customer experience. This also means that AEP is not a CRM, an ERP or, in general, your system of record. AEP is meant to be used to know your customers and, as such, the customer is at the centre of the tool.
You could think of AEP as a huge database. With one caveat: no RDBMS could support what AEP wants to achieve, a completely different technology is needed. However, to keep things simple, I will compare it with a traditional database, as the concepts are very similar. In fact, you can even query AEP using SQL! But this is for a future post.
AEP organises all data in datasets. In case you are wondering what a dataset is (I also did not get it when I first heard about it), it is conceptually like a DB table. Imagine just enormous tables, with information about your customers. There can be many datasets; in fact, a typical AEP implementation has multiple datasets.
There are two types of datasets in AEP: attributes and events.
Attributes are profile characteristics that do not change frequently in time and that you are only interested in the last value. Some examples:
- Email address. People rarely change that and, in general, we use only one email address to receive emails from companies.
- Home address. It is common to move to new houses but, again, only your current address is relevant. You do not live in your previous homes any more!
- Birthday. Unless reincarnation exists, that attribute never changes.
- Phone number. I would even dare to say that this attribute is, today, the most important in our life, as we keep our life in our mobile phone.
So, even though these attributes change over time, we can safely ignore old values.
This is where things get exciting. As opposed to attributes, events are profile characteristics that can have multiple values, each of which will have a timestamp. These represent the interactions customers have with the company. Examples of events are:
- Website visits. If you are familiar with Adobe Analytics, this is the whole raw data in data warehouse. This incudes all interactions with the website.
- Purchases. A list of all the purchases customers have done, including the products.
- Loyalty card. All transactions where the loyalty card was used.
- Visits to stores/branches. Every time someone identifies himself in a brick-and-mortar shop.
- Emails. All email sent and received, either individually (e.g. to get support) or as part of a marketing campaign.
- Call-centre calls.
I hope that it is now getting clearer why a new technology was needed. To put things in context, one of my Analytics customers had a data feed of 5GB/day… compressed. Now, combine this information with all the other data sources you can have and try to imagine the scale at which the technology must work.
Another core component of AEP is the identity service. Each profile dataset can have its own primary key:
- Email address is a typical primary key.
- Your CRM software will generate its own CRM ID.
- Website visits will use the ECID.
- Purchases may use a variety of primary keys:
- Online transactions will have the ECID and, maybe, a CRM ID or a loyalty card number.
- Offline purchases may have a loyalty card number, but only if the customer has one and presented it at the till.
- Calls to the call-centre will know your telephone number.
So, how do we know that the same person has purchased offline, called to register the product and visited the website? The answer is conceptually simple: if a record contains 2 or more different keys, AEP can keep track of that and start linking records. And this is what the identity service does: link all IDs corresponding to a single customer. With this information, AEP can then go to the datasets and find all interactions of a customer, using all the IDs.
Example of identity service in action
Let’s consider the following events:
- A customer signs up up for the loyalty card. During the registration, she provides an email address.
- Some emails are sent to her with product promotions.
- She clicks on an email link with a product she likes and reviews it online.
- She goes to a shop and purchases that product. At the till, she presents the loyalty card to accumulate points.
AEP can now link all activities together and attribute the purchase to the email and to the website content. How is this even possible?
- During registration, the loyalty card number and the email address are linked.
- When clicking an email link, the email address and the ECID are linked. At this point we have loyalty card, email address and ECID interlinked.
- During the website visit, all interactions use the ECID but, through the interconnections just mentioned, can also be attributed to an email address and a loyalty card.
- At the till, the loyalty card is associated with the purchase. Again, with the previous links, the purchase can be associated with the visit to the website and the click on the email link.
The following image should (hopefully) clarify this example:
Putting it all together
If you have been following until this point, I hope you have already understood the power of AEP. By querying the datasets, using the identity service as the glue, you can create very sophisticated segments. These segments can be defined based on any information you have about a customer. The example I gave in my previous post about the home insurance is now possible in a single place. You can also go beyond segmentation and calculate next best offers for each of your customers, using ML algorithms.
However, you may be wondering why so much hype about just this. And you would be right if AEP did not offer anything else. Remember that this post explains the boring foundations of AEP. I wanted to start with them before I go to the next step.
Not only is AEP a data-in system, but also a data-out one. Once you have created your segments or next best offers, you can send them to multiple destinations to action upon the data. Adobe is actively working on integrating all its solutions with AEP. Besides, you can also send them to any 3rd party system.
Some corporations, with an advanced data science team, may have enough with these foundations. They already have all the systems needed to exploit this data and all they need is an API to query the data.
However, most AEP clients will want to use the services that Adobe is building on top of these foundations, which is where the magic really happens. These services require their own post, but let me summarise some of the most common services:
- Customer Journey Analytics (CJA). If you are familiar with Adobe Analytics Workspaces, now imagine taking them to the next level. With that same UI, you can now create reports and analysis based on all the datasets, not just web traffic.
- Real-Time Customer Data Platform(RT-CDP). To keep the description simple, think of this services as AAM on steroids. Since AEP has all the data you need, it can action on changes in real-time.
- Journey Orchestration (JO). For those of you familiar with Adobe Campaign or any marketing automation tool, this service.
- Data Science Workspace. Instead of using your own servers to run you AI/ML algorithms, use Adobe’s integrated solution.
I hope that, by now, you have a clear understanding of the origin of AEP and what it does. Any questions, let me know below.
Image by rawpixel.com