How One Customer Story Proved the Power of Adobe Experience Platform to Meet Customer Experience Needs

[#1]

Authors: Sunish Verma and Jody Arthur

This post provides a synopsis of a proof-of-concept that Adobe Experience Platform developed for a large B2B company to meet several business needs that are common across almost all industries. Our developers solved every one of the customer's uses and in the process, proved that Adobe Experience Platform's architecture can stand up as the core foundational platform for delivering customer experiences, one that augments all our other channels and products and supports their use at high scale.

This blog details a single implementation or customization on Adobe Experience Platform. Not all aspects are guaranteed as general availability. If you need professional guidance on how to proceed, please reach out to Adobe Consulting Services on this topic.

In 2019, one of our enterprise customers, a large B2B company, presented Adobe Experience Platform with a set of problems it wanted to solve. Unable to resist a challenge, our pool of enterprise architects and developers quickly got to work on developing a proof-of-concept (POC) for the customer. When they were finished, their POC not only illustrated the power of the platform to meet the company's needs but also how well Adobe Experience Platform's architecture can stand up as the core foundational platform for the company, one that augments all our other channels and products and supports their use at high scale.

At Adobe, a POC is a scaled-down, expedited engagement with clearly defined success criteria that is meant to prove the technical capacity of Adobe Experience Platform in support of the customer's highest-value use cases. In just four months, our developers were able to meet every one of the company's use cases within Adobe Experience Platform.

At the time, Adobe Experience Platform was evolving quickly to become the first purpose-built customer experience platform that it is today. As a platform, we did not yet have all the products and services we now offer. So, in developing our POC, we used a crawl, walk, run approach to developing the foundational architecture necessary for Adobe Experience Platform to meet the company's needs.

With this in mind, we worked with the company's key use cases and its customers' journeys, and the capabilities of Adobe Experience Platform architecture to meet them.

The company's high-level business objectives are common to many enterprises today (Figure 1). Successful businesses today understand the importance, and promise, that data-driven decision-making has for driving business results. The company chose Adobe Experience Platform because it wanted a platform that could work with its data to develop a true, 360-degree view of its customers allowing it to offer more personalized experiences to its customers through cross-selling and multichannel upsells.

Figure 1. The company’s high-level business objectives it wanted to meet using Adobe Experience Platform.

Our POC had many benefits for our customers (Figure 2). Through its development, the company was able to leverage its online and offline data to achieve client level unification, a richer and more accurate Unified Profile, which in turn allowed it to better leverage platform capabilities to meet its segmentation and activation goals:

• Customer propensity – At the time, the company was doing some segmentation in which it was leveraging some of its data in Adobe Campaign and some from its traditional databases. However, by working within Adobe Experience Platform architecture, the company was able to use multi-entity features and much more advanced segmentation.
• Customer engagement and retention – Prior to our POC, the company was able to activate through each of its three primary channels, its website, mobile app, and email and to do some reporting. Through our POC, we were able to find several efficiencies in how the company could standardize it's messaging and engagement, making it visible could be standardized and made visible across all channels.
• Customer churn – Perhaps the biggest win for this customer was the ability to add machine learning and predictive modeling to its marketing toolbox. At the time, the company didn't have a set up for machine learning and running predictive models. Now, with the model we developed in Data Science Workspace, the customer is able to be far more proactive in its marketing across the board.

Figure 2. The company’s key use cases, shown in terms of their ability to meet them prior to the development of our POC for the company and its new capabilities with the foundational architecture we built for the POC.

Figure 3 illustrates the components of Adobe Experience Platform's foundational architecture that we used to develop our POC. It also shows the new capabilities the POC provided the company as well as additional capabilities the company wants to implement in the future.

Ingestion of several data sources, building the Unified Profile, Query Services, Data Science Workspace, advanced segmentation and the ability to export segments for activation are all key components of Adobe Experience Platform. Integration between Adobe Experience Platform and Adobe Campaign will depend upon the campaign hosting. Products such as Microsoft Power BI (as shown in Figure 3), Tableau and others are supported within the platform. And now, with Adobe Real-Time Customer Data Platform, there are more activation destinations supported than what were leveraged at the time this POC was developed.

Figure 3: Components of Adobe Experience Platform’s foundational architecture used to develop our proof-of-concept.

Customer cases that tested the power of our platform

In order to meet these objectives, the company identified a number of specific problems, use cases it asked Adobe Experience Platform to solve. The company already had some data in Adobe Experience Platform as well as other data it wanted to add to the mix in order to better understand its customers to more effectively and proactively market to them.

Segmentation

Our POC leveraged the company's data to create different segments to address a number of key problems the company needed to solve in order to improve its bottom line:

• Customers' purchase history and browsing affinity – Looking at previous online and offline records for a customer, we were able to see that there were customers who had previously purchased products in one category who had also searched in another, a different category. Here, the question was, how do we find this audience so that we can deliver the right messaging to drive more sales in that second category?
• Customers' purchase frequency on the web – For this case, we needed to categorize customers based on their industry and data on how often they buy and/or how much they buy to develop thresholds for each industry within the company's customer base. With these thresholds in place, the company was able to identify those customers in each industry whose purchases are below the threshold versus those whose purchases are above, which allows for far more effective marketing.
• Anonymous users that convert to known customers – Every data-driven business operating in the Customer Experience Era business knows the challenge inherent in resolving customer identity. Existing and potential customers may browse a company's website without logging in, making their browsing data unattributable to any specific user or profile. Our POC provided the company the ability to recognize where the products that anonymous users are looking at on its website are similar to those that other, known customers previously purchased. By matching up these two types of data, the company can create a segment of pseudonymous customers. We don't know exactly who they are, but the company now knows enough about them to more effectively market to their needs.

All of the questions driving the creation of these segments revolved around the company's desire to provide more relevant experiences for its customers. Answering them involved the marriage of online and offline data and building intelligence from the available data.

Activation

At the time we were developing our POC, Adobe Experience Platform was just beginning to develop some of the capabilities that our users can now take for granted. One of those was the ability to export segments and the Unified Profile. Today, Adobe Experience Platform customers can activate segments leveraging the Adobe Real-Time Customer Data Platform and Journey Orchestration. However, when we began developing the POC, there still weren't many ways to activate segments within Adobe Experience Platform. Given this, the company needed a way to export a file containing all the users that qualified for a given segment for use in their marketing automation tool of choice.

To address this need, we developed a custom export utility, which we call the Segmentation Automation Workflow (SAW). The SAW can export segment files and the company's Unified Profile out of Adobe Experience Platform and drop them into a common secured location where they can then be imported into other tools.

The customer's biggest win – solving for churn

Perhaps the most important use case we were presented with was the problem of churn. Based on a customer's purchase patterns from month to month, the company was often able to predict how much the customer would purchase in the following month. However, when that didn't happen, it was often 3-4 months later when the company would see that in its data, which is of course, too late to effectively act on that information.

Churn is a universal problem for almost all businesses. The key to solving it is having the ability to predict what customers will buy and when. In Adobe Experience Platform Data Science Workspace, machine learning makes that now possible with the ability to leverage the data in platform to develop predictive models based on a Unified Profile.

At the time, our Data Science Workspace was still quite a new service, and the company was one of the first Adobe Experience Platform enterprise customers we worked with on predictive modeling. For our POC, we used Data Science Workspace to build a churn model for the company to identify customers who have purchased items in a specific product category according to an established rhythm but who are falling off that cadence, either in terms of the total dollars spent or the frequency of their purchases.

The problem with churn introduced two questions. How do we find the customers that are dropping out more quickly so we can act on that information? And, how do we identify the common attributes of those customers so that we can be proactive in our messaging to prevent the churn? Our model answered both of these questions allowing the company to develop more timely and relevant offers, significantly reducing the churn it had previously been experiencing.

Our developers built the POC within the company's provisioned environment and with the company's existing data with the goal of illustrating how Adobe Experience Platform could help the company accomplish its business goals.

We started with data integration. The use cases presented required the ingestion of several different data sources, including Adobe Analytics, customer invoice data (with item data), customer master with firmographics, call and chat logs, and other types of client-specific data to create a Unified Profile for the customer. We used the Analytics data in order to better leverage the data already in the platform and brought the other data from the company's offline sources. The customer invoice data was the largest data set in terms of volume but also provided much of the insight we needed to address the company's use cases.

Bringing the company's data into Adobe Experience Platform where it could be used to do advanced segmentation allowed us to meet a number of the company's segmentation and activation goals. To develop the churn model, we used the company's invoice data within Data Science Workspace to determine what products are being purchased by each segment and when those purchases were occurring. We used a sample of 50,000 records across each month for a nine-month period for our analysis.

After some initial data cleansing to remove outliers, we looked at the drop in orders over a specified time period to determine when the churn was occuring in the sales cycle for the customers. Our basis for analysis was customers making any purchase in the first four months of the nine-month timeframe. For the purposes of this analysis, we defined churn as customers who made a purchase in the first four months but did not make a purchase from the fifth month onwards. These results are shown in Figure 4, and the results for the corresponding drop in revenue are shown in Figure 5.

Figure 4: Results for our analysis of the drop in orders from time periods 1–3 (the first three months).

Figure 5: Results for our analysis of the drop in revenue from time periods 1–3 (the first three months).

We used 80% of the profiles for training the model and the remaining 20% for testing its accuracy and robustness​. The machine learning technique we used was the Random Forest​ method, which gave us 78% accuracy (Figure 6).

Figure 6: Results from our churn model using the Random Forest Method.

Based on this model, we were able to identify the key attributes impacting churn for the company (Figure 7). Using the model, we found that purchases made in the first month were the most important predictor of churn, and purchases made under product category “O” were the most likely to fall off. Now, with better insights earlier in the game, the company is now able to focus its messaging on those areas to create relevant and timely offers to customers in different segments to keep them engaged and buying beyond that fourth month.

Figure 7: Key attributes impacting churn for the company (anonymized for this post).

New capabilities that drive better results

At the time, Adobe Experience Platform was evolving quickly to become the first purpose-built customer experience platform that it is today. Even then, the platform's core foundational architecture was more than capable of supporting the customer's needs and has only gotten stronger since then. Future blogs will explain our integration of this and other integrations with Adobe Real-Time Customer Data Platform. With the continued development and release of new products and services, Adobe Experience Platform is helping companies better leverage their data to drive business results through better customer experiences.

Follow the Adobe Experience Platform Community Blog for more developer stories and resources, and check out Adobe Developers on Twitter for the latest news and developer products. Sign up here for future Adobe Experience Platform Meetups.

Resources

1. Adobe Experience Platform – https://www.adobe.com/experience-platform.html
2. 360-Degree View of the Customer – https://www.adobe.com/experience-cloud/use-cases/customer-intelligence.html
3. Unified Profile – https://www.adobe.com/experience-platform/real-time-customer-profile.html
4. Adobe Campaign – https://www.adobe.com/marketing/campaign.html
5. Machine Learning – https://www.adobe.com/sensei/ai-innovations.html
6. Adobe Experience Platform Data Science Workspace – https://www.adobe.com/experience-platform/data-science-workspace.html
7. Adobe Experience Platform Query Service
8. Advanced Segmentation – https://www.adobe.com/analytics/advanced-segmentation.html
9. Microsoft Power BI – https://powerbi.microsoft.com/en-us/
10. Tableau – https://www.tableau.com/
11. Adobe Experience Platform Real-Time Customer Data Platform – https://www.adobe.com/experience-platform/real-time-customer-data-platform.html
12. Customer Experience Era – https://www.youtube.com/watch?v=4dftnjUeA0s
13. Resolving Customer Identity – 

    https://experienceleaguecommunities.adobe.com/t5/adobe-experience-platform-blogs/adobe-experience-pl...

14. Adobe Journey Orchestration – https://www.adobe.com/experience-platform/journey-orchestration.html
15. Marketing Automation: https://www.adobe.com/marketing-cloud.html
16. Adobe Analytics: https://www.adobe.com/analytics/adobe-analytics.html
17. Random Forest​ Method – https://docs.adobe.com/content/help/en/target/using/activities/automated-personalization/automated-p...

Originally published: Feb 18, 2020