At SAPPHIRE NOW this year, we demonstrated a powerful new tool that I use as chief data officer (CDO) at SAP. It’s based on existing SAP offerings and customized for the specific needs of the CDO. Although it is not yet an available product, we wanted to show it to our colleagues – and tell you about it in this blog.
Internally, we call this tool the “enterprise data experience.” Essentially, it’s a dashboard that tracks KPIs of relevance to the task of ensuring trusted data globally for SAP. As a company, we have set ourselves the goal of “making data work at SAP.” The ultimate objective is to deliver a better experience for our customers and continue our own evolution toward becoming an intelligent enterprise.
The importance of governance
As CDOs know, proper data governance is a bedrock requirement for any company that wants to use data as a source of value, insight, and business advantage. In the past, teams here at SAP would update master data by email or other ad hoc means.
To make data work for SAP, we first needed to establish clear data governance processes. Now any changes are made through a ticketing system. This has been a critical move because now, our team has visibility into what’s changing and what’s causing issues.
From data operations to data innovation
In addition to this ticketing system, a few years back, SAP moved to a shared service model where common company-wide tasks were consolidated into regional centers. This move alone has helped us improve efficiency. But in a digital economy, companies can never stand still. Our key objective now is to move people working in these shared services centers from the job of data operations to data innovation.
What does this mean? Data operations can be a tedious task involving a lot of spreadsheet work and manual data entry. But the human brain is much more useful for coming up with new creative ways to use data for improving processes and delivering better experiences. The more tedious tasks should be automated.
To this end, we use our ticketing system – and all of our trusted data flowing through streamlined processes – to perform intelligent analysis. We do this using machine learning algorithms that can detect patterns. Let’s say that the algorithm identifies a consistent issue with a specific field in one of your regions whenever a salesperson attempts to finalize a deal. Time after time, the problem gets escalated to your shared services team for a manual fix. Once the machine learning identifies the problem, you can then isolate the root cause and thus alleviate the burden of manual fixes.
War on duplicate records
Duplications are the bane of every CDO’s existence. Duplicate records impede visibility and cause confusion. Take master customer records, for instance. If you want to develop strong customer relationships and drive loyalty, at the very least, you need to present a single face to the customer across channels – by phone, online, in-person, by mail, or wherever.
In industries such as retail and consumer products, this is the “omnichannel” ideal. Duplicates, however, make it all but impossible to achieve. Silos across interaction channels typically contain conflicting duplicate records of the same customer. This is why, when you as a customer are having trouble with online shopping and make a phone call to complete your transaction, you often need to provide all your info a second time. The company doesn’t know who you are from one channel to the next.
At SAP, my team is helping to lead a war on duplicates using machine learning. One example has to do with account creation. Whenever a salesperson creates a new customer record, we quickly run the request through an algorithm to detect duplicates. The complexity comes into play with big accounts that may have a global presence with multiple lines of business in different regions – all doing business with SAP.
Is the new record a net-new customer or a subaccount of a larger customer? The machine learning algorithm is designed to detect patterns in our customer record hierarchies and make decisions about how to categorize new records. In instances of 90-100% certainty, the decision is made automatically to allow the new record or reject it. Anything less requires human intervention.
Ongoing analysis by the algorithm
Yes, this may mean a lot of human intervention, but the machine learning algorithm also analyzes the decisions made by the data expert working on the issue. Over time, the algorithm has been learning more and more about how to categorize new customers. The result is fewer duplication issues, leading to fewer tickets escalated to our shared services team.
This helps us improve the customer experience by minimizing confusion regarding who our customers are. The goal is always a 360-degree view of each one so that we can deliver the right products and service at the right time and place. A commitment to proper data governance and advanced technologies to continuously improve are helping to get us all the way there.
For more on how emerging technology can help shape the customer experience, see Digital Transformation, Customer Experience, And The CFO’s Role In Innovation. For more on the role of the CDO, see Oil, Data, And The Chief Data Officer In The Digital Economy.