Personal Insights
Enterprise Data Strategy that enables AI Adoption
Whether or not you are involved in Artificial Intelligence or Machine Learning (AI/ML) empowered workflows in your business, you most likely have heard AI is transforming the way businesses run today.
The IDC FutureScape 2022 report predicts that AI penetration in business will be 65% by 2024. For example, European organizations are increasingly investing in AI and AI powered technologies and platforms to build on them AI models and algorithms.
You certainly don’t want to be left behind in this transformation. However, if you haven’t started adopting AI, or you started the journey but are struggling to realize the value of AI, what should be your next move?
Should you
- hire some data scientist and count on them to get the job done?
- shop around to get the coolest AI tool and call yourself an intelligent enterprise?
- copy what your top competitor is doing because they have already adopted AI?
None of the above is the right next step. What you need is a creating a framework that fits your unique business status, that is centered around your own business data, and that brings you a systematic solution so you can implement it. We call this framework the Enterprise Data Strategy.
What is Enterprise Data Strategy?
A successful Enterprise Data Strategy is shaped like a triangle, which includes
- Data Business Strategy
- Data Operational Strategy
- Data Technical Strategy
Figure 1. The data strategy triangle
1.1 Data Business Strategy
What is the business outcome are you looking for? What are your business priorities in the next 6-12 months? You may say “I want to increase revenue and reduce cost”. Sure, this sounds like a universally true business goal, but to get down to the “How” part of the strategy, you need to be more specific.
The way to think about it is defining a specific goal for your business and come up with a North Star metric to evaluate your success. (You could break down one key metric into few proxy metrics. Such a discussion is out of scope of this article.)
For example, you are in the manufacturing industry. You produce computers and computer accessories. Your main business challenge (and goal) for the next 6-12 month is finding and maintaining a proper level of inventory, so that your storage cost is low enough and you also won’t have a shortage of inventory that slows down your sales. For this goal, you want to use Days Sales of Inventory (DSI) as a key metric to evaluate how successful you are. “I’d like to reduce my DSI from 60 days to 45 days in 6 months”.
Defining your business goal and knowing how to measure it is the first step. You then need to think about do you have the right people, and if the people have the right skills to implement the strategy.
On the business side, you should have roles like business analysts, controllers, power consumers (decision makers) who analyze, review data insights, and even make decisions. On the technology side, you may already have application developers and admins, but you also need to have data steward, data engineers, and data scientists who know how to manage, process data, and run AI/ML jobs to generate insights. Different people touch different stages of your workflow, but it is their collaboration that makes your strategy successful.
1.2 Data Operational Strategy
What does your business process look like? If you touch purchasing, logistics, manufacturing, and sales, you certainly have multiple processes running. Each process may need some data as input and generate some other data as output.
Hence, it reflects your business’ operational model
- Which business process generates what kind of data?
- Are these data relevant to your business goal?
- Are there data compliance requirements that you need to meet?
- How is the quality of your data, and do you have relevant process to control them?
- Who owns vs who consumes the data at each state of our business process?
Your operational strategy serves your business goal in the way that it helps you define all kinds of requirements for the data along your business processes.
1.3 Data Technical Strategy
Now it is the technical enablement part.
Don’t rush into buying and setting up some systems and tools. Build your IT landscape with a proper data architecture in the solution first
Some typical questions to answer for this dimension includes (and is not limited to),
- What data need to be collected (including third party data if applicable)?
- How do you collect the data?
- Where do you store the data?
- What is the end-to-end data flow across your IT landscape?
- What kind of transformation is need for your data to be prepared for AI/ML jobs?
- Based on your business goal, do you have certain scalability and performance needs that you need to meet?
The business, operational and technical dimensions of data strategy help your organization build a holistic view of what processes, people, and systems you should put in place and how they work together. Without data strategy, you could easily end up having data silos in your IT landscape. It could be slow for you to get data insights and you could even get inaccurate insights. With a proper Enterprise Data Strategy, it is easy for you to get an out-of-box view of all your data, whether it is SAP or non-SAP data. You can easily and quickly understand insights based on a unified data models and established business context.
How SAP Business Technology Platform (BTP) helps you
SAP BTP provides a wide range of products and services that help you enable and implement your data strategy.
That said, you still need your data business strategy and data operational strategy in place first. When it comes to enabling and implementing the technical strategy, SAP BTP helps you sort out the lifecycle of your data flow and provide tools to help you get the jobs done along the journey.
Figure 2. The data lifecycle
Let’s look at them step by step.
1). Your data lifecycle starts form Data Integration. Basically, you need to figure out how do you establish a powerful data layer to connect distributed data sources and inherit business entities & semantics for later usage in data modeling.
2). With the data source connected and integrated, next you need to consider where and how do you persist data using powerful secure multi-model data platform to extend applications with agility and flexibility.
3). Next, you need to start managing your data. The key question is, how do you gain a centralized view of data across your landscape and enable data governance with easy-to-use search and data lineage?
4). Now we are getting to the data processing step, and data processing itself starts with how do you prepare a robust and agile data foundation to build reliable insights?
5). Prepared data can be used for AI/ML jobs. Meanwhile, to serve other use cases like BI analysis, visualization, and reporting, you need to create data models by reusing business semantics and content packages.
6). Now it is time for IT and business to “shake hands”. Hence, how do you empower enterprise-wide business & IT collaboration for enterprise-grade data sharing is the key
7). Finally, you can build up own intelligent analytics and planning applications supported by machine learning to accelerate time-to-insight by leveraging pre-built business content
Along this journey, BTP has SAP Data Intelligence Cloud to serve steps 1, 3, 4. SAP HANA Cloud can serve steps 2, 4, 5. SAP Data Warehouse Cloud can serve steps 5, 6. SAP Analytics Cloud can serve step 7.
Figure 3. SAP BTP Data and Analytics Products
[Side note] Our focus in this article is not introducing SAP BTP, but if you want to learn more about BTP, here are a couple of very good resources
Case Study
Having discussed Enterprise Data Strategy and how SAP BTP can help, let’s look at a real customer case. For privacy consideration, we won’t share the customer’s name.
This company is in the manufacturing industry. Their products are produced in batches on the manufacturing lines. Previously, they had a high rate of disqualified products. Disqualified products can’t be sold, but manufacturing materials had been consumed, human effort had been spent, thus it resulted in high production cost and low profit.
Technically, they were unable to access sensor data and ERP data in real-time, which made it hard for them to take preventive actions to improve manufacturing quality.
With SAP’s help, they first decided that improved manufacturing quality is their business priority. Then they defined metrics to measure the quality for each product batch.
On organization setup and operating models, they clearly identified plant managers, business analysts, data scientists and IT experts as the key roles and enhanced their collaboration from process perspective.
On the technical solution side, with SAP BTP data and analytical tools, they were able to access sensor data and ERP data in real-time. They reused SAP semantics to build data models and gain end-to-end view of all data. They were able to leverage machine learning to advise product quality in real time.
Finally, the comprehensive approach enabled them to identify top factors that are driving product quality issues and take preventive actions to avoid quality decline. With saved cost and effort, they were able to increase their profit.
Figure 4. ML use case solution architecture
Summary
In this article we discussed what is Enterprise Data Strategy, how SAP BTP data and analytics tools can help you implement your strategy, and how having the strategy implemented can enable you adopt AI technology to gain data insights.
With AI based insights, you gain more values from your own business data. This, in turn, make your organization an intelligent enterprise that is more sustainable.
Enterprise Data Strategy is not an IT specific job to be done, rather, it is a comprehensive framework that helps you be more successful in your digital transformation. Know what business outcome you want, sort out your business requirements and processes, implement it with data and analytics technologies and tools. Finally, take it to the next level by adoption AI.