Featured field and capability
Model-Based Systems Engineering
Nowadays, business plans and annual reports contain an abundance of digital ambitions and doublespeak. Every company wants to become ‘data-driven,’ and most expect to benefit from technologies like artificial intelligence, robotics, and the Internet of Things (see also i BCG digital transformation survey 2020). But often, these terms are signs of undefined expectancies and giveaways of hope that computers someday make life easy. Unfortunately, it’s not so simple. Becoming data-driven requires thorough strategy and state-of-the-art technologies. This article explains what paradigm a company needs to adapt to overcome its limitations.
What is the current state of things?Businesses design their processes and all assets required for the execution of their processes. Assets include roles and responsibilities in the organization and the facilities sheltering staff and machinery. And since the era of computer technology, data has become an asset too.
The computer is expected to help make decisions and measure effectiveness when discussing being data-driven. We want our businesses to make the right decisions based on the right data at the right time and improve continuously.
Image 1: Data as part of the plan-do-check-act cycle of continuous improvement.
Most companies design their businesses by cutting up their processes and organization into small pieces. This works fine for business units that function entirely autonomously. But, most organizations have interwoven dependencies and require well-managed interfaces between business units to make their strategies a reality. And if the output of one is the input for another, so is their data (often input and output are nothing more than data). Surprisingly enough, businesses design processes, organizations, machinery, and facilities thoroughly, but most tend to forget to do the same for their data.
Companies still stick to purchasing software from various vendors and have our IT departments tie it together. No matter how effective the organization’s design and processes are, a company’s data landscape will still look like a patchwork quilt. This siloed approach leads to enormous expense to create and maintain all sorts of software interfaces and sets up a digital strategy to fail. This is one of the reasons why businesses struggle to become data-driven.
The paradigm needs to shift: “data first, technology second.”
So, how do businesses become data-driven? Let’s start with some essential preconditions.Evidently, a business’ processes need to be standardized, and an organization needs to be in place. Only uniformly executed processes lead to predictable inputs and outputs in terms of data. And only if data is uniform can software help. Don’t expect actively supportive software for ‘black-box’ processes. Those processes are usually supported by all sorts of digital spreadsheets that can be changed easily to the whim of their author. You will only become genuinely data-driven when your processes are aligned with their outcomes in terms of data and the availability and useability of your data for different software applications provided by (probably) various suppliers. To make this happen, you need to take ownership of the design of how your organization uses its data. Unless your process matches your data by design, you are not in the driving seat. Your software supplier is.
What does it take to shift the paradigm?It starts by designing your data for its given purpose (the processes it’s used in) and software to be able to handle this data (for processing in an automated way). For software to help in the first place, information needs to be accessible for software and interpretable by software. There is not one single solution and not a single technology to become data-driven. As long as the technology chosen allows for your company to have control over the following:
1. Design your dataThe design of the data needs to be predetermined. IT people commonly mistake designing (structured) data for creating data structures (e.g., for setting up a database), but that is not what is meant by ‘data by design.’ Information architects need to design the semantics (meaning) of a company’s data and the interrelations between pieces of data. For instance: how and where do we use an ‘author,’ and how do records of authors relate to records of ‘documents’? Such a design should be agnostic of the structure in which the data will be stored. We call this design an enterprise semantic model that acts as a blueprint for all software within the company. Either to derive how data should be stored, how data should be queried, or how data should be mapped with proprietary solutions.
Image 2. Laces use knowledge graphs to design an endless string of meaningful data within its context.
2. Make your data accessibleHaving data by design isn’t enough to be data-driven, however. Our information systems (man and machine) need to be able to access all common data required for their various purposes. Common data must be available for multiple queries by software applications used in various processes. Because every software supplier speaks its language, sharing data between applications is burdensome and very costly. To access the same data with different software, businesses need uniform agreements (protocols) to access various data stores (like databases and files). These agreements are made available by standardization bodies like W3C and ISO, for instance. Applications become providers and consumers of commonly available data instead of siloed and locked-in in more and more different software. (see also ii Harvard Business Review Analytic Services, 2021)
Image 3. Laces uses W3C’s web protocol for accessible and distributed data as a fabric.
3. Help your softwareAs we design data and make it accessible, we have to keep one thing in mind: data needs to provide the correct semantics for our computers to interpret the meaning for a given purpose. As companies have ambitions to automate more and be assisted better by our software, we have to keep the needs of our software in mind as well. Data limits AI and IoT ambitions most of the time (the current nature). (see also iii McKinsey Global Institute (MGI), 2016) Because the data we have isn’t designed that way. To interpret data, software needs uniform data (both in meaning and format). And data, in turn, can only be uniform when it’s processed consistently and uniformly—both by man and machine.
Image 4: Laces uses semantic networks for interpretable data
4. Choose your software freelyThe first three topics addressed the semantics of data first, the nature of data second, and now we have come to the last step: software applications. Notice that with the old ‘technology first paradigm,’ we couldn’t have any more influence on the semantics of data and its nature. That was up to the software vendor. But data without software isn’t anything. We need tools to process data and have software present us with the correct information at the right time to support our decisions and monitor performance. Now that data has been put first and technology second, your company is back in the driving seat. Companies can ask software suppliers to provide excellent functionalities, accessing and processing data the way we have designed instead of how software forces us to.
ConclusionStarting with software applications is not the way to become data-driven. Software should be supportive of people handling the data they have designed for their given purpose, derived from their processes, which in turn were derived from the company’s strategy. Only by shifting the paradigm from technology first and data second to data first and technology second will your business become data-driven and reap its fruits.
Interested in reading more?
i BCG digital transformation survey 2020
ii Harvard Business Review Analytic Services, 2021, PULSE SURVEY | EMBRACING DATA ANALYTICS FOR STRATEGIC VALUEiii McKinsey Global Institute (MGI), 2016, THE AGE OF ANALYTICS: COMPETING IN A DATA-DRIVEN WORLD