When choosing what you would wear today for work, you didn’t just pluck a completed outfit from your closet in one go. You searched one dresser drawer for your favorite socks and another for a pair of pants. You scanned your shoe rack—or maybe just all your shoes scattered messily across the floor—for just the right pair. Finally, you picked a matching shirt, accessorized from a tie rack or jewelry box, and brought your outfit together.

Picking an outfit is more than the physical act of grabbing individual items. You’re mentally scanning through different “boxes,” like multiple dresser drawers, closets, racks, and locations, to identify the attributes of each piece. You connect pieces based on those attributes and craft a coordinated outfit for your end goal: looking halfway decent at work.

When you process this information every morning, even when your brain isn’t fully awake yet, you’re doing a type of graph thinking.

What is graph thinking?

Let’s commute from your morning routine to your office and put graph thinking in a business context.

Graph thinking is a mindset that recognizes the connectedness of your organization’s data and prioritizes discovering those relationships in context to solve specific business problems.

When your organization gets graph thinking right and data/analyst teams are supported by the right graph and artificial intelligence (AI) tools to automate and abstract complexity, you get people with superpowers.

The concepts behind graph thinking, like data connectedness and the intersection of “human vs. computational thinking,” have been around for decades.

Seymour Paper coined the term computational thinking back in 1980, later popularized by Jeannette Wing in a 2006 article, which is a problem-solving technique for breaking down complex problems into smaller ones until computing can solve them. More recently, Denise Gosnell and Matthias Broecheler published The Practitioner’s Guide to Graph Data, which codifies their version of graph thinking and details the history of database technologies, from hierarchical to relational to NoSQL and beyond, over the last few decades.

The latest stage of database technology progression is the graph database, which helps organizations store and organize data in a way that’s aligned with their graph thinking. Imagine you used a whiteboard to sketch out your data and create relationships based on what matters to the problems you’re trying to solve. At the end of the day, you want to retain this information for later use—would you open up your laptop and transcribe the content of your whiteboard into a spreadsheet? Not only would that be difficult, it would strip away all the value of your graph thinking.

Unlike relational databases, which use rigid schemas, tables, rows, and columns to store data (think spreadsheets), graph databases store information based on how it’s connected to others. Nodes operate as discrete data units and are connected to other nodes through edges. Edges are relationships, defined by properties based on the type or category of said relationship.

Let’s go back to the outfit-picking analogy. Inside your brain, your clothes are nodes, all of which have different properties, such as color/pattern, style, texture, formality, etc.

When you pick a piece of clothing as part of your outfit, you leverage those properties to build connections with other items. For example, you connect a button-up white shirt to a pair of dark brown leather loafers/pumps because they share a certain degree of formality. It no longer matters that they’re in different “boxes,” because you can see how the context of their relationship solves your core problem.

When implemented correctly, graph thinking takes your team from plopping data into an increasing number of siloed boxes to intelligently organizing your data into an interconnected network. When you feed connected graph data into an engine, a process or algorithm reasons over your graph data and the full context of its relationships, you get insights or instructions that solve a business challenge, not just create another dashboard.

How is graph thinking different from other methods of solving problems with your data?

We’ve almost exclusively stored and structured data in relational databases or spreadsheets for the last few decades. Most of us think of data as tables, rows, and columns—boxes within boxes.

When we’re confronted with a business problem we don’t have an answer for, we use complex queries or spreadsheet formulas to retrieve or process the data, then convert those into visualizations or dashboards for analysis. Spreadsheets might be the perfect solution for certain use cases, but quickly reach their limits when solving more complex problems, like mapping your customers’ journeys and creating predictive recommendations optimized for their needs.

On top of that, relational databases and spreadsheets are fundamentally siloed from one another, like the gap between your dresser drawers and your closet and your shoe rack by the front door. You can merge spreadsheets or integrate their data elsewhere, but the scale of data quickly becomes unmanageable.

Making valuable insights relies entirely on luck, like finding a problem-solving connection between cell Z24 in Spreadsheet A to C172 in Spreadsheet Z.

Graph thinking, on the other hand, strips you of thinking in boxes and helps you organize your data based on its potential value. It’s like organizing all your work outfits—shirts, pants, shoes, and everything else—into a single dresser drawer. Instead of opening more boxes to find what you need, you get straight to work solving complex problems.

Why is graph thinking becoming more popular?

Graph thinking is currently experiencing a surge in popularity due to a number of interconnected factors:

• Major technological advances in graph technology, particularly within the open source community, like ONgDB.
• Advancement in artificial intelligence (AI) and machine learning (ML) tooling to help ingest and/or process data faster and with more sophistication.
• More challenging business problems, which require more comprehensive solutions.
• The sheer scale of data that most organizations store.
• Costly data and knowledge worker burnout due to the lack of the right tools to help them work more efficiently and productively.
• The popularity of an “data-driven enterprise” mentality.

These factors have pushed decades of using spreadsheets or relational databases to store and understand enormous datasets to their absolute limits. When knowledge workers can’t process data requests fast enough, or develop new methods of solving complex business problems, organizations stagnate.

The varied uses of graph thinking

In a business setting, graph thinking proves valuable whenever you face a question or problem that seems to be most intuitively solved with a whiteboard and drawing connections between what you already know.

These can be simple or complex questions. Perhaps you want to better understand which attributes your recent customers share or need to design a marketing campaign based around recent trends and what’s been successful in the past. They can also be enormously complex analytical problems, which, when solved, deliver enormous value to your customers or streamline complex processes to give your people the superpowers we mentioned earlier.

• Search engines for all of human knowledge: Both Google and Microsoft use graph thinking, graph databases, and other AI resources to understand your query and deliver the most optimized results.

• Fraud detection at governmental scale: The United States Department of the Treasury augments 5,000+ financial agents with knowledge graphs to help handle the 68+ million cases they investigate on an annual basis.

• Open source intelligence (OSINT): By tracking the relationships and communications between a global network of individuals and entities, using data that’s freely available on the web, organizations can define and crack down on intricate money laundering networks.

• Risk analysis of financial sanctions: A multinational enterprise connects terabytes of loosely-related data, such as social media posts and ads, with the latest sanctions list from the U.S. government to get answers in hours, not weeks.

• A global graph of professionals: LinkedIn analyzes the relationships between your contacts to build a valuable professional network based on whether people are 1st-, 2nd-, or 3rd-degree connections.

• Personalized digital referrals for patients: A telemedicine startup leverages graph thinking and GraphGrid’s Connected Data Platform (CDP) to build an API-driven platform that helps doctors identify other networks and trusted partners where their patients can get vital care.

A practical example of graph thinking: optimizing a physician’s day

To ultimately get paid for their work, physicians need to “code” the procedures they completed. Medical coding transforms all the diagnoses, procedures, and services they provide into standardized codes, which are converted into a claim for insurance providers to analyze and pay.

Many physicians do their coding work at the end of the week. They need to go back through multiple 14-hour days in the operating room, into the records and documentation for each patient they saw to remember which procedures they completed and tests they ordered, and map those to the appropriate diagnostic and procedural codes. That process is time-inefficient enough, but they’re also trying to remember a week’s worth of patients, increasing the risk they code a procedure or recommendation inaccurately.

Each mistake costs the physician money. They don’t get paid properly and must spend more time rectifying the error.

One healthcare startup recognized that physicians carry an enormous amount of knowledge in their head. The language they use to describe their medicine, the procedural and diagnostic codes they use, all the tests and treatments for each condition, and so much more.

They wondered: If we developed connections between all this knowledge, could a machine understand this language and help physicians code faster and more accurately?

Being physicians, they were driven by a practical goal of lowering this tedious and time-consuming burden, unaware that they were applying the fundamentals of graph thinking. They recognized the value in the relationships between the verbiage and standards of their domain, and understood they couldn’t do that work in boxes.

While the startup operated as experts in medicine, GraphGrid stepped in to help move all the disconnected procedures, codes, and terminology into a graph database, which they modeled around the goal of streamlining the coding process. They collaborated on connecting codes with the natural language used by physicians in their records and documentation, and stood up the composable platform they’d need to store and process data.

The teams then collaborated on training natural language processing (NLP) models around the connectedness of their graph data to create an engine, which took an unstructured string of text, like “abrasion on lower right leg,” and mapped that to a sequence of appropriate codes. They took the system one step further by naming dictation—physicians could dictate the work they’d completed, using their phone and their natural way of speaking, on the spot rather than days after the fact.

The NLP engine delivered 90% accuracy for the automated coding and gave physicians easy ways to modify or update the results, dramatically increasing coding accuracy and reducing the burden on physicians’ time. Instead of spending their weekend making often-inaccurate codes, they could dictate information in real time and get back to the real work: helping patients.

How to get started with building knowledge

Most organizations don’t just jump into graph thinking headfirst and get it right the first time. It’s more of a journey, not a culture or a process you can simply switch on.

First, ask yourself why you’re storing data at all. What problem are you trying to solve? Once you know that, you can start to spin up graph thinking by recognizing that connected information in context is far more valuable to solving your problems than putting everything into more boxes.

From there, you need to start breaking down the major hurdles, objections, and big questions that precede any long-term graph thinking journey.

• Break the cultural reliance on boxes. You’ve probably used relational databases and spreadsheets throughout your entire career. You’re used to writing SQL queries, sorting tables, and creating charts to understand or solve problems. To break these habits, workshop and whiteboard with your team to understand how you could map and categorize your information in different ways based on what’s important to your mission.
• Onboard the right technology. Graph databases aren’t nearly as ubiquitous as their relational counterparts, so you’ll need to deploy new services and platforms to put graph thinking into practice. This will include a graph database and other tools that help you query, search, automate, and visualize connected data.
• Ask big questions about your data architecture and analytical practices. As you progress into a graph thinking journey, you wonder if there are better ways to store information to reduce the tax and tedium of leveraging it to solve problems. Could you organize data differently to identify new patterns? How can you integrate technology and culture to keep your people from opening more boxes?

Find a graph thinking collaborator

Much like the healthcare startup that identified a problem in medical coding and wondered if machines could help physicians do the work faster and more accurately, your organization might not be ready to take on the entire graph thinking journey alone.

GraphGrid, for example, isn’t just a technology provider. We’re graph thinking experts with decades of experience and partners like the U.S. Treasury, National Institutes of Health, Peloton, and more, who can look at your organization objectively, as a set of interconnected data elements, and kickstart the process of making connections between the data you already have. We model your graph data, spin up ML or NLP pipelines to automatically process data, and optimize your organization’s knowledge for solving problems or generating original ideas.

When you partner with an expert and learn about graph thinking alongside them, you create a long-term solution to the ongoing crisis of knowledge worker burnout and turnover. Instead of asking your people to work harder, brute-forcing data into more boxes, you make them far more effective and innovative in the long haul.

Graph thinking just might be your next double-sided win, improving your internal productivity and solving problems far more complicated than picking out an outfit, with human thinking at scale.