<p><strong>Level:</strong> Beginner · <strong>Topic:</strong> Data</p>
<p>A <strong>data graph</strong> is a map of things and the links between them: this invoice belongs to that job, that job belongs to this customer, this number came from that file. It is not a pretty picture for a slide. It is how you answer “where did this come from?” without guessing.</p>
<h2>Tables vs a graph (in one minute)</h2>
<p>A spreadsheet is a list. Rows sit next to each other. That is enough when one table is the whole story.</p>
<p>A graph is a set of <em>nodes</em> (the things) and <em>edges</em> (the relationships). Customer —has— Job. Job —produced— Invoice. Invoice —used— Rate card. Once those links exist, you can walk them. You do not have to remember which tab hid the rate.</p>
<p>You already think in graphs when you say “who owns this?” You just have not drawn it.</p>
<h2>Why this shows up when you use AI</h2>
<p>Models are good at sounding sure. They are not a source of record. If you ask “what did we charge last time?” and the answer is not tied to a node you can point at, you have a paragraph, not evidence.</p>
<p>A data graph does not make the model smarter. It makes the facts findable. Then you can tell the model: use only these nodes. If the link is missing, say missing. That is the same idea as a prompt contract: inputs you can name.</p>
<h2>A small work example</h2>
<p>You have a quote, a site visit note, and a labour rate. In a folder they are three files. In a graph they are three nodes with two links: visit —informed— quote, rate card —priced— quote. When someone asks “why is this line $180?”, you follow the edge. You do not re-litigate the week.</p>
<p>You do not need a huge platform to start. A page with boxes and arrows is a graph. A tidy set of IDs in a table that point at each other is a graph. The test is the question, not the software.</p>
<h2>What a data graph is not</h2>
<p>Tags and maps are how you declare your own work. They are not a certificate from anyone else. It is not the same thing as a course exam, and it is not a reason to dump a customer database into a chat window.</p>
<p>Keep the map small. Name the things you actually reuse. Leave the rest in the folder until a question appears twice.</p>
<h2>What to do this week</h2>
<p>Pick one decision you explain often. Draw five nodes and the links. If you cannot name a link, that is the gap — not a tool to buy. Easy, on paper, on your phone photo if you want a copy.</p>
<p>Instilligent Courses is for people who already have a job and need practice that fits. The Durable Stack is the live course on what work stays yours when a model does the first draft. We do not sell an enterprise catalogue. We do teach you to name sources before you trust a sentence.</p>
<p><a href="https://courses.instilligent.com/course-detail/the-durable-stack-whats-left-when-ai-does-the-work/43735/">Practise this week — The Durable Stack</a> · <a href="https://courses.instilligent.com/blog/what-is-a-prompt-contract/">What is a prompt contract?</a></p>


A data graph is a map of things and the links between them, so you can answer where a number came from.

What is a data graph?

Independent practice exams and structured courses from Instilligent. CCA-F prep and the Durable Stack. Practise this week. Not affiliated with Anthropic.

What is a data graph?

What is a data graph?

Related Posts