Real-world example
To help visualize how the data model works in practice, let's walk through a real-world example. We'll take a prompt, look at the AI's response, and see exactly what our platform "sees" and the insights we can generate from just this single piece of evidence.
The Prompt: "Best running shoes 2025"
What our platform “sees”
Our system reads through the response and identifies every important Entity (a brand, product, publication, etc.) and creates a structured Mention record for each one. A Mention captures not just what was said, but how it was said.
Here’s a sample of the Mention records generated from this one Observation, including their calculated Visibility Score:
Entity Name | Entity Type | Mention Type | Visibility Score | Why? |
Adidas Adizero EVO SL |
|
| 98 | A top-ranked (#1), positive recommendation at the very beginning of the response. |
Adidas |
|
| 95 | Directly associated with the top-ranked product. |
Runner's World |
|
| 65 | A highly positive citation used as proof for the top recommendation. |
ASICS Novablast 5 |
|
| 85 | A #2 ranked, positive recommendation. High score but lower than #1 due to position. |
RunRepeat |
|
| 55 | Cited multiple times as a validating source for several top products. |
Brooks Ghost 14 |
|
| 15 | Mentioned passively and compared unfavorably ("16 is the current iteration") to another product. |
Saucony |
|
| 70 | Associated with the #5 ranked product, appearing later in the response. |
Immediate insights from a single observation
Even before looking at historical trends, a customer like ASICS could log in and immediately see valuable insights generated from this single data point:
"We are the #2 recommended daily trainer. The AI positioned our Novablast 5 directly after the top racing shoe, which is a huge win."
"RunRepeat is a key validator for us. They were cited as the primary source crowning our shoe 'Best Overall Daily Trainer'."
"Our main competition in this response is Adidas. They secured the top spot and the highest
Visibility Score.""The AI is citing specific publications as evidence. We need to make sure our relationships with RunRepeat, Runner's World, and Believe in the Run are strong."
Insights explosion
Now, imagine this process happening hundreds or thousands of times a day. This is where the real magic happens. Each new observation is another piece of evidence that updates our larger "map."
Relationships get stronger: This observation would increase the
strengthof theASICSCOMPETES_WITHAdidasrelationship. It would also increase thestrengthof theRunRepeatVALIDATESASICS Novablast 5relationship.History is recorded: A snapshot of these new
strengthscores is sent to our Tinybird "time machine."Deeper questions get answered: After a month of collecting this data, the ASICS marketing team can now ask much deeper, strategic questions that look at trends over time:
Is our position improving? - "Show me a chart of our
COMPETES_WITHrelationshipstrengthagainst Hoka and Adidas over the last 90 days.""Who are the real kingmakers?" - "Which
publicationhas the highest Eigenvector Centrality? Who is the most influential voice in the running shoe space that we need to connect with?""Is our PR campaign working?" - "After we seeded our shoes with key publications, did our
VALIDATESrelationships with them show positivemomentum?"
Hopefully this shows the power behind this system. We start with a single piece of evidence, structure it, and then layer it over time to build a dynamic map of our customer's entire brand universe.