Methodology

Fan data that tells you
what to do next,
not just what happened.

Understand the model behind FanTrendTracker, how to interpret the results, and how we address data security.

The emotion model

11 emotions, multi-label, never collapsed.

We measure 11 distinct emotions — an extension of Plutchik's classic emotion wheel that adds love, optimism and pessimism to the original eight. Real fan reactions are rarely a single feeling: a comment can be joyful, surprised and a little anxious all at once, and our model picks up every layer.

How to interpret each emotion?

Analytical dimensions

One pass through the data, eight ways to read it.

Every comment is enriched once and then re-interrogated through every lens below. Same canonical labels — different cuts, depending on the question on the table.

How does that work?

How the models were built

Open base. Closed-corpus fine-tuning. Independent validation.

A multilingual baseline, sport-specific re-training, and an evaluation methodology designed to catch the failure modes a generalist model wouldn't — so the headline numbers we cite hold up under scrutiny.

How was the model pipeline developed?

Local & private by design

Your data never leaves your environment.

Every other vendor in this space ships your fan data to a third-party API for inference. We don't. The whole pipeline — from ingestion through theme labelling — runs on hardware you control.

How do we address data security?

Platform capabilities

Six ways we understand your fans

Every capability is designed to answer a specific question a sports organisation, broadcaster, or brand actually needs answered.

What are the ways to interpret the data?