Data questions
What are Mettle's sources?
Mettle’s data comes from three main sources: 1. What companies say themselves: annual and sustainability reports, user conversations 2. What the market hears about a company: Traditional media (e.g. BBC), trade press (e.g. Retail Week), and social media (e.g. Reddit, X) 3. What the market does as a result: Share price movements and market reactions. We curate that data daily in 15 languages (including global English), with records dating back to January 2010.
What is Open Source Intelligence?
Open-Source Intelligence, or OSINT, is intelligence produced by collecting, evaluating and analysing publicly-available information to answer specific questions. There are six sources of OSINT: traditional media (newspapers, magazines, etc); trade press (industry journals, academic publications, etc); social media (forums, blogs, Reddit, Twitter, etc); public government data (especially public hearings); companies’ own data (annual reports, quarterly earnings calls, etc); and grey literature (technical reports, working papers, newsletters).
What is 'wisdom of crowds' and why is it applicable to me?
The “wisdom of crowds” concept suggests that collective input often outperforms individual expertise. This idea, dating back to Aristotle, concludes that a diverse group of people, when combined, can achieve better outcomes than even the most skilled individuals. For example, the average of many guesses about the number of jellybeans in a jar closely matches the correct number. At Mettle, we leverage this principle to answer complex business questions.
What is sentiment?
Sentiment reflects how individuals or groups feel about an organisation, brand, or entity, influencing their likely future behaviour. Positive sentiment suggests support (e.g. repeat business, regulatory approval), while negative sentiment indicates challenges (e.g. customer loss, reputational risk). Most sentiment tends to be neutral - there is apathy, with no clear indication of future behaviour . Using machine learning (ML), we analyse sentiment to classify texts as positive, neutral, or negative. This process also accounts for emotional nuances and involves human oversight to improve accuracy continuously.
How does Mettle calculating sentiment?
Mettle calculates sentiment as "volume-weighted net polarised sentiment" by analysing publicly available conversations across traditional, trade, and social media. We compare positive and negative commentary, weigh it against conversation volume over a 720-day trailing average, and express results as a percentage (-100% to +100%). This method has shown strong correlations with subsequent share price movements.
Is Mettle's analysis accurate?
Our sentiment analysis, underpinned by machine learning models, achieves accuracy levels exceeding 85%, compared to human benchmarks of around 75%. These models have been trained on our own extensive, and constantly developing, datasets since 2015. Accuracy is continuously refined, especially when incorporating new languages—Mettle currently processes conversations in 15 languages using native-language training data.
How does Mettle ensure the accuracy and reliability of its data?
At Mettle, we prioritise data accuracy and reliability through a combination of advanced machine learning techniques, continuous model refinement, and rigorous validation processes. Our approach begins with sourcing a wide range of publicly available data, which we then analyse using sophisticated machine learning algorithms to identify patterns and insights. To ensure the accuracy of these insights, we collaborate closely with leading academic institutions, including UCL, Oxford Brookes, and Bath, where our models and findings are regularly reviewed and validated by experts in the field. This academic collaboration helps us maintain a high standard of data integrity and ensures that our approaches are aligned with the latest research. Additionally, our models are constantly updated and trained on fresh data, enabling them to adapt to changing trends and ensure that the insights we provide are relevant and timely. We also incorporate a human-in-the-loop approach to review and correct any discrepancies, ensuring that our analysis is both precise and contextually sound. Through these robust processes, Mettle delivers reliable, actionable insights that clients can trust to inform strategic decisions and manage risk effectively.
Can I get some sample data?
Of course! Use the “Contact Us” form to request data on specific companies, industries, or sectors. We’re happy to guide you through the data and answer any questions.