Exciting news from your team at Gazelle. We're thrilled to announce the latest update to our G-Scores, now live and ready to supercharge your business attraction efforts.
How Do G-Scores Work?
Our G-scores are completed and assessed against key criteria sets, including firm, industry, and region. Our analysts examine hundreds of data points to complete our analysis using artificial intelligence continuously learning and evolving with the new iteration.
For a deep dive into the inner workings of how G-Scores work, check out our short interview with Gazelle.ai's Chief Economist, Dr. Hugh Kelley here.
What Has Changed?
The 2024 run scored 15.5 million firm - industry combinations and based on 1,478 predictors (totaling 68.7 Billion data points).
We have observed that the overall share of firms predicted to expand across all industries has moved marginally higher. Firms classified as expanding and ‘to-be-called’ in 2024 and early 2025 lie between 22% and 51% (previously 17% and 54%) with an industry average of 37% (36%). Overall, this suggests that firms have slightly higher to the same propensity to expand in 2024-25 as they did in 2022-23, but the industry mix of expanders and trends related to this, have somewhat changed.
Industry variables have seen a small increase in their importance, and this gain comes at the expense of Regional variables. And again, Firm level variables are the second most important group. Some regional workforce characteristics do remain important with 2 (previously 5) in the top 20. Overall, these results suggest that to identify expanding firms across multiple sectors, Industry factors, particularly financial indicators, as well as regional and firm-level funding rates, are the most important predictors of expansion for the current economic climate.
How Has The Algorithm Improved?
There are a few quick hits we'd like to highlight, but if you want to dive into the technical side, find our the full technical whitepaper of our Chief AI Officer, Dr. Hugh Kelley here:
New AI model fits with an increase in the number of predictors to 1478 signals across 16 years (from previously 1,298 for 14 years).
Updating to the most recent data for all high (weekly or monthly) frequency, i.e. contemporaneous, data and updates to regional-industry financials data.
Use of NAICS Assigner/CScore for training to verify the accuracy of NAICS codes
Disaggregation of Professional Scientific, Technical and Engineering workers industry group into 2 separate groups focusing on Scientific and Legal activities versus Marketing and Advertising, as they were determined to have unique growth drivers.
Matching/Merging and AI code modifications to allow unlimited number of records and to mitigate any current or future hardware or security software limitations.