Which G-Scores Have Changed?
In short, more than ever before! We have undertaken our largest-ever data update, scoring absolutely every company for which we had enough data to do so. It's never been a better time to find companies that are ready to expand.
How Can I Tell Which G-Scores Have Improved?
At present, there isn't a way for us to display which companies have increased or decreased in the score. It's worth explaining why: since our G-Score is intended to be a snapshot of a company's propensity in a given moment, it's not necessarily appropriate to compare the predictions from two different runs (similar to the questionability of comparing the grades of students who participated in a class at different times or in different cohorts. A company's rising or falling G-Score can be due to a very wide range of factors (such as firm performance) but may also be related to changes in data used, mathematical structures of the algorithms, or the presence of unique market shocks; note the latter causes do not affect the relative ranking of firms as they are common to all firms at a given point in time. In the future, we will overcome this limitation by generating time trends and generalizing G-Scores to different points in time using a common dataset, algorithm structure, and parameter set.
How Has The Algorithm Improved?
While we don't want to elaborate upon this to the extent our Chief AI Officer Dr. Hugh Kelley's would like, and you can read his public white paper here, there are a few quick hits we'd like to highlight:
The overall percentage of companies that the algorithms predict will expand (i.e. G-Scores 2, 3, 4, and 5) has declined slightly. This is due to a wide range of factors, but it's rather apparent that the COVID pandemic plays a role.
Firm-level variables seem to be more significant for this latest round of tests than previous ones, meaning that a company's individual performance (as opposed to its geographic location, industry, or other factors) explains more about a company's willingness to expand than previous sets of scores have shown.
Deep learning has made our predictions in several industries, such as NAICS 32 (Wood, Chemical, and Plastics Manufacturing) and NAICS 33 (Fabricated Metals) more accurate.
G-Scores of 1 (No expansion predicted) have become significantly more accurate for Computer Manufacturing (NAICS 334xxx) and Primary Metals Manufacture (333xxx to 3339xx).
For more insights and technical details, we highly recommend reading Dr. Kelley's white paper. Further explanations are also available by request.
Why Are There Fewer Verified Projects (G-Score 6) Than Before?
As part of our regular data updates, we removed all of the projects from 2016 and 2017, all of which had overrun their projected project timelines and taken place. To avoid confusion by making companies seem like they have active projects when they no longer do, we instead scored these companies anew and gave them the G-Scores you see now.
Why Don't Branches Have G-Scores?
As part of our latest update, we decided to change some of our procedures when it comes to assigning G-Scores to company entries in Gazelle. The biggest change was that we stopped assigning the G-Scores of parent companies to their individual branches, as we found that this artificially inflated results and led to much confusion - i.e. that this particular branch was prone to expand, rather than the company as a whole.
That's not to say that we're removing branches - far from it, we're constantly working on expanding the company trees for all of the companies in Gazelle.ai. You will still be able to search for these branches, including HAT searches.
Why Does X / Y / Z Company Not Have A G-Score?
Broadly speaking, there are three possible reasons a given company might not have a G-Score.
Newly-Added Company: Because of the very time-consuming process of constructing G-Scores and scoring them, we create our scores in large batches, generally the entire platform as it exists at a given time. As part of our more regular data updates, such as trade shows and featured lists, we will of course also continuously create new companies in the time between G-Score rounds. These companies will not have G-Scores until the next scoring round and are marked with a N/A to reflect this.
Branch Location: As mentioned above, we have decided to remove G-Scores from branches of companies for the sake of clarity and readability of the platform.
Invalid NAICS Code: Gazelle.ai is a tool designed specifically to identify expanding companies, especially companies with compound annual growth rates (CAGR) over 20%, which are most responsible for economic dynamism, growth, and hiring in their respective industries. Because of this, there are some NAICS codes, like agricultural activities, NAICS 812xxx to 819xxx, and NAICS 91 codes for government services, that we do not include in our data set, as they fall outside the scope of our data thesis. In the event that such a company is added to the platform through one of our data providers, its G-Score is listed as "N/A."
What If My Question Isn't On This List?
Just send us an email! We'll be happy to help in any way we can.
Happy Gazelle-ing!