Forbes, which seems to have cornered the market on list-making, recently dropped their latest roundup: America's Best Startup Employers 2024. Here's why we're intrigued: 1) We're kind of a startup sleuth ourselves, 2) We've already looked into many of the top 100, and 3) With a storm brewing outside in NYC, what's a better pastime than diving into a good ol' data mining and analysis?
Now, Forbes swears by a certain 'methodology,' as seen below. Sure, we'll bite. We sifted through the first 100 companies, collected relevant data points using our system, scrutinized each metric, mapped out visual aids, and yes, even ran a PCA-based multilinear regression (how very 2016 data scientist boot camp of us).
Source: https://www.forbes.com/lists/americas-best-startup-employers/?sh=58d20b312ad7
Despite our best efforts, this Forbes list remains as cryptic as ever. What mysterious alchemy determined this ranking? We cannot tell; there seems to be no coherent trend across many of the data dimensions.
We pored over specific datasets for each of the 100 companies to fuel our analysis. Here's the scoop on what we reviewed: Glassdoor Reviews, CEO Approval Ratings, Blind Reviews, LinkedIn Job Openings, 6-Month Total Headcount Growth, Total Employee Numbers, Indeed Reviews, and the Year of Founding. See the complete data here.
The Glassdoor CEO approval rates of the top 100 companies show no consistency, with the X-axis representing ranking.
The Glassdoor company ratings of the top 100 companies show no clear trend, with the X-axis representing ranking.
For the top 5 companies: they boast an average Glassdoor review of 4, a CEO approval rating hovering at 83%, an Indeed review of 3.8, and a headcount growth of a whopping 1% over the past 6 months. How these metrics stitch together into "best startup employers" is still up for debate.
Our ‘fancy’ machine learning, sorry, ahem PCA multilinear regression spat out an R-square value shy of zero. Our model, consisting of the data points we mentioned above, does not effectively predict or account for the variation in the 1-100 ranking we are analyzing.
Here's a fun fact to ponder: Glassdoor, often mired in controversy over score manipulation, has ratings that are a bit like fairy dust—sprinkle them where you may, but don't take the magic at face value. Compared to Blind and Indeed reviews, Glassdoor scores are inflated by 10% and 14%, respectively. On a scale of 1 to 5, that's significant. It just goes to show that a pinch of skepticism can add a lot of clarity to those numbers. Generally, the bigger the discrepancy we observe, the more likely it is that the Glassdoor rating has been edited.
But hey, we're all about sharing. Take a gander at our data and tell us if we missed a trick. Despite the mystery, there are indeed some gems among these companies worth tipping your hat to. Again, you can access via Airtable here.
But there are still many companies that have maintained consistent scores across platforms and withstood our scrutiny; many maintained consistent ratings or growth despite the startup downturn over the last 6 to 12 months. OpenAI have met the criteria of having above 4 ratings across employer review sites, a good CEO approval rate (above 80%), and positive headcount growth:
OpenAI
Glassdoor Review: 4.2
Glassdoor CEO Review Approval: 100%
Blind Review: 4.4
6 Months Total Head Count Growth: 75%
Indeed Review: 5.0