Welcome to the revamped Scale diversity report. Scale launched expanded diversity/demographic questions for the 21x conference we've spent two years revamping how we analyse and report that data. We now have the start of trends to report on with two conferences under our belt with the new information.
The expanded report required code to analyze which of course we have open sourced.
Further, this year we have some interactive graphs rather than flat images.
As always, we got the data differently for the CFP/Speakers vs Attendees. So let's jump right in with the CFP data.
tl;dr
The full analysis of our date follows, but first, here are a few tidbits we found interesting for folks that don’t want to dive into the weeds.
- Our CFP has nearly identical percentages going in (CFP submissions) and out (accepted talks) for our new demographic axes: age, ethnicity, education
No Response/Prefer Not To Sayis on the rise both for CFP submissions and for Attendees- Despite much discussion in the industry that these sorts of conferences are only appealing to an increasingly older demographic, we saw a diversity of ages in our attendees, with our highest percentage being 25-34
With only two years on all of our new data categories, it's not yet possible draw too may conclusions about trends, but hopefully over time this new data will help us to understand trends both in the industry and at SCALE.
NOTE: Prefer Not To Say is abbreviated in legends as PNTS.
CFP Analysis
As a reminder, the goal of this analysis is to look at the input and output of the CFP system sliced and diced by different demographics. This helps inform both any bias in the CFP system, but also where we can do more recruiting for the CFP system. Note that while the CFP system is not anonymous, reviewers are not shown the demographic information.
Further, we cut the data two ways. The first is the acceptance rate for each case (for example, for each gender). This means for people who identify as X, they had Y% of their proposals accepted.
The second is the percentage of all submissions and acceptances for each case. So of all submissions, what percentage of them identified as X, and same for acceptances.
CFP: Gender
We added two new options starting in 21x: non-binary and prefer not to sayNon-binary is obvious, but prefer not to say was important to differentiate people who chose not to respond at all vs. those who wanted to communicate that this was a question they didn't want the answer included in demographic stats. This had an impact: no response definitely went down (for most of these analyses it's useful to ignore 19x which was our first year back post-pandemic and the stats were pretty off).
For acceptance rate, the new non-binary option topped this chart both years: 77% in 21x and 63.6% in 22x having the top acceptance rate of any self-identified gender. From 21x to 22x, 12% of people chose not to respond to this question, and the percentage of people who chose prefer not to say rose from 45% to 60%. It's not unreasonable to assume the current political environment affects whether people feel comfortable sharing this sort of data. Here's the graph:
The submission and acceptance rates for gender looks very consistent with previous years. Submissions from 20x to 21x saw a 0.9% drop in other and a 5% drop in no response that were reflected in 2.8% prefer not to say and 1.3% non-binary, and acceptances have similar numbers. More interestingly, from 21x to 22x, looking at submissions we see a 4.6% drop in submissions from male, a 2.2% increase from female and a 0.2% increase from non-binary. Acceptance rate mirrors this with a 4.5% drop from male, a 2.3% increase from female, though a slight drop of 0.5% from non-binary, and a 2.6% growth in prefer not to say. All in all, the numbers didn't change really outside of their normal ebb and flow here, but we have some more granular data. Here are the relevant graphs:
CFP: Age
Age is a new demographic category for us in this report. We started collecting it in 21x and thus have two years of data, which isn't enough to see trends, but is enough to see where we are.
I thought it was interesting that folks under 18 had the top score for acceptance rate. But this is likely because there are fewer under-18s submitting to conferences, and of course SCALE has a dedicated track for such speakers.
The prefer not to say folks take the next slot, but then most age brackets are clustered together in the 40-50% range. Interestingly, no response was the lowest at 12.5% exactly, both years.
As a new stat, what I was hoping for here is that no age bracket proved to be a negative outlier, and that does seem to be true.
On the submission and acceptance front, the percentages that came in are nearly identical to each other - you actually have to look closely to tell the difference between the two charts. And that is really beautiful.
CFP: Ethnicity
On the ethnicity front, there also does not appear to be an outlier. For 21x Black / African American was lower than the rest, but it was the 3rd-highest acceptance rate in 22x, so it doesn't seem like a trend, and is likely within the normal band. Various different ethnicities criss-crossed each other, which is healthy.
Similar to the age graphs, the submissions percent and acceptance percent are nearly mirrors of each other.
CFP: Education
Education looks pretty similar to Ethnicity. Other and no response end up hanging out consistently at the bottom, and it's unclear to me why, but they are also the smallest number, making the percentages a bit less forgiving.
I won't bore you with too many words on this since it's similar to the above, and I'll just show you the charts. Acceptance rate:
And submission and acceptance breakdowns:
CFP: Employment Status, Marital Status, and Household Income
The last three demographic questions had a 0 or near-0 answer rate, and so we have no data for them.
Attendees Analysis
Attendee data is a bit simpler - people show up or don't, so there are fewer ways to cut the data. However, since we do have quite a high number of no response, for each graph I also showed a stacked graph removing no response so that it's clear what our breakdown is among those who did respond. I only had the raw data to do this for the last two years, however, so these graphs will become more interesting with time.
Attendee demographics shows us less about bias in the CFP system (although that would certainly be reflected in our attendees), and more about what people feel comfortable and represented at SCALE. Let's dig in!
Attendees: Gender
As before, we've had gender data for a long time, but only recently added prefer not to say and non-binary. Let's start with the graphs:
As with our CFP system, we saw a rise in people not responding, culminating in 53% this year. This just makes all the numbers go down and makes the graph misleading. If we remove those people altogether, we see that there's a slight drop in those identifying as male, and this seems mostly picked up by those identifying as female (14.4% - 15.9%) though we do see the new non-binary option at 1.4%. All in all, nothing too surprising here.
Attendees: Age
As with CFP, this is a new stat so only two years of data. Let's start with the graphs:
Once again no response dwarfs the data, so looking at the second graph we see that 25-24 is our biggest group at 25.4%, so we appeal to more than just the founder's generation of old Linux carmudgeons. :) Further, age groups are fairly evenly distributed except for those under 24, with under 18's at 1.3% and 18-24's at 10%. I don't know of any other conference publishing this kind of data, but given that the majority of conference-goers are sent by their companies, these numbers feel pretty high. It's also worth noting that under 18 is primarily recruited by schools we partner with, one of our many efforts to reach out to groups we might otherwise reach.
Attendees: Ethnicity
As usual, no response makes the data hard to see, so looking at the second graph shows a more interesting picture. White/caucasian is the biggest, but less than half, followed by Asian and Latino, then prefer not to say. While that's probably representative of the industry as a whole, I'd love to work on getting more people from different origins and backgrounds at SCALE.
Attendees: Education
Education also looks pretty similar to the industry with bachelor degrees being the biggest share followed by master and then high school/equivalent and everything else being small contingent.
Attendees: Employment
Full-time employment is so large here it even stands out on the first graph, and totally dominates the second graph. This isn't surprising since most conference attendees are sent by their company. However, we strive to be affordable for students, under- and unemployed folks as well, and I was pretty happy to see 32.8% of folks coming from some other answer.
Attendees: Marital Status
No real surprises here with single and married making up the vast majority of respondents, but it was interesting they had a nearly even split between them.
Attendees: Household income
Our final category here is household income. As mentioned before, we strive hard to keep our ticket prices low as well as give out lots of discount codes to schools and other organizations in order to ensure our event is welcoming to students and those who are under- or unemployed. It's no surprise that the $100k-500k ranges are the largest, we also see solid representation from the 10k-100k ranges and even ~1% from below 10k. If you want to attend SCALE and can't afford to, please do reach out to us!
A word on the code
Prior to this year, the numbers were generated from a variety of kludged together spreadsheets, scripts, and manual math. This year, all numbers were generated with code you can find in the scale-dei repo.
Google Sheets is still the number-to-graph generator, and I think that works well with these new interactive graphs this year. The scripts generate a single line I can copy and paste into Sheets and that's it, so now that I have all the tabs and graphs wired up properly, generating these reports should be significantly easier and faster in the future.
Parting Thoughts
That said, looking at the graphs, while there is room for improvement, I am pretty happy that we still seem to be a conference that has a reputation for welcoming people of all backgrounds, identities, looks, etc. We've had several interesting conversations with newcomers and existing staff on how we can better reach out to underrepresented communities and help them feel more comfortable attending and hopefully also speaking at SCALE. I hope this aggregated data will encourage more folks to answer the demographics questions going forward so we can get a more complete picture. And just to re-iterate: any demographics data is only ever used in aggregate.