Singular Black Female
By Tom Adeyoola
After a tumultuous week dominated by the US election that ended with a sore loser incumbent failing to go quietly, we’ll focus on one positive aspect of the provisional result. The historic.
Kamala Harris has become the first female Vice-President (-elect) of the United States of America. The first Black and first South Asian in that role. For a nation built on immigrants, the image below is clear and stark.
On Saturday, Harris made a promise to the country.
“While I may be the first woman in this office, I will not be the last, because every little girl watching tonight sees that this is a country of possibilities.”
Extend Ventures released a report on ‘The State of the Nation for Diverse Entrepreneurs’ the day before the US election, which also for the first time highlighted a stark and clear historical fact. Only one Black female founder in 10 years has raised funding publicly classified as Series A or above. Just one ‘Singular Black Female’.
Given the report published by the British Business Bank just a week earlier concluding that ethnic and female founders face systemic disadvantage, particularly with reference to access to finance, it would be difficult to come to the same hopeful conclusion as Harris that the UK today is ‘a country of possibilities’ for the diverse and underrepresented.
But there is no reason that it shouldn’t be now that data exists to point the way towards removing the barriers to opportunity and solving for market failures.
The British Business Bank report stated that ethnic founders tended to be more qualified and spend more of their own time and money on developing their ideas, yet achieved worse outcomes. “Controlling for 30 explanatory factors….ethnicity and gender are intimately tied to worse outcomes,” it said.
The British Business Bank report is based on survey information, but at Extend Ventures we want to understand absolute numbers, to truly understand the state of the nation, to classify all entrepreneurs and business owners by gender, ethnicity, place and educational background.
Why is this important?
You cannot improve what you don’t measure and too often the anecdotal is easily dismissed with a counter example. And more than that, the anecdotal allows for the denial of the existence of unconscious biases. You will be hard pushed to find an investor who will say that they don’t purely invest in the best businesses and that an entrepreneur with a great business idea, no matter their background, could not successfully secure investment from them.
Indeed, it has been fairly impossible to get any venture capital firm to publish figures showing how many investments they have made into ethnically diverse founders. The Rose Review opened up the discussion on female founders. But it has always felt clear to me that each firm did not want to publish their data knowing that they had none or at best only a few and were worried that they would be worse than the firm next door.
As the Extend Ventures report found, 38 Black founders have raised venture finance in the last 10 years, 0.24% of the total raised in that time period. Ten (0.02%) of them were female. There you go. The cards are on the table. No need to hide the data. You can all do better and for all the talk this year post George Floyd and Black Lives Matter, we can now see whether any of the precipitant action increases those two numbers meaningfully. That’s the bottom line. Will another Black female founder have raised a Series A by October 2021?
Capturing diversity data
Extend Ventures is a Black-led group of volunteers and it has, like many Black founders that have gone before, struggled to raise funding for what is a non-profit working on crucial research to help diversify access to finance.
We have taken an innovative approach to capturing diversity data for businesses to enable us to classify them by the gender, ethnicity, place and educational background of their founding teams for the first time.
Why hasn’t this been done before, especially with reference to ethnicity? To be frank, I think out of fear of getting it wrong and falling foul of the Equality Act (2010) and, in particular, the self-determination characteristic of ethnicity. Your ethnicity is what you determine it to be, not someone else. As surveying founders for ethnicity is not mandatory, researchers that have gone before have deemed it too difficult to capture this information universally.
However, we believe perceived ethnicity in a funding situation is what matters, i.e. what the gatekeeper to financial capital deems the entrepreneur to be, as that is what will spark any potential unconscious or fully conscious bias that exists.
Returning to Kamala Harris; a case in point.
I can well believe that a non-ethnic organisation would be fearful of taking a ‘perceived ethnicity’ approach for fear of making crass assumptions or just getting it wrong.
For us, the benefits far outweigh what we believe to be manageable risks. We started by accumulating all publicly announced venture funding deals in the UK over the last 10 years. By starting with venture capital, which is a small subset of funding sources, we could prove our methodology before scaling up to larger datasets. Next up are all businesses that have received grants over the last 10 years, which is more than 40,000, compared to the 2,002 venture-funded businesses found in our research. We are thankful to Innovate UK for sponsoring this phase of our research, which has the added benefit of being able to not just classify the companies that received grants but also the ones who applied and weren’t successful.
From the publicly available data on venture-funded deals, we’ve been able to create algorithms to capture information on founders and then crawl the internet for publicly available image and educational background information. 3,496 founders were involved in the creation of the 2,002 businesses. Where founder information did not exist, those businesses were not included in the dataset. Given how important PR, networking and awareness is to startup businesses, this pathology is a rarity.
These days, facial and ethnic recognition computer vision technology is practically a commodity. One of our interns, a Warwick University Masters student, employed the techniques live in a real-world example in his interview process. Off-the-shelf software, however, is not fully tuned for specific use cases and in particular has been used on datasets that can lead to inherent biases, i.e. overfitting to Caucasian from a surfeit of white faces in the training datasets. Given the dataset size, we were well placed to train and manually review results that in particular expressed a degree of ambiguity.
My experience of working with computer vision and AI as the founder of Metail, where we had 12 world class PhDs on the team and also collaborated with Cambridge University on bleeding edge R&D, taught me that you always need to follow the 80:20 rule. This means not hiding away from having a ‘person-in-the-loop’ to ensure robustness of results. It is both efficient and effective. You can think of it as algorithms automating the decision-making for 80% of the data, leaving the remainder for fast manual checks. In our instance, the starting point is better than that and will continue to improve over time. For self-driving cars, 97% or even 99% is not enough. The ‘person-in-the-loop’ doesn’t work in this case. This is why the huge optimism five years ago that the era of driving automation was upon us has been replaced by the realisation that solving for the level of accuracy required for mass adoption will take a while.
Leaving money on the table
We now plan to go a step further with AI technology to parse Companies House data on accounts for businesses, filings that can be in any format, from handwritten scrawls to audited documents. This will enable us to build up a picture of business performance alongside the classification by the gender, ethnicity, place and educational background of their founding teams. It is all eminently doable and important to build up a picture of the economy that should ultimately enable policymakers to do a better job of maximising economic effectiveness and opportunity. Furthermore, it should enable financial gatekeepers to understand where money is being left on the table, where possibilities are being left un-nurtured and unexploited.
The data we have highlighted in our report is shocking and startling, but represents to me the potential at last for generating measurable systemic change. Information is power and if we positively take advantage of all the clearly under-invested potential the UK has to offer, then maybe, just maybe every British little girl regardless of their ethnicity or socioeconomic background will be able to see too “that this is a country of possibilities”.
Read our reports here.