Measuring Inclusion Without Turning Founders Into a Diversity Statistic

AVODA never split the 99 alumni of its completion study by sex, and it has no gender-split figure of any kind, so the definitions and counts that follow come from frameworks, statutes and studies written elsewhere. IRIS+ metric OI8197, “Founders: Female”, counts the founders of a company who are women and still hold an active role at the end of the reporting period, and it rolls up into OI2209, “Founders: Total”. That definition is how a gender-lens investor using IRIS+ sees a woman founder: as one unit in a numerator. A program can measure inclusion without shrinking her to that unit if it settles four things before the intake form goes out: why each question is asked, who receives the answer, how small a published table may get, and what she receives back.

Key Takeaways

The 2X indicators and IRIS+ define a woman founder for reporting as a share: OI8197 divided by OI2209. Under the 2020 guidance note, a company qualifies on founders only when women are the majority, so a firm founded by one woman and one man falls short.
Split every program metric by gender, as American ecosystem research advises, and a 25-founder cohort produces cells holding one woman. Her row is her business.
Re-identification needs few facts. A 2019 study found 15 demographic attributes made 99.98% of Massachusetts residents unique, and the UN humanitarian office’s data team judged 38 of 59 datasets too risky to publish as received.
Kenya’s Data Protection Act lists sex and marital status as sensitive personal data. Uganda’s Act requires a collector to tell the person, before collection, who will receive her data and how long it will be kept.
The Founder’s Copy sets four entries beside every inclusion question: purpose, recipients, smallest published cell, and what the founder receives back.
The sharper African complaint is that sex-disaggregated data is scarce. Count, with consent, and return the count to the people in it.

OI8197 and the founder who becomes a submetric

Start with what the frameworks achieved. The 2X Challenge, launched by 2X Global at the 2018 G7 Summit, reports raising US$34 billion in gender-lens investments between 2018 and 2023. Its criteria page calls the 2X Criteria “the global industry standard” for assessing such investments. Investors and companies “self-report” alignment, which requires basic safeguards, at least one gender data requirement and at least one of six criteria, with thresholds set by country and sector. That capital, aimed at women’s businesses, comes first in any honest account.

Then look at what the reporting does with her. In 2020, CDC Group and the Global Impact Investing Network published a guidance note mapping the 2X indicators onto IRIS metrics. Indicator 1B, “Percent of company founder(s) who are female”, becomes OI8197 over OI2209. A company qualifies under 1B when women form “the majority of the total founders”. Ownership qualifies at 51% of shares held directly by women. The founder counts if she keeps an active role, which the note says may be advisory.

Run two familiar East African cases through that arithmetic (both invented for illustration). A husband and wife who founded a poultry feed company together register as 50%, below the line, whatever she does in the business each day. A woman who founded a firm, sold most of it and now advises its male managing director counts in full. The metric tracks a cap table and a founding story. Her experience inside the enterprise sits outside it, and the IRIS+ page says as much: “No single metric is sufficient to understand an impact.”

Disaggregated by gender, in a cohort of twenty-five

The one outcome split AVODA has published is what 99 alumni told us in Who Actually Finishes: 15% registered, 25% informal, 45% with no business. Dividing 99 people by sex and then by sector could leave cells of a handful, so the questions below fall on our own forms first.

Banu Ozkazanc-Pan and Susan Clark Muntean’s 2022 book for Cambridge University Press, Entrepreneurial Ecosystems: A Gender Perspective, pushes past the headcount. Its first chapter complains that metrics, scales and questionnaires often treat women as a type of entrepreneur, and that inclusion work slides into a “fix the women” approach. Their closing chapter asks support organizations to set measurable goals, with metrics such as dollars invested, entrepreneurs mentored, event attendees and loan recipients, “all disaggregated by gender”. They also suggest exit interviews with women who leave, so a program records why.

The advice is sound where it was written. Their evidence comes from American cities such as Boston and St. Louis, where leading organizations track female participation in their membership and programming. At city scale, a split by sex averages over many people.

East African cohorts are smaller, and the split behaves differently. Take a cohort of 25 founders, nine of them women (the numbers are invented for the example). Split the women by sector and two cells hold one person each. A published row giving average revenue growth for women in food processing, with one founder behind it, publishes that founder’s revenue under the funder’s logo. Everyone in her cohort knows who she is, and so does her landlord if he reads the report. The more carefully a program follows the book, the more exactly it names her.

In 2019, Luc Rocher, Julien Hendrickx and Yves-Alexandre de Montjoye estimated in Nature Communications that 15 demographic attributes would make 99.98% of people in Massachusetts unique. That is a state of millions, and it took fifteen facts. In a room of 25 founders, sex plus sector plus district can be enough.

A 2019 guidance note from the data team of the UN Office for the Coordination of Humanitarian Affairs (OCHA) explains that removing names and phone numbers still leaves respondents exposed when location or ethnicity are combined. From the start of 2018, the team assessed 59 datasets uploaded to its sharing platform and judged 38 of them too risky to publish as received. Disclosure control lowered the risk to an accepted level for 14, and the team later tightened its default threshold from 5% to 3%. The note is frank about the cost: every protective step removes some information value, and the balance is set in advance with the people who own the data.

A program manager can borrow that discipline without the software. Fix the smallest group you will publish before the results arrive, merge sectors or districts until every cell clears it, and print the rule beside the table.

Before it crosses to Kampala, then, the book’s advice needs a size condition: disaggregate inside the program, where the numbers drive decisions about her, and aggregate before the numbers leave.

What Kenyan and Ugandan law already asks of an intake form

Kenya’s Data Protection Act of 2019 defines “sensitive personal data” to include marital status, family details and the “sex or the sexual orientation of the data subject”. Section 25 requires that personal data be collected for explicit purposes, limited to what is necessary, explained when it touches family or private affairs, kept in identifiable form only as long as needed, and sent outside Kenya only with proof of safeguards or with consent. A results platform hosted by a funder in another country is a transfer.

Uganda’s Data Protection and Privacy Act of 2019 makes the collector “accountable to the data subject” and asks for her “transparency and participation” in how her data is used (section 3). Section 13 lists what she must hear before collection: the purpose, whether answering is discretionary or mandatory, the consequences of declining, who will receive the data, and how long it will be kept.

Ozkazanc-Pan and Clark Muntean advise checking websites and promotional material so women appear in proportion to men, a fair goal for any program brochure. In September 2023, Kenya’s Office of the Data Protection Commissioner fined Casa Vera Lounge on Ngong Road about 1.9 million shillings for posting a patron’s photograph on social media without consent. A woman founder’s portrait on an impact report page is her personal data too. Proportion on the poster comes after her signed permission.

A Ghanaian respondent and the case for counting more

Pollicy, the Kampala data organization, published a 2024 study of gender data written by Zoe Bobina with My Data Rights. It ran focus groups and interviews in Côte d’Ivoire, Ghana, Kenya and Zambia. It found that even basic sex-disaggregated data “remains minimal” in official sources and that few policies require it. A civil society respondent in Ghana put the cost in one question: “How well can you tell that you have been able to empower as many males as you have done females?”

The book makes the same case from Boston: “Transparency and measurement provide the opportunity for stakeholders to keep leaders accountable” for their commitments, where good intentions alone leave nothing to check. Read together, the two sources describe the commoner East African failure, which is the uncounted woman. A program that cites dignity to avoid publishing any split is protecting itself, and a funder is right to press it.

That point stands, alongside the study’s own caution that its interviews are few and its findings cannot be generalized. The Founder’s Copy, set out next, answers the Ghanaian question: the program keeps every split it needs internally, publishes the splits large enough to be safe, and returns the result to the women it describes. The count survives. What changes is the order of its readers, with the founder ahead of the report.

The Founder’s Copy

Here is one inclusion question written up so that the founder herself could read it and agree. Field: sex. Purpose: funder template and the program’s check on whether women finish at the same rate as men. Recipients: program staff, then the funder in aggregate. Smallest cell: the floor printed on the form. Returned: the cohort’s completion rates by sex, shared with every founder at graduation.

That entry is a Founder’s Copy, and a program writes one for every question on the intake form, the baseline survey and the exit interview. The purpose names the decision inside the program that the answer feeds. A question on who else must sign before her business borrows can feed guarantor support and session timing. A sex field that only the funder’s template uses is marked as a template field, and the form says so and makes it optional, as Uganda’s section 13 already requires.

The recipients are listed by role: program staff, the evaluator, the funder, the funder’s reporting platform and the country it sits in. A founder who reads the list can choose what to answer. The smallest cell is the minimum number of founders behind any published figure, written down before the cohort starts. Where a cell falls short, merge categories or report it as below the floor. A program that sets the floor after seeing the results has set a convenient floor.

The return is what each question owes the founder: her own figures beside the cohort’s, the pattern from the exit interviews, and a date when her identifiable answers are deleted, which both statutes expect.

Bobina’s study draws a line that sharpens the purpose entry. Sex-disaggregated statistics count women and men, while gender data goes further and asks about the norms that shape their outcomes. A count of women in the cohort serves the funder’s numerator. A question about who controls the business account serves the founder’s next twelve weeks. The Founder’s Copy keeps both and labels each.

The king’s census and the lost sheep

David ordered the census “so that I may know how many there are”, and Joab, his commander, asked the question a Purpose entry would ask: “why does my lord the king want to do such a thing?” The count went ahead. It took nine months and twenty days, it returned a total of fighting men from Dan to Beersheba, and afterward “David was conscience-stricken” (2 Samuel 24). That number served the king’s sense of his own strength. In Luke 15:4 to 6 a shepherd with a hundred sheep knows the number well enough to notice when it reads ninety-nine, and his count ends with one sheep carried home on his shoulders. The Ghanaian respondent’s demand for a count is just, and the two stories settle whom the finished count belongs to: the king’s ended in his own regret, the shepherd’s with the one carried home.

Frequently Asked Questions

What is IRIS+ metric OI8197?

OI8197, “Founders: Female”, is the IRIS+ metric that counts an organization’s founders who are women and still hold an active role at the end of the reporting period, which can be an advisory role. Divided by OI2209, “Founders: Total”, it gives the founder share that the 2020 2X guidance note uses to qualify investments under the entrepreneurship criterion.

Is a founder’s sex sensitive personal data in Kenya and Uganda?

Kenya’s Data Protection Act of 2019 lists sex, sexual orientation, marital status and family details as sensitive personal data, with stricter processing grounds. Uganda’s Data Protection and Privacy Act of 2019 requires consent and, before collection, notice of purpose, recipients, retention period and whether answering is optional.

How small is too small for a gender-disaggregated table?

No universal number fits every program. Set a minimum count of founders behind any published figure before results arrive, then merge sectors or districts until each cell clears it. OCHA’s humanitarian data team treats re-identification risk above 3% as too high by default, and it agrees the threshold with each data owner.

Should accelerators stop collecting gender data from founders?

Keep collecting it, because sex-disaggregated data is already scarce across African programs and policies. Change how it travels: state the purpose and the recipients on the form, make template-only fields optional, analyze every split inside the program, publish only cells above a fixed floor, and give founders the results.

What should a founder ask before answering a program’s gender questions?

Ask who will see the answer, whether it leaves the country, whether the question is optional, how long it is kept and what you receive back. Uganda’s law already requires a collector to tell you most of this before collection, and a program that welcomes the question is a safer place to answer it.

Next Step

A Founder’s Copy for each inclusion question, a smallest cell fixed before results arrive and a consent line in plain words are the three outputs of a strategy call with Joshua built around your intake form. Send the form and the funder’s results template ahead of the call. A form corrected before the next intake protects every founder who fills it in.

Sources and Evidence

Ozkazanc-Pan and Clark Muntean, Entrepreneurial Ecosystems: A Gender Perspective, Cambridge University Press, 2022: metrics that treat women as a type of entrepreneur; metrics “all disaggregated by gender”; exit interviews; transparency and accountability.
2X Challenge, 2X Criteria: self-reported alignment, the gender data requirement, country and sector thresholds, US$34 billion raised 2018 to 2023.
CDC Group and GIIN, How to Measure the Gender Impact of Investments: Using the 2X Challenge Indicators in Alignment with IRIS+ (2020): indicator 1B as OI8197 over OI2209, the majority-founder rule, the 51% ownership rule, the active-role definition.
IRIS+, Founders: Female (OI8197): metric definition, active role, submetric of OI2209.
Rocher, Hendrickx and de Montjoye, “Estimating the success of re-identifications in incomplete datasets using generative models”, Nature Communications, 2019: 15 attributes make 99.98% of Massachusetts residents unique.
OCHA, Guidance Note: Statistical Disclosure Control (August 2019): re-identification through combined variables, 38 of 59 datasets too risky, the 5% and 3% thresholds, the information-loss trade-off.
Kenya, The Data Protection Act, No. 24 of 2019: section 2 definition of sensitive personal data; section 25 principles on purpose, necessity, explanation, retention and transfer.
Uganda, Data Protection and Privacy Act, 2019 (ULII text): sections 3, 7, 13 and 18 on accountability, consent, notice before collection and retention.
Capital FM, “Data Commissioner fines 3 firms Sh9.4mn over data infringements” (26 September 2023): Casa Vera Lounge fined about Sh1.9 million for posting a patron’s image without consent.
Nairobi Law Monthly, “Data Protection Commissioner fines school Sh4.5 million for posting minors’ pictures without consent” (26 September 2023): the same September 2023 penalty notices.
Bobina, Unlocking the Power in Gender Data: An Afro-feminist Approach to Data Governance (Pollicy, 2024): scarcity of sex-disaggregated data, the Ghanaian respondent, gender data beyond sex counts, the study’s limits.
2 Samuel 24:1-10 (NIV), Bible Gateway and Luke 15:3-7 (NIV), Bible Gateway: David’s census and the shepherd’s hundred.