The Self-Report Problem in Every Founder Assessment, Including Ours

Our own Business Assessment Matrix asks a founder to score her venture on seven criteria, weighted 50% business model, 30% innovation and 20% team strength, and to pick the statement that “fits reality.” Every one of those seven scores is her answer about her own business. That is the self-report problem, and it sits inside almost every founder assessment, application form and alumni survey in Kampala: the evidence comes from the person the decision is about. A self-report serves well when little rides on it and grows fragile when a seat, a grant or an impact figure depends on it, so grade each answer by who stands behind it and check the few that carry the decision.

Key Takeaways

Most founder application forms collect self-report: revenue, staff, registration status, customers served. The founder is the only witness to each answer.
People shade answers toward what the asker seems to want. In 450 interviews in Bangalore, respondents were about 2.5 times more likely to prefer a device they believed the interviewer had built, even when the two devices were identical.
Asking directly can still beat elaborate measurement. A World Bank study in Sri Lanka found that simply asking owners for profit gave a more accurate figure than computing it from detailed revenue and cost questions.
The Said, Shown, Checked Ladder grades every field by its evidence: what she says, what she can show on a document she already holds, and what someone other than her confirms.
In Uganda, 61% of businesses in a 2024 national MSME survey were unregistered, so the Shown and Checked rungs rely on mobile money, receipts and customers more than on registries.
AVODA’s published alumni figures are self-report too, and they should carry that label.

What does a founder application form actually measure?

Picture a Kampala accelerator’s program manager with three hundred application forms to score before Friday’s selection meeting, or a donor impact report due at quarter end. The form asks for monthly revenue, employees, registration status, customers served and women on staff. The founder fills in each field about her own firm, while hoping for a seat.

Forms like this are good tools. They are cheap, they scale to hundreds of applicants, and they treat founders as adults who know their own businesses. They also give a fair start to a founder who lacks a network to vouch for her.

Data-Driven Decision Making in Entrepreneurship, edited by Nikki Blacksmith and Maureen E. McCusker (CRC Press, 2024), argues for standardized assessment from the investor’s side. Blacksmith, Kelly Diouf and McCusker call unstructured interviews and reference calls, the methods most investors use, costly and of little predictive value. Nilima Ajaikumar and Reece Akhtar describe psychometric assessments as typically taking the form of surveys and as “more objective than relying on an individual’s intuitive judgment.” The same chapter notes in passing that resumes get dressed up, then recommends surveys in which the founder describes herself.

Why does an honest founder still give a shaded answer?

An answer on a form carries information when shading it costs something or when someone can check it. Attach a prize to the answer and leave it unchecked, and the answer starts to describe what the founder hopes the reader wants to hear. Nobody has to decide to lie. Courtesy and a little rounding up do the work.

Nicola Dell and four co-authors, three of them at Microsoft Research India, ran 450 interviews in Bangalore. Respondents were about 2.5 times more likely to prefer a technology they believed the interviewer had developed, even when the alternative was identical. When the interviewer was a foreign researcher working through a translator, the bias rose to about five times. The respondents were being polite to the person in front of them, and the data recorded the politeness.

Money sharpens the same effect. Dean Karlan and Jonathan Zinman, in a study published by Innovations for Poverty Action with data from Peru and the Philippines, start from the practice they wanted to test: lenders usually judge how borrowers use their loans from “unverified self reports from client loan applications.” When independent surveyors asked borrowers directly, the answers understated how much of the loan went to uses outside the business, compared with a method that let borrowers hide individual answers from the surveyor. An accelerator applicant sits in the same seat as those borrowers, asking a funder for money.

The strongest case for simply asking

Suresh de Mel, David McKenzie and Christopher Woodruff tested profit measurement among microenterprises in Sri Lanka for the World Bank. They asked owners directly for their profit, and they also built profit up from detailed questions on revenue and expenses. They sent research assistants to observe firms at random times 15 to 16 times in a month. Their conclusion: firms underreport revenues by about 30 percent, and “simply asking profits provides a more accurate measure of firm profits than detailed questions on revenues and expenses.”

I concede the point in full. A well-asked direct question can beat a complicated constructed measure. Uganda’s own evidence base runs on asking. When Nathan Fiala evaluated loans, grants and business training for 1,550 microenterprise owners in four semi-urban districts, the outcomes were measured through follow-up surveys of the owners. Serious research in our region depends on what founders say.

Look at how de Mel and his colleagues knew the direct question was more accurate, though: research assistants sat with the firms and watched. Their finding rests on a check. It tells us which question to ask, and leaves open which answers to trust when a cohort seat depends on them. So the argument here is about stakes and verification, and the question stays on the form.

The Said, Shown, Checked Ladder

This framework grades any field on a founder form by the evidence behind it, in three rungs.

Said. The founder’s own answer. Monthly revenue typed into a box, “five employees”, “registered”, “we serve 400 customers.” Fast and unverified.
Shown. A document she already holds that supports the answer: a mobile money statement, a receipt book, an invoice, a delivery note, a photograph of the shop with the date on it, a short work sample. The book’s list of predictive methods includes work samples, where the candidate performs the task in front of the assessor; they belong here.
Checked. Someone other than the founder confirms the answer: a search of the business registry, a call to two named customers, a visit on an unannounced day, an attendance record kept by the program rather than by the founder.

Victoria Mattingly and her co-authors in the same book supply the Checked rung in their own words. On measuring whether a founder runs an inclusive team, they write that inclusion “should be measured through ‘other-report’ data,” because what counts is how the team experiences the founder.

Verification costs time, so spend it where the decision jumps. For each selection round, find the two or three fields that carry the most weight in the score, and the applicants sitting near the cut line. Those fields, for those applicants, deserve a Shown or Checked step. A founder far above or far below the line keeps her Said answers, because checking them would change little.

Where the book’s American frame breaks down in Kampala

Data-Driven Decision Making in Entrepreneurship is written for American venture capital, where a startup leaves a thick paper trail of incorporation records, payroll, bank accounts and audited statements. In that world a survey of the founder’s traits is the soft layer on top of hard records. In Uganda the hard layer is thinner.

The State of Entrepreneurship in Uganda 2024 study, run by Ichuli and Ipsos with support from the Mastercard Foundation and published by the Ministry of Trade, Industry and Cooperatives, surveyed 3,062 micro, small and medium enterprises. It found that 61% of businesses were unregistered, rising to 70.2% in the Eastern Region. Only 8.3% used software for operations such as accounting, and 53.4% lacked both a smartphone and a computer. For most Ugandan applicants, then, a registry search comes back empty and a request for management accounts produces a blank page.

That changes how the ladder is built here. The Uganda Registration Services Bureau runs public registry searches, some of them free, so a “registered” answer can move to Checked in an afternoon. An “informal” answer sits outside the registry, so its Shown rung must come from what the founder actually holds: a mobile money statement, a receipt book, a supplier’s invoice. Its Checked rung comes from people: the customers, suppliers and landlord who deal with the business every week.

The book also misses the grant economy. Its fourteenth chapter, by Jessica Hart and co-authors, describes an impact-linked carry model in which customer surveys are “administered by startups,” and third-party verification is reserved for the startups that exit. When the party paid on a number also collects that number, the number drifts toward the payment. In a donor-funded ecosystem, where a founder may answer forms for several programs a year, that pressure is higher still. The book leaves it unexamined: a search of the full text for faking and social desirability turns up only a passing citation on impression management in the chapter on text analysis.

Applying the ladder to our own numbers

Turn the ladder on AVODA, because a program that grades only other people’s evidence is teaching the wrong lesson.

Our Business Assessment Matrix is Said from top to bottom: seven self-scored criteria and an instant result that stays private to the founder unless she saves it. At those stakes it works as a mirror, and a founder gains little by flattering herself in one. Our Founder Readiness Assessment asks written questions such as “How do you know it will be profitable?” and tells the applicant to use numbers. Those answers are Said as well, though a reader can judge the reasoning even when the figures are unverified.

AVICS, the AVODA Interest and Credit Score, is described on our program page as built on punctuality, financial accountability and cognitive assessments, and the same platform hosts classes, payments and attendance. Where a score input comes from that platform’s record of what a founder did, it sits on the Checked rung, which is stronger evidence than anything she tells us about herself.

Then there are the alumni figures we have published. In Who Actually Finishes we reported a drop-out split from 99 alumni: 15% registered, 25% informal, 45% without a business. The essay’s own title calls it what 99 alumni say. Those figures are Said. The registered share could move to Checked through the same registry search we would ask of any applicant, and the informal share needs the people around each business. Any alumni survey also has a blind spot beyond every rung: the alumni who stayed silent. The split should travel with that label until the checking is done.

Honest weights on both sides of the scale

Everything above rests on one premise: a founder’s honesty on a form responds to the price the form puts on it. Verification raises the cost of drift, and that is right as far as it goes. Yet the ladder reaches only what can be checked, and in an economy where most firms sit outside the registry and few keep books, most of what a program learns about a founder will always be Said. Proverbs 11:1 puts it in the language of the marketplace: “A false balance is abomination to the LORD: but a just weight is his delight.” The verse holds both parties to account. The trader is responsible for the weight she reports, and the program that sets the scale is responsible for whether it rewards a heavy thumb. A form that prizes a flattering answer is a false balance of its own. Stewardship of a selection process means pricing honesty correctly from the program’s side, and it means asking founders for a just weight in the many places where the only one watching the scale is God. That integrity is the part verification cannot supply.

Frequently Asked Questions

What is the self-report problem in founder assessment?

It is the gap between what a founder says about her business and what is true, which widens when a decision rides on the answer. Most application forms, readiness tests and alumni surveys are filled in by the founder herself, so the evidence comes from the person being judged. Stakes and checking decide how far to trust it.

Should accelerators stop using self-reported data?

Keep it. Self-report is cheap, scales to hundreds of applicants and, when asked well, can be accurate. A World Bank study in Sri Lanka found a direct profit question beat detailed revenue and cost questions. Add a Shown or Checked step to the two or three fields that carry the most weight in each decision.

How can a program verify a founder’s answers in Uganda?

Match the check to the answer. A registered business can be confirmed through a Uganda Registration Services Bureau registry search. Revenue can be supported by a mobile money statement or receipt book. Customer numbers can be tested by calling two named customers. Attendance and payments recorded by the program are stronger than any claim.

What is the Said, Shown, Checked Ladder?

It grades every field on a founder form by its evidence. Said is the founder’s own answer. Shown is a document she already holds, such as a mobile money statement or invoice. Checked is confirmation from someone else, such as a registry, a customer or a program record. Verification is spent on the heaviest fields near the cut line.

Are AVODA’s own alumni figures self-reported?

Yes. The drop-out split we published from 99 alumni (15% registered, 25% informal, 45% without a business) reports what alumni told us, and alumni who stayed silent are outside the figures. The registered share can be checked against the business registry; the informal share needs customers and suppliers to confirm it.

Next Step

Before your next selection round or impact report, take your application form and mark every field Said, Shown or Checked, then pick the two or three Said fields that carry the most weight. If you would like a second pair of eyes on that form before the meeting, book a strategy call with me this month. Thirty minutes on the form now is cheaper than defending an unchecked number to a funder later.

Sources and Evidence

Blacksmith and McCusker (eds.), Data-Driven Decision Making in Entrepreneurship, CRC Press, 2024: chapters by Ajaikumar and Akhtar (psychometric surveys, resumes), Blacksmith, Diouf and McCusker (interviews and reference calls, work samples), Mattingly et al. (“other-report” data) and Hart et al. (startup-administered impact surveys). DOI 10.1201/9781003196853.
de Mel, McKenzie and Woodruff, Measuring Microenterprise Profits: Don’t Ask How the Sausage is Made, World Bank Policy Research Working Paper 4229 (2007): Sri Lanka; revenues underreported by about 30 percent; direct profit question more accurate; 15 to 16 random observation visits a month.
Dell, Vaidyanathan, Medhi, Cutrell and Thies, Participant Response Bias in HCI, CHI 2012: 450 interviews in Bangalore; about 2.5 times preference for the interviewer’s device, about 5 times with a foreign interviewer and translator.
Karlan and Zinman, List Randomization for Sensitive Behavior: An Application for Measuring Use of Loan Proceeds (Innovations for Poverty Action): Peru and the Philippines; lenders rely on unverified self reports; direct questions understate non-enterprise use of loans.
Innovations for Poverty Action, Evaluating the Impact of Loans, Grants, and Training on Small Businesses in Uganda: Nathan Fiala’s study of 1,550 microenterprise owners in four semi-urban districts, with outcomes measured by follow-up surveys.
Ministry of Trade, Industry and Cooperatives, The State of Entrepreneurship in Uganda 2024: Profile of Uganda’s MSMEs: 3,062 MSMEs surveyed by Ichuli and Ipsos; 61% unregistered (70.2% in the Eastern Region); 8.3% use software such as accounting software; 53.4% have neither smartphone nor computer.
Uganda Registration Services Bureau, URSB Registry Searches: business name and company searches, some free of charge, detailed reports with a payment reference number.
AVODA, Business Assessment Matrix: seven self-scored criteria weighted 50% business model, 30% innovation, 20% team; results private unless saved.
AVODA, Founder Readiness Assessment: written self-report questions, including “How do you know it will be profitable?”
AVODA Blue program page: AVICS described as built on punctuality, financial accountability and cognitive assessments, on a platform that hosts classes, assessments, payments and attendance.