Ravikanth Andhavarapu

What 800 nonprofit filings show about proving your own numbers

Tue, 28 Jul 2026 03:30:00 +0000

I spent a week reading the public filings of 800 American nonprofits, and I want to describe one pattern I did not expect to find, along with the thing that made it visible.

The organizations are all mid-sized — between five and twenty-five million dollars of annual revenue — working in human services, education, youth development and international development, spread across 54 states and territories. I pulled every Form 990 available for each, up to thirteen years apiece.

I was looking for something narrow: whether the numbers these organizations report to their funders can be traced back to the systems that produced them. What I found first was something simpler, and stranger.

Revenue that moves without the business moving

A Form 990 splits incoming money into a few lines. Two of them carry almost everything a service organization receives. Contributions and grants covers donations, foundation grants and government grants. Program service revenue covers money earned by delivering the mission — fees, contracts, payments for services rendered.

In 118 of these organizations — 16% of those with at least five years of comparable filings — one of those two lines moves by at least 25 percentage points of total revenue in a single year, while total revenue stays essentially flat.

Not the organization growing. Not a funder arriving or leaving. The same money, landing on a different line.

One organization reported 93 to 96% of its revenue as contributions for eleven consecutive years, then 5% the following year — with total revenue slightly up. Another sat at 98–99% for nine years, then dropped to 11%, while revenue rose by five and a half million dollars. A third has crossed that boundary in both directions, twice, and currently sits roughly where it started a decade ago.

Thirteen organizations oscillate — the line moves one way, then back again.

Why this is not an accounting scandal

I want to be careful here, because the obvious reading is wrong.

There is a genuine, well-known ambiguity underneath this. When a government pays a nonprofit, someone has to decide whether the government is buying a service — an exchange transaction, which belongs in program service revenue — or funding a mission without receiving commensurate value in return, which makes it a contribution. For Medicaid-funded services, state contracts and pass-through county money, that question is arguable in good faith.

It was arguable enough that in 2018 the Financial Accounting Standards Board issued a standard, ASU 2018-08, specifically to reduce the inconsistency. It pushes in one direction: most government grants and contracts are contributions, because the government does not receive commensurate value.

So a reasonable person would expect these moves to cluster around 2019 to 2021, when that standard took effect, and to run in one direction.

They do not.

Year of move	Count
2012–2018	76
2019–2021	30
2022–2024	32

More than half the moves predate the standard entirely. They run in both directions. And when I read the audited financial statements behind a sample of these organizations, only two of twenty-three reports mention ASU 2018-08 at all — one of them as boilerplate about a pronouncement not yet adopted.

I also went looking for the most obvious false positive. Federal relief money arriving in 2020 and leaving again produces exactly this signature: a line jumps, then returns. I found 30 organizations in that pattern and excluded every one of them. The 118 are what remains after that.

So this is not fraud, and it is mostly not a standard being adopted. It is something less dramatic and more interesting: a judgment call that gets made again each year, by whoever is closest to the filing at the time.

Why it matters more than it looks

If this were only about presentation, it would be a curiosity. It isn’t, for three reasons.

The public support test. A public charity has to demonstrate, on a rolling five-year lookback, that its money comes from a broad base. Contributions and program service revenue are weighted differently in that calculation. A line that moves eight million dollars inside the lookback window doesn’t just make year-over-year comparison awkward — it moves an input to the test that determines the organization’s tax status.

Funder reporting. Grant reports, board packets and renewal applications are built on multi-year comparisons drawn from these same categories. If the underlying definition changed in year three and the comparison wasn’t rebuilt, the report quietly compares two different things. Nobody notices, because the report is assembled by a person who knows what they meant.

That last clause is the actual finding. A number that shifts category without a documented reason is a number that no system is producing. It is being decided. And a decision that lives in one person’s judgment rather than in a rule is a decision that leaves when they do.

The document nobody reads

Partway through this I realized I was using the wrong source.

The Form 990 is the famous one — public, searchable, the basis of every charity rating site. But it is a derived document, and it is old. Because organizations take the automatic six-month filing extension and the IRS then needs months to process and publish, the most recent fiscal year with near-complete coverage in my sample was 2023. Only 12% had a 2024 filing available.

Meanwhile, any organization spending a million dollars or more in federal awards must file a Single Audit — a full financial statement audit plus a compliance audit against each major federal program — with the Federal Audit Clearinghouse. Those are due nine months after fiscal year end, with no extension, and they post quickly.

In a 250-organization sample, 42% had a Single Audit on file. Of those, 82% had a 2024 or 2025 audit available — against 12% with a 2024 Form 990.

The audit package contains the audited financial statements the 990 is derived from, the notes that explain the accounting policies, every finding the auditor raised, and the name and title of the person who signed for the organization. It is free, public, and available through an API.

It is a strictly better document, roughly two years fresher, and almost nobody outside the compliance world reads it.

The most interesting thing I found came out of that comparison. One organization’s audited statements state plainly that it adopted ASU 2018-08, under which grants and contracts are treated as contributions. Its Form 990 for the same fiscal year shows contributions falling from 100% of revenue to 20%. Two documents, one year, one organization, pointing in opposite directions.

I am not naming it, and I want to be explicit about why. There is almost certainly an explanation, the organization has done nothing wrong, and publishing a list of charities whose filings look odd would be a way of manufacturing alarm in order to sell a remedy. The pattern is worth writing about. The names are not mine to publish.

What I think this says

Every organization here has an auditor, a finance lead and a board. The inconsistency isn’t a competence problem, and it isn’t a software problem — most of them have perfectly good accounting systems.

It is a seam problem. Financial data lives in the accounting system. Delivery data — who was served, how often, with what result — lives in case management systems, service logs, timesheets and spreadsheets. Funder reports need both, cut on calendars that match neither the fiscal year nor each other. Federal awards run October to September, many state contracts July to June, foundations on their own grant anniversaries.

Nobody’s job title covers the join. So it gets done by hand, by someone who is good at it, every reporting cycle, forever. It works. The reports go out. The audits come back clean.

And the cost of it is invisible, because it’s absorbed rather than spent — until the person who knew how the spreadsheet worked takes another job, and the organization discovers that a decade of its reported history rested on their memory.

The analysis used the ProPublica Nonprofit Explorer API for Form 990 data and the Federal Audit Clearinghouse API for Single Audit data. Both are public and free. Figures are counts from the sample described, not national estimates.

I do independent assessments of exactly this problem for nonprofits — whether the data can prove what the programs actually do. If it’s relevant to you, my email is ravi@helloravi.com.

How do we ensure every human can get a loan of 10USD?

Mon, 12 Mar 2018 05:42:38 +0000

Winance is a KYC-powered Blockchain-enabled financing system of the world.

Winance is about addressing the finance of the human beings in the world. WIN is the name of the coin.

Why?

Access to capital is a very crucial tool to eliminate poverty and increase productivity. How can we ensure that every human can get at least 10USD worth of loan? If we can get them a small loan they will be able to get a bigger loan as they build trust on the system. The initial number is not important.

What?

Here is an example of how a loan transaction might happen:

Tanvi(in India) is looking for loan to buy a bike from Royal Enfield. The bike costs 1,00,000INR. Keithr equests KYC data on the chain which Tanvi provides and chooses to fund depending on the data.Keith from Europe wants to fund the loan. Keith buys WIN token worth 1 Lakh INR and transfers it to Tanvi on condition that it is used for buying a bike. Tanvi can use that money only for bike purchase by transferring that WIN token to Bike owner. Bike owner converts it into INR through an exchange. Tanvi keeps paying back to Keith in installments through WIN token which he can use it to loan to someone else.

How?

Current crypto currencies have an inclusiveness problem. WIN coin is an inclusive decentralized system where every human being starts with some coin as soon as he/she does their KYC(after ensuring that there is a reasonable probability that it is not duplicate) is needed. This way every human being is interested in both doing their KYC and particpating in the economy.

Loans disbursed in this system are a lot more safer than ordinary loans. Why is that? Loan disburser can be sure that the money is being used for the right purpose by tracking where the loan amount is going.

There is an entire futures and options market place that needs to run to peg the WIN coin to a certain amount for each individual contract.

Coins get created when somebody does KYC. We ensure some inflation exists which incentivizes players who attract other people to come and do their KYC and generate their own KYC. We will need to involve people who verify other people’s kyc and contribute to the authenticity of KYC.

What are the benefits this offers?

It is inclusive. Everybody will want to be part of the financial system because their KYC brings some cash and makes loans accessible to them. Cheaper capital can fund more expensive capital.
It creates an ecosystem for finance guys to make money by being part of the loan disbursing system (through offering options and contracts to those who want to give loans)
The coins accrue value whenever a company accesses the data of these people.
The data accumulated will hold a huge amount of long term value.

The most important thing that we need if you want to work with me

Thu, 13 Apr 2017 05:42:38 +0000

Pro-activeness is absolutely necessary. If you don’t have it, we don’t want you to be part of the team, You may be a nice person but we don’t want to work with you.

How does it translate on a day to day basis?

Wrong thinking -Nobody told me what to do so I am waiting and doing nothing.
Right thinking - How do I find out what I should do? Should I ask Venky or Ravikanth?

Wrong thinking - I already asked Ravikanth, he did not tell me anything
Right thinking - how do I make sure Ravikanth tells me what to do? Should I call him again?

Wrong thinking - I have fever, I will wait for Venky and Ravikanth to call and find out.
Right thinking - I will leave a message to them so that they are not concerned about it.

Wrong thinking - Ravikanth did not ask me. I did not tell.
Right thinking - Over-communicate. Don’t make us guess.

Getting started with phabricator

Thu, 13 Apr 2017 05:42:38 +0000

This has been shamelessly copied from the post written https://www.facebook.com/notes/abhilash-inumella/moving-to-code-conversations-from-feature-conversations/10153441383084384/ by my friend abhilash-inumella

Writing software is at the heart of a consumer internet company like ours. All of the heavy lifting is done by the code we write. But the irony of the current situation is that we’re talking very less about the code. To overcome this problem Facebook, Quora, Dropbox, Uber, Pinterest, KhanAcademy internally use http://phabricator.org/. It’s a tool to communicate code. It’s for reviewing code. It’s a tool to get the shit done. It’s the only review tool I’ve seen that doesn’t suck.

Action items:

some_install_path/ $ git clone https://github.com/phacility/libphutil.git
some_install_path/ $ git clone https://github.com/phacility/arcanist.git

Now add some_install_path/arcanist/bin/ to your PATH environment variable. When you type “arc”, you should see something like this:Usage Exception: No command provided. Try 'arc help'. source /path/to/arcanist/resources/shell/bash-completion

Software Engineer Workflow:

git pull: pull from remote master
git branch -b 'feature-awesome' : you make a feature branch
git commit : implement the feature
arc diff : send it for review by including a proper test plan
git commit --amend : incorporate changes suggested by the reviewer
arc diff: send it for review again
git commit --amend : incorporate changes suggested by the reviewer
arc diff : send it for review by including a proper test plan
git commit --amend : incorporate changes suggested by the reviewer
arc diff: send it for review again
git commit --amend : incorporate changes suggested by the reviewer
....... ITERATE UNTIL THE REVIEWER IS SATISFIED ......
Once the reviewer is satisfied: make sure you're upto speed with changes others have made
git checkout master;
git pull;
git checkout feature-awesome
git rebase master
git push

From now on all the code that goes onto master will be reviewed first on Phabricator. Let me know what you guys think!

Hands On Introduction to Machine Learning

Thu, 13 Apr 2017 05:42:38 +0000

One of the important aspects of being good at solving business problems with tech is to understand the scope and limitations of different technologies.

As a part of this I asked a friend of mine(Bracket) from HyperVerge who also happens to the co-founder of the company to point to resources which introduce machine learning.

The most common suggestion is to do the Andrew NG course on machine learning.

Given that I can code, I was looking for something more hands on.

http://neuralnetworksanddeeplearning.com/ - Hands on intro to machine learning
Tom Mitchel’s machine learning - A light weight but great book
https://shapeofdata.wordpress.com/introduction/ - Provides a very intuitive explanation of many machine learning techniques
Coding a simple ML algo: Quora Answer
General advice on Machine Learning: Quora Answer

MySQL Cheat Sheet

Sun, 02 Apr 2017 05:42:38 +0000

On OSX MySQL the following are start, stop and restart commands

sudo /usr/local/mysql/support-files/mysql.server start

sudo /usr/local/mysql/support-files/mysql.server stop

sudo /usr/local/mysql/support-files/mysql.server restart

Logging in

mysql -u root

mysql -u root -p (will prompt for password)

More Commands

How to reduce software development cost by upto 90 percent

Wed, 22 Mar 2017 05:42:38 +0000

The worst question to ask a non-technical person

What do you want?

When you ask this question they never communicate the problem they want to solve. They usually come up with a solution based on their experience of usage of different websites for the problem they have in mind. They communicate the solution which they came up with instead of communicating the problem they are facing. What is the limitation of doing this?

The person coming up with this solution does not know all the different options available if she is not a technical person. That is not the only problem. Somethings are extremely easy for machines and some seemingly very simple things are very hard for machines. If the actual problem can be solved by avoiding the hard parts you will have a solution which is fast and easy to build.

If you are a non-technical person communicate both the problem and the solution you have in mind. Communicating both of them is important. This provides enough detail into the problem for the tech person to address the problem.

Brutality is the key to execution

Sun, 19 Mar 2017 05:42:38 +0000

Key to execution

The only thing which has worked for me in ensuring execution is brutality. Be absolutely brutal about the things you want to execute well. Keep this list small. Don’t fill the day with lists of things you want to do and try to be brutal about it. I have two things I am brutal about and probably going to add one more but I am still unsure of it.

How to be brutal

Tie each of your important activities to any of the compulsions you have.

Hunger is a compulsion that I have and you might too. For instance, don’t eat your lunch until you are done with having a 3 hour focussed time for yourself at work. Having focussed time to work can be a rarity in many offices. This also helps manage the social distractions. Tell people that they are denying you lunch by interrupting you.

Using my phone is a compulsion for me I have a a rule to not touch my phone or use it until I am done with my yoga in the morning. Only when I follow this rule strictly I end up doing everything in the morning. If I slack off a little then it is gone.

Making execution easy

If you want to be ensure a certain activity, one of the thing that works is to do it for 45 days. Then you dont have to do it. You just have to prevent yourself from not doing it. That is huge advantage in itself. Seinfeld popularized a similar strategy if you are interested.

I don’t know of any other way of ensuring execution. Let me know if you do.

Criteria and Job Description for Choosing a Tech Lead and junior developer

Wed, 11 Jan 2017 05:42:38 +0000

Criteria for Choosing a Tech Lead

Should know design patterns, anti patterns. If they have used UML diagrams in a project that would be a plus.
Should know test driven development and behaviour driven development
Very good code quality(no code smells)
Knows enough about software design to build a flexible application.(should have built an application being conscious of costs associated with design)
Should have mobile full stack experience and some Android experience, and spent some time building apps on Nodejs
Should be a good listener
Mentees should speak well of him
Communication should be clear

Criteria for Choosing a Junior Developer

Should be able to build a full stack app on their own in any of the full stack frameworks. Skill in node is highly preferable
Should be able to break larger problems into smaller problems
Should be able to learn quickly with little help on their own
Good listener
Bonus points for people who can ask thought provoking questions in a discussion
2-3 hour session time

My Favourite Story

Tue, 13 Sep 2016 05:42:38 +0000

This happend in ancient India. A man who is unable to tolerate the tantrums of his wife and the burden of the family decides to become a “Sannyasi”. In those times divorce was not an option in India. “Sannyas” was misused as a way to get out of the family burdens in a respectable manner.

Sannyasi is someone who spends his whole time on attaining realization or enlightenment. He leaves his family and travels to a forest to do his “tapas”. Once he reaches the forest he realizes how comfortable his family life has been where he did not have to worry about food. He finds that it is a lot of work to gather and cook food for himself.

With determination he continues doing his “tapas”. One day he finds a wolf with both its front legs chopped off. To his amazement he finds that wolf looks well-fed. He wonders how the wolf has been able to hunt and fend for itself given its handicap. To find out how this is happening he hides himself behind a nearby stone and watches the wolf.

To his shock he finds that lion walks in to the place where the wolf stays. He thinks to himself that the wolf is dead. Surprisingly, the lion actually brings meat and feeds the wolf. Looking at the whole situation the Sannyasi infers that this is a message from God. God is asking him to focus on his tapas and forget about the silly worry of feeding himself.

Determined with the message from God, he completely ignores his hunger and just focusses on his tapas. Two days pass by and he is extremely hungy. Five days pass by and he is weak. Seven days pass by and he starts moaning. He does not have the energy to gather food and he cannot bear the hunger.

A Yogi who is passing by hears his moaning and takes pity on him. He feeds this moaning man and enquires about his situation The Sannyasi narrates the whole story of the wolf and the lion and the message from God. The Yogi responds: “It is definitely a message from God. But why did you choose to be the crippling fox instead of choosing to be the generous Lion”

Ravikanth Andhavarapu

What 800 nonprofit filings show about proving your own numbers

Revenue that moves without the business moving

Why this is not an accounting scandal

Why it matters more than it looks

The document nobody reads

What I think this says

How do we ensure every human can get a loan of 10USD?

One pager

Why?

What?

How?

What are the benefits this offers?

The most important thing that we need if you want to work with me

Getting started with phabricator

Hands On Introduction to Machine Learning

MySQL Cheat Sheet

How to reduce software development cost by upto 90 percent

The worst question to ask a non-technical person

Brutality is the key to execution

Key to execution

How to be brutal

Making execution easy

Criteria and Job Description for Choosing a Tech Lead and junior developer

Criteria for Choosing a Tech Lead

Criteria for Choosing a Junior Developer

My Favourite Story