This information is based on an evidence-based methodology

The ledger

The ledger contains 384 metric rows across 29 public projects. Each row records one year of one metric in its native unit: pounds of food, program visits, volunteer hours, kilograms of CO2e, dollars raised, etc. Projects have their own source and one of four confidence grades: Verified (published or audited), Probable (derived from a verified figure by a standard conversion), Estimated, or Modeled. The site refreshes from the ledger frequently. If a number changes here, it changed in the ledger first, in version history.

How a dollar figure is made

Modeled social value for a project is calculated:

value × multiplier × CPI factor = modeled social value (2026 USD)

I care about social return on investment (SROI). I track the projects that I work on over time so that I can ensure that what I'm working on is meaningful. SROI puts an outcome in dollars. You take a count of something that happened, like pounds of food delivered or hours volunteered, and multiply it by an evidence-based price for that outcome. Each price comes from a published source, and it applies the same way to every row that uses it.

That evidence-based outcome price is the multiplier in the formula above. Each metric gets exactly one, with a vintage year (the year its source estimated it) and a citation. I don't add growth assumptions or projections, or round anything up.

I restate every price in 2026 dollars using the Consumer Price Index for All Urban Consumers (CPI-U), the Bureau of Labor Statistics (BLS) measure of what urban households pay for goods and services. The CPI factor converts each price from its vintage year, referenced to May 2026 (335.123), so a 2019 price and a 2023 price can sit in the same total.

Why one pound of food carries more than one price

I measure the same 7,007,257 pounds of rescued food four ways on this ledger, and I price and sum three of them. Denver Food Rescue's grocery value is reported at zero, on purpose, and Fresh Food Connect's is still priced.

• Grocery Value of Food Distributed, the grocery value of the food, where it is reported only. Its multiplier is zero, so it adds nothing to any total.
• Wealth Created, the grocery value of the food, where it is still priced. $0.15 a pound after inflation adjustment, $1,018,485.
• Meals Provided, the meal the household didn't have to buy. $3.28 a pound after inflation adjustment, $22,990,498.
• Food Waste Prevented, the retail sales the seller never made. $2.88 a pound after inflation adjustment, $20,211,802.

Together the priced lenses come to $6.31 a pound and $44,220,786, about 47.3 percent of the portfolio. The cost per meal is household grocery spending, and ReFED's figure is foregone retail sales, built from point-of-sale data. And the grocery value is that same shelf price again, so they're all separate routes to one number, the retail price of the food. The strict test for double counting asks whether the same value gets counted twice for the same party, and pricing them together fails it.

I report all of them anyway, because each one answers a different question a reader might actually have, and collapsing them to one throws the other answers away. If you want the conservative reading of this portfolio, use the largest single lens by itself: $22,990,498 against $93,473,860.

CO2e Avoided prices the greenhouse cost, which ReFED reports separately from the financial value, so it stays out of the overlap. So does Health Care Costs Avoided, which prices a health improvement as a cost insurers avoid. That saving lands on a different payer from the household.

One pound of rescued food carries $12.32 of modeled value across five lines:

• Health Care Costs Avoided, $5.54 a pound. $38,847,237.
• Meals Provided, $3.28 a pound. $22,990,498.
• Food Waste Prevented, $2.88 a pound. $20,211,802.
• Wealth Created, $0.15 a pound. $1,018,485.
• CO2e Avoided, $0.47 a pound. $3,284,087.

Of that $12.32, the $6.31 above is the overlap, one value counted more than once. The rest lands on people other than the household, so it's additional value. That $12.32 shows how much this ledger rests on one physical pound, and you can check my arithmetic against the same 7,007,257 pounds every lens above is measured over. On the home page the pounds figure is bigger, because it adds the attributed slice of the Upcycled Certified program. None of these lenses was measured on that tonnage, so I leave it out here. And since Health Care Costs Avoided is recorded on Denver Food Rescue rows only, that venture's own stack runs slightly higher than this portfolio-wide figure.

Health Care Costs Avoided is the largest priced outcome here, and it has the lowest confidence grade on the page. Its multiplier comes from an association in the research between food insecurity and health care spending. No study here measures what rescued food does to a person's health. Booking it as an avoided cost assumes the association is causal, and that the food moves a person from food insecure to food secure. Neither one is established. If you want the most cautious reading of this site, set this line aside.

Denver Food Rescue records 1,340,344 visits to its no-cost grocery programs. Visits aren't people, so I divide them by 46.36, the organization's own 2023 ratio of visits to unique participants. That year it counted 241,186 visits from 5,203 people. What's left is 28,915 person-years. I keep 85 percent of those, the share of participants in Denver Food Rescue's survey who reported better health from eating more fruit and vegetables. Then I multiply by 0.80 for attribution, the provider-level fraction NEF Consulting used for FareShare, and by 0.75 for deadweight, since some of these households would have found produce somewhere else. What remains prices at $1,863.17 per person-year, from a 2018 study by Berkowitz and colleagues. That figure is in 2015 dollars. Adjusted for inflation, it's $2,634.37.

The visit count and the 46.36 divisor come from Denver Food Rescue's records, and the 85 percent comes from its survey. Attribution and deadweight are judgments. One comes from a UK study of a different organization, and the other is mine, with no published number behind it. If you doubt any of them, swap in your own and redo the multiplication.

The Upcycled Certified line

The Upcycled Food Association wrote the Upcycled Certified standard and launched the program in 2021. Member companies do the actual diverting. In 2021 and 2022, the two years the program existed while I ran the association, certified companies diverted 772,467 tonnes. The association sold the program to Where Food Comes From in 2024. Everything diverted after 2022 stays unpriced here.

I price one thousandth of that tonnage, which comes to 1,702,998 pounds. Those pounds price at $2.88 a pound after adjusting for inflation, the same ReFED figure I use for every rescued pound on this site, and they book avoided emissions at the same rate as the rest. Together the line comes to $5,710,334. The 0.001 is my judgment call, and no published figure backs it up.

In 2016, Garrett and colleagues measured the potential additionality of two farm certification standards at 0.14. Probst and colleagues found in 2024 that under 16 percent of issued carbon credits represent real reductions. Keep a tenth of that 0.14 for the association and you get 0.014. That's fourteen times my figure.

I go lower because certification changes behavior where applicants don't already comply, and verifies it where they do. Blackman and Naranjo made that case in 2012. A company applies to certify a product it already makes.

The standard's counting rule, Section 4.3.6, counts input weight, including moisture that gets removed later in drying or baking, so the tonnage runs larger than the food kept out of waste. Its own example diverts 35 tons of blueberry pomace, dries it down to 10 tons, and reports 35. On top of that, a certified ingredient's inputs count again inside a certified finished product, so one mass of food can show up under two certificates.

I use a single retention factor on this line, because splitting the 0.001 into two factors would deduct the same fact twice. The SROI Guide warns that attribution often sits inside a deadweight estimate, and it says to take care not to take off more than you should.

Published social return studies of food rescue put deadweight at 60 percent or below, an order of magnitude from where I sit. But those studies price the work of collecting and moving food, and this line prices a label on food somebody else diverted. Of the eighteen certification and convening bodies checked, none prices a share of its members' outcomes at all, which cuts against pricing this line in the first place.

At 0.005 this line would be $28,551,671, and at 0.014 it would be $79,944,678. The whole rest of this ledger models at $87,763,527.

What is never summed

• Dollars raised represents the money in to fund the projects. It shows up as its own figure and never enters modeled social value.
• My own footprint. For my personal CO2e I use the US average footprint for each ledger year plus a 10 percent margin, since I assume I consume more than the average American. I price it with the same social cost of carbon as the portfolio's avoided emissions and report it on its own page. The net figures subtract it from the portfolio's recorded modeled value: the Net SROI chart on the home page, and the net modeled total and ratio on the home and ledger strips. Every other figure on the site shows the portfolio before that subtraction.
• Rescued pounds carry value in more than one lens today: meals provided, retail sales the store never made, and the grocery value of the food itself. Those lenses measure different things and share the same pounds. I price meals and retail sales. The grocery value is reported at zero for Denver Food Rescue, deliberately, and still priced for Fresh Food Connect.
• Program-scale diversion, nearly all of it. Member companies do the diverting that the Upcycled Certified program measures. I report the full certified tonnage on the Upcycled Food Association's page as program reach and price one thousandth of the 2021 and 2022 tonnage. The rest prices nowhere.

The ratio

Modeled social value is stated in 2026 USD; dollars raised is stated at face value, per that metric's own methodology note. Dividing value by raised is the input-versus-outcome comparison the two figures exist for.

That works out to $93.5M of modeled value against $38M taken in, or about $2.46 of modeled value for every dollar in.

Across the portfolio, 91 percent of the money in that denominator is a nonprofit total-revenue line from a Form 990. That line doesn't split donations from program income, so I can't claim it as contributions, even though most of it very likely is. Another 7 percent is Fresh Food Connect's earned software revenue. One percent is capital committed to Small Capital's fund. Well under one percent is money I can name outright as a grant or a gift.

Internal rate of impact beta

SROI says how much value each dollar became. The internal rate of impact, or IRI, says how fast.

0 = ∑_t (V_t − I_t) ÷ (1 + IRI)^t

I_t is money in during year t, and V_t is the priced social value it produced. It is the internal rate of return investors use, with priced social value in place of cash.

Run on a nonprofit's books year by year, that equation breaks. Money comes in and value goes out every year, so it finds no rate, two rates, or a rate set by a few early dollars. So I follow each dollar to what it paid for.

• Programs. Money that runs a year of work buys that year's value, which I count a year after the money. Every program year is the same one-year bet, so a program's IRI is its multiple minus one. The programs are Denver Food Rescue, Fresh Food Connect, Small Capital, and Upcycled Food Association.
• Builds. Money that makes one lasting thing, like a film, a mural, or a book, buys every later year it keeps producing. I count each year's value in the year the ledger records it. A build still producing reads "so far," and one under 1x reads "not yet paid back." The builds are Colorado Poet Laureate Anthology, Durango Brewery Mural, Durango Poet Laureate Program, Durango Solar System Project, and Moving Line Documentary.
• Ideas. Ideas split the same way, on projections. A build idea's launch capital buys each year of projected value after launch, and I count ten years. A program idea's capital is one year of program money for one year of value.

The portfolio's IRI starts every project's clock at its own first dollar and solves for one rate across all of them. That works out to 106 percent a year.

IRI restates money in to 2026 dollars, the basis the value is already in, because a rate across two price bases means nothing. So a project's IRI can read lower than its ratio on this page suggests.

I color a rate gold under 7 percent a year and green at 7 percent or more. Seven percent is the upper rate federal agencies use to discount the future, and about a market return.

Every IRI is gross of the carbon it took to run each organization. IRI is in beta, so the method may still change.

Projections

Any public idea, built or not, can have a researched projection: a conservative annual steady state, stated per year and per metric. Each one is labeled a projection where it renders, and none of them enter the ledger's recorded totals, the leaderboard, or the ratio above.

For an idea that doesn't exist yet, I anchor capital to launch, projected annual value, and projected people served to named comparables. A built venture anchors to its own recorded track record: the average of its last few complete ledger years, carried forward flat unless a documented growth factor says otherwise. Ventures with a stated goal report the goal as a goal, with its target year and the per-year rate that falls out of the two. Reach goals stay unpriced where pricing them would overstate modeled value.

Every projected figure traces to a published comparable, to the venture's own ledger rows, or to a named anchor with a source, a vintage year, and a confidence grade. The first anchor I registered is per-viewer cultural value for film and media work: $5 per viewer. That's a market-price floor set at standard digital rental pricing, $3.99 to $5.99, and it's under half of Cinema United's 2024 average US movie ticket of $11.51. Five dollars is about an eighth of the recorded cultural-reach multiplier above, $38.50 per attendee at a 2022 vintage, because an idea that doesn't exist yet hasn't earned the recorded rate. If I can't anchor an idea one of these ways, it ships without numbers until the research exists.

Sources

IRS Form 990 filings

For the nonprofits in the portfolio, the primary financial record is public, so you don't have to take my word for the money.

Data refreshed Oct 15, 2026, 4:21 PM Mountain Time.

A note on what these numbers are

Every dollar figure on this site is a modeled estimate: a count of something that happened, priced with a published outcome price and restated in 2026 dollars. Nothing on this site is investment or financial advice or an audited statement of any organization's finances. I did my best.