Summary
What do they do? The modeled pathway is one household receiving 26 weeks of access to a staff-led Community Market. SFMFB opened two markets in June 2025 and says it aims to operate eight markets serving up to 4,500 households weekly within three years. The markets use a grocery-style, client-choice model and connect participants with peer navigators and support services. More
Why this approach interests us
- Reliable client-choice food access can relax a household budget constraint, improve diet fit, and pair food with peer navigation and benefit support.
Our main reservations
- The causal anchor is a different bundled pantry intervention, current Community Market outcomes are unpublished, and network accounting does not identify a marginal market course cost.
What do you get for your dollar?
SFMFB reports serving 44,000 households weekly and nearly 56 million meals-equivalent in FY2025. Its Community Markets offer client choice and support, but SFMFB has not published a causal food-security estimate. Using an operating-cost allowance oriented by pantry accounts and heavily discounted external randomized evidence, we estimate about $6,000 per additional household not experiencing very low food security at 12 months. A separate health-utility transfer produces a very-low-confidence central estimate of about $59.6M per 10 QALYs.
- cost per better life
- $59.6M — per 10 QALYs; very-low-confidence food-security health-utility transfer
- positive-effect sensitivity
- $515.3K–$83.5B — per 10 QALYs; joint assumption scenarios, not confidence bounds; null or harm possible
- native outcome model
- ≈ $6,000 — per additional household not experiencing very low food security, before the separate donor-additionality discount
- funding room
- Not published — the $100,000 gift is a scenario, not a current marginal offer
1. What do they do?
The modeled pathway is one household receiving 26 weeks of access to a staff-led Community Market. SFMFB opened two markets in June 2025 and says it aims to operate eight markets serving up to 4,500 households weekly within three years. The markets use a grocery-style, client-choice model and connect participants with peer navigators and support services.
Enroll a household
A household enters one defined Community Market cohort; a recommendation-grade plan would publish eligibility, waitlist, baseline food security, and nonparticipation.
Provide client-choice food
Participants choose food in a grocery-style setting rather than receiving a fixed parcel.
Connect support
Peer navigators and partner services can help households access benefits and other resources.
Measure the outcome
Follow every eligible household for 12 months with the USDA 18-item food-security module and report retention, missingness, and a credible comparison.
Scope of this review. The model covers one 26-week Community Market household course. It excludes neighborhood pantry visits, home delivery, CalFresh outreach, food-bank-wide meals-equivalent, and organization-wide reach. Households—not individuals—are the outcome unit.
2. Monitoring and information sharing
Households reached through SFMFB's pantry network and two Community Markets
Audited accounts plus organization reporting. SFMFB reports 44,000 households served weekly and nearly 56 million meals-equivalent in FY2025. Audited Neighborhood Pantries expense was $101.97 million, including $75.54 million of donated food distributed.
Our assessment. This establishes scale and accounting context, not unique households, causal food-security gains, or the marginal effect of a donation. The audit narrative separately says more than 49,000 households weekly, so the denominator is not perfectly stable.
228 adults randomized to Freshplace or traditional pantry access in Hartford, Connecticut
Randomized controlled trial. Freshplace combined client-choice food access, monthly motivational interviewing, and referrals. Participants were reported as less than half as likely as controls to experience very low food security, with gains in self-sufficiency and fruit and vegetable intake.
Our assessment. This is the most relevant causal anchor we found, but it is not SFMFB evidence and does not isolate client choice from case management or referrals. Our 5-point best guess is a heavily discounted transfer, not an observed SFMFB effect.
Twelve food-bank-based interventions, including five randomized trials
Systematic review. The review found heterogeneous interventions and outcomes and did not pool results in a meta-analysis.
Our assessment. Food-bank interventions are not one uniform treatment. The evidence supports testing a defined Community Market pathway, not treating every meal or pantry visit as an equivalent outcome.
3. Qualitative assessment
Reliable client-choice food access can relax a household budget constraint, improve diet fit, and pair food with peer navigation and benefit support.
Key reservations
- SFMFB has not published a causal Community Market effect on food security, health, or well-being.
- Freshplace bundled client choice with motivational interviewing and referrals in another city and population.
- The 44,000-household annual-report denominator conflicts with the audit narrative's more-than-49,000 figure and may include recurring households across weeks.
- The donor-cash cost subtracts recognized donated food and services; it is not a total social-resource cost and does not value donated food at zero.
- Neighborhood Pantries accounts cover a broader network than the two Community Markets, which opened in June 2025.
- The eight-market expansion goal is not a published marginal funding gap or capacity curve.
- Public contracts, commodities, private gifts, and donated food interact; displacement for the next gift is unknown.
- The separate $85.7 million per 10-QALY estimate uses explicit judgment calls for the food-security state transition, adult allocation, observational utility transfer, and duration. None is an SFMFB measurement; the conditional range spans approximately $1.11 million to $24 billion, and a null has no finite positive upper bound.
Benefits not included in our estimate
- Reduced worry in the home-delivery survey because it has no comparison group
- CalFresh benefits accessed because benefit dollars are not equivalent to attributable welfare gain
- Health-care savings, diet quality, social connection, peer-navigation benefits, and avoided food spending
- Food-bank-wide reach because households can recur and Community Markets are only one delivery channel
- Child health, development, school attendance, caregiver relief, and within-household spillovers
- Diet quality, stress, mental health, medical spending, mortality, and benefits-enrollment gains not captured by the adult food-security utility transfer
- Food value, participant time, volunteer time, donated services, public commodities, and other social-resource costs outside the donor-cash boundary
- Benefits from neighborhood pantries, home delivery, CalFresh outreach, policy work, and other Food Bank programs
4. What do you get for your dollar?
Roughly $6,000 per additional household not experiencing very low food security—if the transferred effect is real.
We isolate approximately $25.8 million of FY2025 Neighborhood Pantries cash and recognized operating expense after removing recognized donated food and donated services, divide by 44,000 weekly households and 52 weeks, and model a 26-week course at $300. We then assign a 5-point absolute reduction in very low food security—far below the bundled Freshplace trial result—to reflect program and population transfer. The $300 is a judgmental donor operating allowance, not a verified cash-flow or marginal cost: the accounting subtraction still includes depreciation. It is not a total social-resource cost; the comparable full recognized accounting cost is about $1,160 per course.
A better life is our comparison unit of 10 additional quality-adjusted life years (QALYs), potentially spread across people. These are uncertain estimates, not measured returns or verified donation offers.
How we calculate the estimate
CONDITIONAL COST PER ADDITIONAL HOUSEHOLD NOT EXPERIENCING VERY LOW FOOD SECURITY: $300 ÷ 5%
= $6,000
Model inputs and assumptions
- Modeled cash cost per 26-week household course
- $300 (range: $200–$600). FY2025 audited Neighborhood Pantries expense of $101.97 million less $75.54 million of recognized donated food distributed and $0.66 million of donated services and supplies, divided by the annual-report denominator of 44,000 households served weekly and 52 weeks, then multiplied by 26 weeks and rounded. This cash-accounting boundary does not value donated food at zero and is not a Community Market marginal quote. low.
- Full recognized accounting cost per 26-week household course
- $1,160 (range: $1,040–$1,160). FY2025 audited Neighborhood Pantries functional expense divided by 44,000-49,000 weekly households, annualized and multiplied by 26 weeks. This includes recognized donated food and is a social-resource comparator, not the donor-cash numerator. low.
- Freshplace randomized relative reduction in very low food security
- 50% (range: 50%–50%). The 228-person Freshplace randomized trial reported intervention participants were less than half as likely as controls to experience very low food security. Freshplace bundled client choice with monthly motivational interviewing and referrals and was not an SFMFB program. moderate for Freshplace; indirect for SFMFB.
- MFI estimate of absolute 12-month very-low-food-security reduction
- 5% (range: 1%–15%). Judgmental transfer discounted well below the Freshplace result because SFMFB Community Markets differ in population, geography, service bundle, dose, and follow-up and publish no causal food-security estimate. The low bound is a small positive effect, not a guarantee; a null effect remains plausible. very low.
What would $100,000 buy in the conditional model?
- Favorable positive effect: 75 additional households. 500 modeled 26-week courses · $1.3K per outcome
- MFI best guess: 16.67 additional households. 333.33 modeled 26-week courses · $6K per outcome
- Small positive effect: 1.67 additional households. 166.67 modeled 26-week courses · $60K per outcome
Uncertainty. The numeric range is conditional on a positive effect. SFMFB has not published a causal Community Market food-security result, and Freshplace combined client choice with case management and referrals in another city. If Community Markets do not reduce very low food security, or a gift substitutes for existing cash or donated food, additional outcomes approach zero and no finite positive cost-effectiveness estimate exists.
Our current best estimate: $59.6M per better life (10 QALYs).
Our revised conditional best guess is about $59.6 million per10 incremental QALYs for Community Markets, not the entire Food Bank. Table2's .023 food-insecure-versus-secure utility contrast replaces the old .008 mixed-population policy contrast; a separate50% funding-additionality judgment is now explicit. Other central inputs remain USD300 per26-week course,5% exit probability,35% reaching the study's secure category, one adult-equivalent,50% utility transfer and half an effective year. This is not a measured SFMFB effect or verified marginal offer. It is not a leading health-only use of donations under this model; food consumption and dignity benefits are outside the QALY comparison.
EXPLORATORY COST PER 10 QALYS · ONE BETTER LIFE: $300 ÷ 0.000050312499999999996 QALY × 10
= $59.6M
QALY conversion assumptions
- Additional delivery caused by funding
- 50% (range: 10%–75%). Judgmental share of an additional funded course that expands service rather than replacing already funded operations. Separate from clinical transfer and the causal program effect. No current donor expansion offer verified. judgment.
- Modeled donor-cash cost per household course
- $300 (range: $200–$600). USD300 is a judgmental incremental operating allowance, oriented by legacy pantry expenses net of recognized donated food/services. That subtraction still contains depreciation and is not an exact cash flow or marginal Community Market quote. USD1,160 is a full recognized-accounting comparator, not a complete economic resource valuation. very low as a marginal price.
- Causal exit from very low food security
- 5% (range: 1%–15%). The native model heavily discounts the 228-person Freshplace randomized result for differences in population, geography, service bundle, dose, and follow-up. SFMFB has not published a causal Community Market result; null or harm remains outside the finite positive range. very low.
- Share of exits reaching the study food-secure category
- 35% (range: 10%–75%). With no Community Market transition matrix, the 35% share reaching the study's food-secure category is an MFI judgment. The study dichotomizes USDA adult-module0–2 affirmative responses as secure (including marginal) and3+ as insecure. Exiting very low food security need not cross this threshold; it is not necessarily high food security. judgmental.
- Adult-equivalent beneficiaries per household
- 1 (range: 0.5–2). The central model credits one adult-equivalent per household and excludes child and caregiver effects. The range reflects uncertainty about household composition and within-household benefit allocation; SFMFB publishes neither for the modeled cohort. judgmental.
- Observational insecure-versus-secure utility contrast
- 0.023 (range: Published 95% CI: 0.010–0.036). Table2 Scenario2 estimates SF-6D .804 if all food insecure versus .827 if all food secure: .023 (95%CI .010–.036). This observational counterfactual state contrast is not a measured Community Market effect. Annual QALYs require the utility difference to last a year. Scenario1's .008 compares a mixed status quo with all secure; its prose conversion to80/100,000 is arithmetically inconsistent, though .827−.819 supports .008. We use Scenario2 for modeled individual state transitions, with an explicit transfer judgment. very low as a causal transfer.
- Share of observational utility effect retained
- 50% (range: 25%–100%). Retain50% of the observational .023 state contrast for residual confounding and SF population/program transfer. This does not include donor funding displacement, which is modeled separately; adult allocation and duration have their own inputs. judgmental transfer.
- Effective utility duration
- 0.5 year (range: 0.25–1 year). The midpoint allows for gradual onset during the 26-week course and limited persistence to the 12-month endpoint. The timing and persistence of a Community Market food-security change have not been measured. judgmental.
- Optimistic positive transfer: $515.3K. $200 per household course · 0.0038812499999999993 QALY each
- Central transfer-discounted estimate: $59.6M. $300 per household course · 0.000050312499999999996 QALY each
- Cautious positive transfer: $83.5B. $600 per household course · 7.1875e-8 QALY each
The numeric range is conditional on a positive Community Market effect, some modeled exits reaching the study food-secure category (including marginal security: USDA adult 0–2 affirmative responses), a positive adult health-utility relationship, and donor additionality. SFMFB has not published a causal Community Market result or a current marginal funding plan; the health-utility anchor is observational. If the program has no causal effect, leaves every household food insecure, causes harm, produces only transient change outside the assumed duration, or a gift displaces public funding, commodities, or existing private support, the impact price has no finite positive upper bound.
5. Funding and previous grants
SFMFB has not published a costed site-by-site expansion gap, household enrollment and retention, pre-specified food-security follow-up, or evidence that the next private gift creates additional service rather than replacing cash, commodities, or donated food.
The $100,000 scenario is illustrative. SFMFB describes a goal of eight Community Markets serving up to 4,500 households weekly within three years, but does not publish a current site-by-site budget, committed funding, remaining gap, waitlist, staffing constraint, partner displacement, or outcome forecast for the next gift.
This review does not establish a verified marginal funding offer or a complete history of grants.
Opens the organization’s giving page. A general donation may not fund the specific activity modeled here; confirm allocation with the recipient.
6. Sources
- FY2025 Annual Report. San Francisco–Marin Food Bank. organization-reported. Published: 2025; retrieved: 2026-08-31.
- FY2025 audited financial statements. San Francisco–Marin Food Bank / Baker Tilly. independent audit. Published: 2025-12-11; retrieved: 2026-08-31.
- Community Markets. San Francisco–Marin Food Bank. organization program description. Published: Publication date not stated; retrieved: 2026-08-31.
- A randomized controlled trial of a food pantry intervention. American Journal of Preventive Medicine. peer-reviewed randomized trial. Published: 2013; retrieved: 2026-08-31.
- Food bank-based interventions and health outcomes: a systematic review. BMC Public Health. peer-reviewed systematic review. Published: 2023; retrieved: 2026-08-31.
- Eliminating Food Insecurity in the USA: a Target Trial Emulation Using Observational Data to Estimate Effects on Health-Related Quality of Life. Journal of General Internal Medicine. peer-reviewed observational target-trial emulation. Published: 2023; retrieved: 2026-09-07.
- Food Security in the U.S.: Measurement. USDA Economic Research Service. official measurement definition. Published: Updated 2026; retrieved: 2026-09-01.
Annual expenses: years and sources
Average annual expenses (three consecutive fiscal years): $170,781,819. Organization size is separate from the modeled cost-effectiveness of a donation.
San Francisco Food Bank
Whole legal entity Form 990 Part IX total functional expenses; latest original reconstructed filings checked, fiscal-end year from printed reporting dates. Older fallback rows from IRS extracted API.
Latest original filings checked; ProPublica extracted API lags these original returns. Accounting expense includes donated food; not cash-only budget.