← Cost-effectiveness annex  ·  The Water for Health Fund

Asili · DR Congo

Treated water kiosks in Bukavu and South Kivu: the catalytic-capital worked example, and the portfolio's hardest integrity question.

Asili's kiosks have covered their own operating costs from water sales since 2019, so the honest cost-effectiveness question is not "what does delivery cost" but what does one-time rehabilitation capital buy. This page models exactly that, and it also puts the portfolio's highest credit yield under the integrity lens rather than around it.

PRELIMINARY · v1 · AUGUST 2026 · PHILANTHROPIC-COST FRAME · NOT COMPARABLE TO FULL-DELIVERY-COST FIGURES WITHOUT THE LABEL
01The Program

Kiosks that pay for themselves, capital that doesn't exist locally

What Asili is

A Congolese social enterprise, majority-owned by the Eastern Congo Initiative, operating staffed water kiosks selling treated water in Bukavu and surrounding South Kivu. 69 kiosks operate today, serving roughly 80,000 people; the program acquired a majority of Mercy Corps' Bukavu water system as that funder exited, and the rehabilitation of the 91-kiosk Mazigiro network is underway. The program targets 650,000 people served by 2029. Kiosk water sales have covered full operating costs since the first quarter of 2019.

What the carbon advance does

Virridy prefinances 200,000 carbon credits at $5 per credit, paid across five milestone tranches: $1,000,000 that covers 83.4% of the $1,199,450 Mazigiro rehabilitation budget, with Asili funding the remainder. If DR Congo grants a national Article 6 authorization, deliveries earn a further $10 per credit. The carbon mechanism is expansion capital for a self-sustaining operation, and the model treats it that way rather than claiming credit for the whole system.

Frame discipline: every $/DALY on this page divides by the $1.2M rehabilitation budget, the philanthropic and catalytic cost. Ongoing delivery is beneficiary-financed through tariffs and is excluded, which makes these figures structurally more favorable than a full-delivery-cost model like the Rwanda page's $6,822/DALY. The two frames answer different questions and are never to be compared without this label.

02Results

Catalytic dollars, community-wide health

Unlike the school program, kiosks serve whole communities including children under five, so the randomized mortality-relevant evidence applies and the health value per dollar is far higher.

$355
philanthropic cost per DALY averted, raw: the $1.2M rehab budget against 3,375 DALYs over 2025–2033 at a conservative 12% mortality effect. Evidence-adjusted: $1,133 (composite 0.314).
$196
per DALY under the meta-analytic 25% mortality reduction (Kremer scenario): 6,120 DALYs, 3.60× the cash benchmark on moral-uncertainty weights.
1.99×
cash-benchmark multiple (moral-uncertainty weighted) in the conservative base case; stable across all three moral-weight frameworks.
$1.08/t
philanthropic cost per credited tonne (1.11M credits projected 2025–2033), 172× below the social cost of carbon.
0.47
credit integrity ratio, the portfolio's lowest: baseline residual 0.55 (hard flag: the 0.243 credits/person-year yield predates the 2026 methodology reconstruction that cut the Rwanda program's yield by ~45%) × delivery 0.85.
$2.30/t
integrity-adjusted cost per tonne, still 80× below the SCC. Even at integrity 0.35, $3.08/t (60×).

Why the health economics are so strong here: the denominator is one-time catalytic capital rather than a decade of delivery cost, and the beneficiary population includes under-fives, where diarrheal mortality concentrates. The honest caveats run the other way and are itemized in the worksheet: nearly every health input is a flagged judgment pending country-data lookups, the population target is unresolved (650,000 program target vs a 150,000 crediting cap in the current purchase agreement), and the credit yield awaits methodology reconstruction.

03The Worksheet

The entire model, cell by cell

The complete v1 workbook: every input with its provenance, every calculation, every scenario.

Provenance key: Proforma input Evidence-anchored Judgment · flagged Sensor-measured Formula
↓ Download the workbook (.xlsx)
A · Program series, 2025–2033 (Virridy carbon proforma, base scenario)
Series202520262027202820292030203120322033Total
Population served80,000222,500365,000507,500650,000650,000650,000650,000650,0004,425,000 person-yrs
Credits (tCO₂e)05,94010,89052,470169,290218,790218,790218,790218,7901,113,750

The proforma carries the population endpoints (80,000 today; 650,000 by 2029); the 2026–2028 ramp is linearly interpolated here and flagged. Prefinance tranches: $175,000 / 35,000 credits, $325,000 / 65,000, $250,000 / 50,000, $200,000 / 40,000, $50,000 / 10,000. Contracted backstop: 142,560 credits (scenario S5). Implied yield 0.252 credits per person-year, consistent with the booked 0.243.

B · Cost basis: one denominator, philanthropic frame
Virridy carbon advance$1,000,000Proforma input 200,000 credits at $5, five milestone tranches; 83.4% of the rehab budget
Asili share$199,450Proforma input
TOTAL_PROGRAMME_COST (Mazigiro rehabilitation)$1,199,450Formula the single denominator; $0.271 per beneficiary-year

Excluded, by design and stated on every figure: ongoing operating cost (beneficiary-financed through kiosk tariffs since Q1 2019) and pre-2025 capital. This is what makes the frame catalytic rather than full-delivery-cost. The build fails programmatically if any downstream figure divides by a different cost.

C · Health chain inputs
InputValueProvenanceBasis
Under-five population share18%Judgment · flaggedDRC demographic profile; census data pending.
Under-five all-cause death rate14 / 1,000 child-yrsJudgment · flaggedDRC elevated relative to region; UN IGME lookup pending.
Diarrheal share of under-five deaths9%Judgment · flaggedGBD cause-fraction lookup pending.
Mortality reduction from safe water12%Evidence-anchoredConservative (GiveWell-style); the Kremer et al. meta-analytic 25% runs as scenario S1.
Years of life lost per under-five death33Evidence-anchoredStandard discounted YLL.
Diarrheal incidence, under-five / over-five0.9 / 0.35 per yrJudgment · flaggedGBD lookups pending.
Morbidity reduction, piped treated supply30%Evidence-anchoredWolf et al. 2022 (Lancet).
Delivery effectiveness (kiosk water safe)85%Judgment · flaggedSite-level continuous monitoring build-out pending; the portfolio's pooled monitored-safe rate is 90%.
Household use share of kiosk water75%Judgment · flaggedNo household survey yet. Effective coverage = 0.85 × 0.75 = 0.64.
YLD per diarrheal episode0.002215 DALYFormula0.188 disability weight × 4.3 days ÷ 365.
D · Calculation chain
StepValueComputation
Under-five person-years796,5004,425,000 × 18%
Baseline under-five diarrheal deaths in window1,004796,500 × 0.014 × 9%
Deaths averted76.81,004 × 12% × 0.64 effective coverage
Mortality DALYs (years of life lost)2,53476.8 × 33
Baseline diarrheal episodes, all ages1,986,825796,500 × 0.9 + 3,628,500 × 0.35
Episodes averted379,980× 30% × 0.64
Morbidity DALYs (years lived with disability)842× 0.002215
Total DALYs averted, raw3,375mortality dominates 3:1
$ / DALY, raw (philanthropic)$355$1,199,450 ÷ 3,375
Evidence adjustment (section E)× 0.3143,375 → 1,058 DALYs adjusted
$ / DALY, evidence-adjusted (philanthropic)$1,133$1,199,450 ÷ 1,058
xCash, GiveWell / egalitarian / wellbeing1.93× / 0.84× / 3.02×DALY moral weights 2.3 / 1.0 / 3.6; benchmark 0.00335 units/$
xCash, moral-uncertainty weighted1.99×0.45 / 0.25 / 0.30 prior; ranking stable across frameworks

Conversion note: valuing a death averted as 33 YLL × 2.3 units per DALY yields about 76 units, slightly below the ~117 GiveWell assigns an under-five death directly, so the conversion is mildly conservative.

E · Evidence-adjustment pipeline
FactorScoreBasis
Internal validity0.70No program-specific outcome study; the chain is assembled from literature values.
External validity0.70Meta-analytic evidence transferred to the Bukavu kiosk context.
Publication bias0.80Systematic-review base.
Implementation quality0.80Operating and self-financing since 2019; continuous monitoring build-out pending.
Composite0.314Formula product of the four; reported alongside raw, never silently multiplied in.
F · Credit integrity and carbon metrics (the headline issue)
Booked credit yield0.243 / person-yrHard flag The portfolio's highest, fNRB-adjusted, and it predates the Gold Standard SDWS V2.0 reconstruction, which cut the Rwanda program's yield by roughly 45%. Reconstruction pending.
Baseline-practice residual0.55Judgment · flagged Prices the reconstruction risk explicitly rather than waiting for it.
Delivery effectiveness0.85The same cell the health chain uses. One delivery number, two consumers.
Credit integrity ratio0.47Formula 0.55 × 0.85 × 1.00 leakage
$ / tonne, credited basis$1.08$1,199,450 ÷ 1,113,750 t · 172× below the $185 SCC (philanthropic cost basis)
$ / tonne, integrity-adjusted$2.30÷ (1,113,750 × 0.47) · 80× below the SCC
Self-sustainabilitytariff-fundedKiosk sales have covered operating costs since Q1 2019; carbon is expansion capital, not survival. This strengthens sustainability and sharpens the additionality question, and the model says both.
G · Scenarios (one-way; a preliminary model carries no Monte Carlo by design)
ScenarioVaried inputDALYs$ / DALY (phil.)xCash GWxCash MUReading
Base case3,375$3551.93×1.99×conservative 12% mortality effect
S1 · Kremer meta-analytic mortality25%6,120$1963.50×3.60×the upper evidence anchor
S2 · integrity low0.35$3.08/t60× SCCcarbon-side; health unchanged
S3 · integrity high0.70$1.54/t120× SCCcarbon-side
S4 · population at crediting cap150,000976$1,2290.56×0.57×the 150K purchase-agreement cap vs the 650K program target; the unresolved discrepancy, priced
S5 · contracted credit book only142,560 cr$8.41/t22× SCCcarbon-side floor: even the contracted book alone clears the SCC test by 22×
S6 · Article 6 upside counted as cost+$2.0M3,375$9480.72×0.74×the +$10/credit LOA payment on the 200,000 advance, if scored as philanthropic cost; in practice it is funded from credit value on delivery

Recomputed robustness statement: the climate finding is robust everywhere (10× below SCC at worst). The health finding spans $196 to $1,229 per DALY across scenarios, and the single input that most moves it is the population question, which the purchase agreement itself will resolve before the fourth tranche.

04Critique

The model's critique of itself

DimensionRatingAssessment
Effect-size qualityWEAKNo program-specific outcome study; every health input except the two literature effect sizes is a flagged judgment pending DRC data lookups. Preliminary status is required, and the composite adjustment (0.314) says so quantitatively.
Counterfactual robustnessMODERATEHealth-side funging is low: the prior institutional funder exited and Asili absorbed its system. The sharper question is carbon-side additionality: tariffs already fund operations, so the advance finances expansion, not survival. It covers 83.4% of a budget with no evident alternative financier, but v2 must test that claim rather than assume it.
Moral-weight transparencySTRONGAll frameworks reported; ranking stable; the death-to-DALY conversion is mildly conservative relative to GiveWell's direct valuation.
Attribution and integrityFLAGAttribution is clean (the cost basis includes both parties' shares). Integrity is the headline issue: the 0.243 yield is the portfolio's highest and pre-reconstruction; the 0.47 integrity ratio prices that risk, and the 150,000 vs 650,000 population discrepancy is carried as a scenario, not a footnote.
Excluded factorsFLAGExcluded: tariff-financed operating cost (the frame choice, labeled on every figure); time savings; non-diarrheal waterborne disease; pre-2025 capital. The frame itself flatters $/DALY relative to full-delivery-cost models, which is why the label is mandatory.
Ranking stabilityMODERATEClimate: robust in every scenario. Health: $196 to $1,229 per DALY; the population target is the decisive input, and it resolves contractually before Tranche 4.