Lesson 4: Adjustments — from gross value to counterfactual impact
You’ve estimated a value and a probability. Multiplying them gives you a raw expected-value number. But a raw EV overstates your impact, because it assumes that without your intervention none of this would have happened. In reality, you’re operating in a world where other people are also trying to change things — and where changes come and go.
Three adjustments turn a raw EV into something honest: credit allocation, speed-up, and persistence.
Credit allocation
Corporate campaigns don’t happen in a vacuum. Other NGOs run similar campaigns. Media coverage shifts public opinion. Consumer preferences evolve. Regulators create pressure. If a company adopts a welfare policy, can Feather Forward claim full credit? No — and pretending otherwise inflates your estimate.
Coefficient Giving handles this with explicit credit allocation:
For national wins (policies adopted in one country), Feather Forward gets 80% of the responsibility. Other advocacy groups, media coverage, and background trends account for 20%.
For global wins (multinational companies changing policies worldwide), Feather Forward gets only 10%. Why? Because organisations in many other countries also pressured those companies.
These numbers are judgement calls. You could argue for 70% or 90% for national wins. But making the estimate explicit forces you to think about it — rather than implicitly claiming 100%.
The framework distinguishes two levels:
Responsibility is divided among actors — the organisations doing the work.
Credit is divided among funders. If Coefficient provides 50% of Feather Forward’s budget and other donors provide 50%, Coefficient gets 50% of whatever Feather Forward achieves.
Coefficient’s share = Feather Forward’s responsibility × Coefficient’s share of Feather Forward’s funding.
So if Feather Forward has 80% responsibility for a national win, and Coefficient provides 50% of their funding: 80% × 50% = 40%.
The same logic scales down to your €500 donation. Your share of Feather Forward’s credit is roughly (Feather Forward’s responsibility) × (your donation ÷ their annual budget). For most individual donations the second number is tiny, which is fine — it just means you’re claiming a small, honest slice rather than an inflated one.
Speed-up
Even if your intervention works, you usually aren’t creating change that wouldn’t otherwise happen. You’re accelerating it. If broiler welfare standards would have improved in 2035 without intervention, and Feather Forward’s campaign makes it happen in 2030, that’s 5 years of impact. You count the benefit for those 5 years, not forever.
Coefficient’s default for policy change is 5 years of speed-up. This reflects a belief that most good policies will be adopted eventually — philanthropy just moves them forward.
If you think your cause area is an exception (because no one else would have done this for decades, or because the window is closing), adjust. But make the adjustment explicit.
Persistence
Even changes that happen can fade. A government budget allocation can be cut by the next administration. A regulation can be rolled back. A corporate commitment can be quietly dropped when media attention moves on.
Persistence asks: once the change happens, how long does it actually stick? That number tells you how many years of impact you can honestly count. Coefficient’s working assumption is that most policy wins persist for several years before they start to decay or get overtaken — but cause areas and political contexts differ, so adjust explicitly.
(A related move some BOTECs make is an implementation adjustment — accounting for the gap between “they said yes” and “they actually did it.” Feather Forward’s real-world analogue finds only 40–60% of commitments are fully implemented. That’s really a probability adjustment, not a persistence one, and it fits inside the conditional decomposition you saw in Lesson 4: P(success) = P(commitment) × P(full implementation | commitment).)
The full formula
With adjustments, your BOTEC looks like this:
Cost-effectiveness = (Value × Probability × Credit × Years of impact) / Cost
Each number is a judgement call. The BOTEC doesn’t give you certainty. It gives you:
Transparency — others can disagree with specific inputs rather than the whole conclusion.
Learning — if you predicted 30% and it didn’t happen, you can ask why. If you just said “promising,” there’s nothing to update on.
Comparability — you can compare grants across different cause areas using the same framework.
See it done for real
Now read how Coefficient actually evaluated the broiler welfare grant — numbers, adjustments, and all.
Read: The broiler chicken welfare section (2.4) of How We Use Back-of-the-Envelope Calculations — Coefficient Giving (15 mins)
Notice:
How they estimate animals impacted (Fermi-style decomposition)
Their base rate assumptions for probability of success
The credit adjustments for national vs global wins
The implementation adjustment
What they quantify and what they leave unquantified — and why