The wall of shame. Lessons nobody should pay for twice.
Good intentions are not results. These are cases where money, hope and often real hard work went in, and the outcome was poor, unproven or harmful. We include them because the lessons are expensive and nobody should have to pay for them twice. We criticise decisions and programmes, not anyone's character, we note where organisations responded, and where the evidence is mixed we say so.
It didn’t work The promised result didn’t arrive.
Mixed results Some good, far less than promised, or the evidence is contested.
It did harm The evidence suggests people ended up worse off.
The entries
It didn't work
PlayPumps
2006–2010
Cellar noteA merry-go-round that pumped water. Charming on camera; hard work at the well.
What was promised
Children's play would pump clean water: spin the roundabout, fill the tank, and sell billboard space on the tower to pay for upkeep. In 2006 $16.4 million in support was announced to install pumps across southern Africa, with a plan to raise more than $60 million by 2010.1
What happened
Three years later FRONTLINE/World found pumps standing idle. Women said spinning was hard work without children around, that the old hand pumps had been easier, and that nobody had asked them; one pump had produced no water for six months, and calls to the repair line went unanswered. The rollout ended quietly.1
The evidence
A UNICEF evaluation in 2007, before the big rollout ended, had already raised problems with safety, low water output, installation quality and community accountability.21
How they responded
The rollout was wound down. The reporter whose 2005 story had helped launch the campaign went back three years later and reported what had gone wrong, which is to journalism's credit.1
How “Did it work?” would have caught it
We'd have asked the dullest question first: does it deliver more water, at lower cost, than the hand pump it replaces? It didn't, and the users could have told anyone who asked.
The lesson
Compare against the boring alternative, and ask the people who'll use it. Novelty is not an outcome.
It didn't work
Three Cups of Tea and the Central Asia Institute
1996–2012
Cellar noteA bestselling story about building schools, and a charity whose governance didn't match it.
What was promised
The Central Asia Institute was founded to build schools and bring education to remote parts of Pakistan and Afghanistan, a mission made famous by its co-founder's books.3
What happened
In 2012 Montana's Attorney General concluded that the board had failed to fulfil important responsibilities in governing the charity, and that the co-founder had failed in his duties as executive director and board member.3
The evidence
Under a settlement he agreed to repay more than $1 million to the charity, stepped down as executive director, and the board was to be rebuilt with at least seven members.34
How they responded
The charity accepted the settlement and its governance was restructured under the Attorney General's terms.3
How “Did it work?” would have caught it
We'd have looked past the story to the accounts and the board: how much went to schools, how many are operating and staffed, and who checks. A small board overseeing a famous founder is a warning light.
The lesson
A great story is a reason to look harder, not a substitute for looking. Governance is part of the outcome.
It didn't work
American Red Cross housing in Haiti
2010–2015
Cellar noteNearly half a billion dollars raised; six permanent homes built. Read the definitions before the headline.
What was promised
After the 2010 earthquake the Red Cross raised nearly half a billion dollars, and its chief executive spoke of developing "brand-new communities".5
What happened
A ProPublica and NPR investigation found the charity had built six permanent homes in all of Haiti, that a signature project meant to build hundreds of homes built none, and that a claim of providing homes to more than 130,000 people included people trained in construction and given temporary shelter or rental help.5
The evidence
The reporting drew on internal memos and evaluations, and noted that other groups, facing the same land-title problems, built thousands of homes.5
How they responded
The Red Cross said the six homes were a pilot, that land disputes and residents' wishes led it to shift to repairs, transitional shelters and rental subsidies, and that its work helped about 132,000 people.65
How “Did it work?” would have caught it
We'd have asked what "helped" means, counted permanent homes separately from training and temporary shelter, and noticed that a disaster-relief charity was promising long-term construction it had little experience of.
The lesson
Insist on plain definitions. When a headline number mixes very different things, the outcome you care about is usually the smallest part.
Mixed results
Kids Company
until 2015
Cellar noteReal need, real devotion, and a model that couldn't survive a bad month.
What was promised
A London charity supporting vulnerable children and young people through a demand-led service, which reported reaching 36,000 beneficiaries.7
What happened
It collapsed in 2015. The Charity Commission later found a "high risk business model": heavy dependence on grants and donations, reliance on one key fundraiser, low reserves and demand-led spending, and it made a formal finding of mismanagement over repeated failure to pay creditors, including its own staff and the tax authority, on time.78
The evidence
The Commission said the 36,000 figure's method "should have been clearly articulated wherever cited", and found records too thin to show that heavy spending on a small group of beneficiaries was justified.7
How they responded
In 2021 the High Court rejected an attempt to disqualify the trustees and chief executive, and the Commission agreed there was no dishonesty, bad faith or inappropriate personal gain. We record that plainly: this was a failure of model and oversight, not of honesty.7
How “Did it work?” would have caught it
We'd have asked how many children were helped, measured how, and what happens if a large funder walks away. Without a method behind the number, there's nothing to taste.
The lesson
Reach is not results, and reserves are not a luxury. A charity that can't survive a shock can't protect the people who depend on it.
Mixed results
The $100 million gift to Newark's schools
2010–2016
Cellar noteNot a disaster. A costly lesson in how long it takes to know.
What was promised
Mark Zuckerberg pledged $100 million to Newark's public schools to back a sweeping reform effort; with other donors, private funding reached $200 million.910
What happened
The reforms, launched in 2011 after $200 million in private donations, were wrenching for many communities, with school closures and an expansion of charter schools. Achievement growth fell at first in both English and maths.10
The evidence
A Harvard evaluation using student-level data found that by 2015-16 growth in English had improved significantly while maths stayed about where it had been; most of the English gain, 62 percent, came from students moving to more effective schools rather than schools improving.10
How they responded
Supporters point to the English gains; critics to the disruption. Both are in the evidence. We'd call this mixed, and we'd rather say that than pick a side.
How “Did it work?” would have caught it
We'd have asked, before the first dollar: what result, measured how, by when, compared with what? The independent answer arrived about six years in, which is an expensive way to find out.
The lesson
Big, fast, top-down gifts buy disruption first and evidence later. Decide in advance how you'll know if it worked.
It didn't work
One Laptop per Child
Cellar noteA beautiful machine. The test scores didn't move.
What was promised
A low-cost laptop for every child would transform learning in poor countries.
What happened
In Peru, the programme reached rural primary schools, and a large randomised evaluation in 318 schools found computers per student rose from 0.12 to 1.18 and children used them much more, at school and at home.11
The evidence
There was no evidence of effects on maths or language test scores after 15 months, and only inconclusive signs of better general cognitive skills.11
How they responded
To its credit, the programme was evaluated with a large randomised trial, which is why we know.
How “Did it work?” would have caught it
We'd have separated inputs from outcomes. Laptops delivered is an input. Children reading and counting better is the outcome, and it should have been tested small before it was bought big.
The lesson
Delivering a thing is not the same as changing a life. Count learning, not laptops.
It didn't work
TOMS one-for-one shoes
until 2019
Cellar noteA brilliant marketing idea in search of a problem to solve.
What was promised
Buy a pair, give a pair: every purchase would put shoes on a child who needed them.
What happened
A study of TOMS donations in rural El Salvador found the shoes were liked and worn, but most children already owned shoes, the programme did not reduce shoelessness, and overall impact was negligible. Recipients were slightly more likely to say outsiders should provide for their families.12
The evidence
Fair is fair: a companion study found no statistically significant damage to local shoe markets, though estimates hinted at modest substitution away from local purchases.13
How they responded
In 2019 TOMS replaced one-for-one with giving a third of its annual net profits to grassroots organisations.14 It also let researchers study the programme, which is how we know.
How “Did it work?” would have caught it
We'd have asked whether the children lacked shoes in the first place, and whether shoes were the thing standing between them and a better life.
The lesson
Start with the need, not the product. A gift that feels good to buy can still be a gift nobody needed.
It didn't work
The FTX Future Fund
2022
Cellar noteGrants made with money that turned out not to be what it seemed.
What was promised
A large new philanthropic fund, paid for by the business behind the crypto exchange FTX, would make ambitious grants to promising people and projects.15
What happened
In November 2022, as FTX collapsed, the fund's team resigned, saying they had fundamental questions about the legitimacy and integrity of the business operations funding it, and that many committed grants would likely not be honoured.15 FTX's founder was later convicted and, in 2024, sentenced to 25 years for fraud.16
The evidence
The team said they were "devastated" that many committed grants looked unlikely to be honoured, leaving grantees who had planned around the money exposed. In our view the episode also cost the wider effective-altruism movement, which had celebrated the donor, a great deal of trust.15
How they responded
The team resigned publicly, condemned any deception by FTX's leadership and apologised to grantees. Our criticism is of a donor-concentration risk the movement under-weighted, not of the grantees.
How “Did it work?” would have caught it
Our rule is to judge outcomes, and that cuts both ways: where the money comes from is part of whether a gift can be relied on. Concentrated funding from a single, opaque source is a risk to price in.
The lesson
Don't let one donor become the weather. Diversify funding, and do due diligence on where money comes from as well as where it goes.
It did harm
Scared Straight
Cellar noteDesigned to deter crime. On the evidence, it does the opposite.
What was promised
Taking young people at risk of offending into prisons to meet inmates would frighten them away from crime.
What happened
A Campbell systematic review of nine randomised or quasi-randomised studies in the United States, with 946 participants, found no evidence that these programmes reduce offending.17
The evidence
In the seven studies that reported reoffending, the programmes significantly increased the odds of offending. The reviewers concluded they cannot be recommended.17
How they responded
The review's authors advised that any agency still running such programmes should at least evaluate them rigorously.17
How “Did it work?” would have caught it
Only one way: a controlled comparison. Before-and-after stories looked convincing; the trials showed harm.
The lesson
Intuitive is not the same as true. Some programmes make things worse, and only a fair test will tell you which.
Mixed results
Dumping donated goods: clothes and food aid
Cellar noteThe evidence is strong for one case and contested for the other. We'll say exactly which.
What was promised
Shipping surplus clothes and food to poor countries would help people in need at little cost.
What happened
A study of sub-Saharan Africa estimated that imports of used clothing explain about 40 percent of the decline in apparel production and about half of the decline in apparel employment between 1981 and 2000.18
The evidence
On food aid the picture is less settled. One influential study linked US food aid to longer and more frequent civil conflict; later work argued its method can produce that result even where none exists. We treat the conflict link as unproven.1920
How they responded
Neither practice has stopped. Our view: buying locally or giving cash avoids undercutting the very producers a poor economy needs.
How “Did it work?” would have caught it
We'd have asked who loses when free goods arrive, and whether cash to the same people would do more.
The lesson
Give what people need, not what you have spare. When in doubt, cash respects both the recipient and the local market.
Fairness note: this page describes documented programmes and decisions, using the findings of courts, regulators, official reports, peer-reviewed research and established journalism, each linked below. It is not a judgment of anyone’s character. If an organisation listed here has changed course, or we have misread a source, we will correct the entry and date the change on our what’s changed page.
Sources
Every fact and figure above links here. We checked each source on 24 October 2026. Figures are as published by each source; estimates are theirs, not ours.