Home  /  How we know  /  Chain card

Fits on one sheet of Letter or A4. To keep a copy, print and choose “Save as PDF”.

Did it work? The chain card

Eight questions between a gift and “it worked”

Take any claim that a gift worked and walk it down these eight links, in order. Each link gets its own answer and its own level of certainty. Where nobody can answer, that is the weakest link, and it sets the headline.

How to read a claim with it

  • Report every link on its own. Never add them up into a score or count how many look strong.
  • The weakest link sets the headline: the first one, in order, whose evidence is unknown, contested or only the organisation’s own estimate.
  • A strong link never lifts a weak one.
  • Never fill a blank with a guess of your own. “Nobody knows yet” is an answer.

A method, not a score, and not a verdict on any charity. The reasoning behind each link is at diditwork.org/how-we-know#chain.

diditwork.org/chain-card

Worked through

Read a charity report with the chain

Seven questions, one page of glossy claims, about ten minutes. Here is the procedure run on a report we made up, so you can see what each step catches before you try it on a real one.

Illustrative: made-up numbers, not any charity's results

Brightwater Reading Circle (a made-up charity)

Our tutors transformed 2,400 children's lives last year.

  • 84% of the children who finished our programme were reading at grade level by June, against 62% in nearby schools without us.
  • An independent 2019 study at three of our first schools found reading scores rose.
  • This year we expand to 40 new towns.
  • Just $60 helps a child learn to read.
  • Next year we will report how many children we enrolled.
  1. Step one L0 Defined

    Restate the claim (L0). If you cannot, stop.

    Card Compared with what?

    Report “Transformed lives” cannot be checked. The second line can: reading at grade level by June, children who finished, against nearby schools without the programme.

    So Restate it as that second line and carry on. If the report had only the headline, you would stop here.

  2. Step two L2 Caused

    Ask "compared with what?" and "how was it tested?" (L2). Prefer the evaluator's evidence label.

    Card Who was chosen, and when?

    Report The comparison schools were picked by the charity, and the programme went to schools that asked for it. No lottery, no named design.

    So This is the charity’s own comparison. Schools that ask for tutoring may already be the keener ones, so the gap could be partly who signed up.

  3. Step three L3 Sized

    Ask how many independent studies and whether they agreed (L3).

    Card What is the range, not just the headline?

    Report One older study at three schools, reporting that scores “rose”, with no size and no range. This year’s figure has no range either.

    So One small study, and no size to shrink. Nobody can say how big the effect is yet.

  4. Step four L4 Travels

    Ask whether it fits your setting and the charity's scale (L4).

    Card Will it travel to where my money goes?

    Report The evidence comes from the first three schools; the money goes to 40 new towns.

    So Unknown. New towns mean new tutors, new schools and a charity running at many times its old size.

  5. Step five L5 Worth it

    Read the evaluator's figure in its own unit: modeled, not guaranteed (L5). Decide your own weights between kinds of good.

    Card What does one good outcome cost, and how sure is that?

    Report “Just $60 helps a child learn to read.” No evaluator has published a figure.

    So That is a fundraising price, not a cost per child who reads because of the programme. Don’t turn it into one, and don’t compare it with an evaluator’s modeled figure for another charity.

  6. Step six L6 Adds

    Ask what one more gift adds now (L6); consider the evaluator's fund.

    Card Would it have happened without my gift?

    Report Nothing on how much is already raised or what a further gift would pay for.

    So Unknown. Ask the charity what one more gift funds this year that would not happen otherwise.

  7. Step seven L7 Checked

    Write down what you expect and check in a year (L7).

    Card Will we know in a year?

    Report Next year’s report promises an enrolment count, which says how many came, not whether they read better.

    So Write down what would convince you: next June, everyone who started (not only finishers), against schools chosen the same way. Then check.

L1 Happened. The seven steps fold L1 into the others, so ask it as you go. Here the 84% counts only children who finished. How many started, and how did the ones who left do? Who and what is missing?

Weakest link: L2 Caused

Not established yet. The weakest link is the second step: the only comparison is the charity’s own, between schools that chose it and schools that didn’t. That is no reason to think the tutoring fails. It is a reason to put the open questions above to the charity before treating $60 as the price of a child who reads.

If you can’t get answers, giving through an evaluator’s fund hands these questions to people who ask them for a living (Part 7.5 of how we know).