Our Research Write-Ups Now Carry the Evidence They Rest On

The research shelf in our lab holds sixty-six write-ups. Fifty-five of them are built on a single peer-reviewed paper, one write-up per paper, and those papers are the ones listed in our research library. The other eleven are a different kind of thing: some pull together many studies at once on one dimension of a gift, some work through our own catalogue, and three of them document our own process, including two products we turned down.

All sixty-six were the plainest pages we publish.

You arrived at a title, then a table of filing information, then the article. On the fifty-five built on a paper, that paper sat in the table in small grey type next to a label. Who wrote it, where it was published and when were nowhere on the page at all.

The reason is worth telling, because it is not the reason anybody would guess. Last year we rebuilt every other kind of record in our lab: a header that tells you what you are looking at, a block at the foot holding everything you would need to cite us, a citation tool beside it. All of that was built once and switched on by a single test.

That test asked whether the record had a verdict. A verdict is what we give a product we tested. A research write-up examines evidence, so it passes and rejects nothing, and it has never had a verdict. The test had therefore been false on every one of these pages since the day it was written, and a switch that has always been off looks exactly like a switch somebody chose to leave off. Seven theme versions built blocks behind that switch, and every one of them went past it.

The test now asks what each block actually needs. The one block that prints a verdict still asks about the verdict, because that is the only thing it could honestly ask.

So every research page now carries a record at its foot, holding our own identifier for it and sitting directly above a citation tool that opens for you instead of waiting to be clicked. On the fifty-five built on one paper, that record also links to the paper itself and to its entry in our research library, the citation tool names the paper rather than naming us, and the page opens with one quiet line giving the authors, the journal and the year. On the other eleven none of that appears, because there is no single paper to name and we would rather show you nothing than show you a guess.

Fifty-seven of these pages carry a chart. Those charts now sit in our frame rather than in fifty-seven different ones, and a chart heading that could previously be set larger than the page title now sits below it, on every chart but one. And where a research page has neighbours examining the same dimension of a gift, it now ends with those instead of with the two write-ups we happened to publish either side of it.

One thing is deliberately not here yet. The opening sentence of each page should be the plain-language summary of what the research found, set large. That summary lives in a field which, on some records, holds a sentence written for a reader and on others holds a paper’s own title. Setting that large would hand some visitors a journal citation as the first thing they read. The field gets one job first.

The correction we published alongside this covers the two things we found wrong while doing it.